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,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,4,balanced,0.029674666623274486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,4,balanced,0.030975999931494396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,4,balanced,0.032986665765444435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,4,balanced,0.03751999884843826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,4,balanced,0.04173333446184794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,4,balanced,0.04394666850566864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,4,balanced,0.04538666705290476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,4,balanced,0.04398933549722036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,4,balanced,0.04404266675313314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,4,balanced,0.04577066500981649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,4,balanced,0.046069333950678505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,4,balanced,0.048170665899912514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,4,balanced,0.047824000318845115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,4,balanced,0.049914668003718056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,4,balanced,0.05622399846712748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,4,balanced,0.05602666735649109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,4,balanced,0.05685866872469584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,4,balanced,0.06609599788983662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,4,balanced,0.07034666836261749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,4,balanced,0.09306666254997253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,4,balanced,0.11463466286659241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,4,balanced,0.15026666720708212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,4,balanced,0.18790932496388754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,4,balanced,0.26152000824610394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,4,balanced,0.32547199726104736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,4,balanced,0.47171199321746826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,4,balanced,0.6115573247273763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,4,balanced,0.766266663869222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,4,balanced,1.1901386578877766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,4,balanced,2.2990293502807617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,4,balanced,0.04786133269468943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,4,balanced,0.04823466638724009
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,4,balanced,0.04969066878159841
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,4,balanced,0.059104000528653465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,4,balanced,0.07902400195598602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,4,balanced,0.1125920017560323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,4,balanced,0.1411733329296112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,4,balanced,0.14230400323867798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,4,balanced,0.14308266838391623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,4,balanced,0.1458560029665629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,4,balanced,0.14072533448537192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,4,balanced,0.14382933576901755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,4,balanced,0.14753066500027975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,4,balanced,0.15034666657447815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,4,balanced,0.15938666462898254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,4,balanced,0.1613759994506836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,4,balanced,0.17298134167989096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,4,balanced,0.19200533628463745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,4,balanced,0.21452800432840982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,4,balanced,0.24649065732955933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,4,balanced,0.27723199129104614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,4,balanced,0.3421386480331421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,4,balanced,0.430618683497111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,4,balanced,0.5634986559549967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,4,balanced,0.7265333334604899
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,4,balanced,1.0117759704589844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,4,balanced,1.2855573495229085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,4,balanced,1.6062506039937336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,4,balanced,2.421189308166504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,4,balanced,4.703839937845866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,4,power_law_1.2,0.032979199290275575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,4,power_law_1.2,0.030425599217414855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,4,power_law_1.2,0.031488001346588135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,4,power_law_1.2,0.03592959940433502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,4,power_law_1.2,0.03815680146217346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,4,power_law_1.2,0.048640000820159915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,4,power_law_1.2,0.05408639907836914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,4,power_law_1.2,0.04665600061416626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,4,power_law_1.2,0.05488640069961548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,4,power_law_1.2,0.052076798677444455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,4,power_law_1.2,0.05493119955062866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,4,power_law_1.2,0.06280959844589233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,4,power_law_1.2,0.05678719878196716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,4,power_law_1.2,0.056857597827911374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,4,power_law_1.2,0.07423359751701356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,64,power_law_1.01,0.07727360129356384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,4,power_law_1.2,0.06021760106086731
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,4,power_law_1.2,0.10129280090332031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,4,power_law_1.2,0.06104320287704468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,4,power_law_1.2,0.10944639444351197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,4,power_law_1.2,0.06384000182151794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,64,power_law_1.01,0.0468095988035202
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,4,power_law_1.2,0.111571204662323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,4,power_law_1.2,0.067603200674057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,4,power_law_1.2,0.11393280029296875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,4,power_law_1.2,0.07300480008125305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,64,power_law_1.01,0.04429439902305603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,4,power_law_1.2,0.07616639733314515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,4,power_law_1.2,0.11608959436416626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,4,power_law_1.2,0.07692800164222717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,4,power_law_1.2,0.12067199945449829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,64,power_law_1.01,0.04863359928131104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,4,power_law_1.2,0.08962560296058655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,4,power_law_1.2,0.1245695948600769
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,4,power_law_1.2,0.09761919975280761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,4,power_law_1.2,0.127839994430542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,64,power_law_1.01,0.04834559857845307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,4,power_law_1.2,0.12204159498214721
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,4,power_law_1.2,0.1324031949043274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,4,power_law_1.2,0.1297727942466736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,4,power_law_1.2,0.14045439958572387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,64,power_law_1.01,0.04830079972743988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,4,power_law_1.2,0.18215680122375488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,4,power_law_1.2,0.14552320241928102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,4,power_law_1.2,0.22867200374603272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,4,power_law_1.2,0.1584831953048706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,64,power_law_1.01,0.049830400943756105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,4,power_law_1.2,0.28399999141693116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,4,power_law_1.2,0.18172160387039185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,4,power_law_1.2,0.4063744068145752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,64,power_law_1.01,0.05466880202293396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,4,power_law_1.2,0.1980736017227173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,4,power_law_1.2,0.23752961158752442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,4,power_law_1.2,0.5863743782043457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,64,power_law_1.01,0.05423359870910645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,4,power_law_1.2,0.265881609916687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,4,power_law_1.2,0.8294591903686523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,4,power_law_1.2,0.3426176071166992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,64,power_law_1.01,0.05492479801177978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,4,power_law_1.2,0.4563392162322998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,4,power_law_1.2,0.9503935813903809
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,64,power_law_1.01,0.052883201837539674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,4,power_law_1.2,0.5939583778381348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,64,power_law_1.01,0.05689600110054016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,4,power_law_1.2,0.801632022857666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,4,power_law_1.2,1.4906304359436036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,4,power_law_1.2,1.0065152168273925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,64,power_law_1.01,0.05713919997215271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,64,power_law_1.01,0.060166400671005246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,4,power_law_1.2,1.478502368927002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,4,power_law_1.2,3.0699392318725587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,64,power_law_1.01,0.06578559875488281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,4,power_law_1.2,1.8088703155517578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,64,power_law_1.01,0.06809599995613098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,64,power_law_1.01,0.07052159905433655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,4,power_law_1.2,2.7213504791259764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,64,power_law_1.01,0.0800320029258728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,64,power_law_1.01,0.08926720023155213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,64,power_law_1.01,0.10474239587783814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,4,power_law_1.2,6.640991973876953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,64,power_law_1.01,0.11544959545135498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,64,power_law_1.01,0.1437440037727356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,64,power_law_1.01,0.15828479528427125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,64,power_law_1.01,0.21861119270324708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,64,power_law_1.01,0.2669440031051636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,64,power_law_1.01,0.4103104114532471
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,8,power_law_1.2,0.06947199702262878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,64,power_law_1.01,0.4583615779876709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,8,power_law_1.2,0.04686720073223114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,8,power_law_1.2,0.048767998814582825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,8,power_law_1.2,0.04935039877891541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,64,power_law_1.01,0.5671167850494385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,8,power_law_1.2,0.059385597705841064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,4,4,balanced,0.024351999163627625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,4,4,balanced,0.025216000775496166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,4,4,balanced,0.025514667232831318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,8,power_law_1.2,0.0787775993347168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,4,4,balanced,0.024933333198229473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,4,4,balanced,0.0313226655125618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,8,power_law_1.2,0.08353279829025269
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,4,4,balanced,0.03172266731659571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,4,4,balanced,0.031285333136717476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,4,4,balanced,0.033413333197434746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,4,4,balanced,0.033333333830038704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,4,4,balanced,0.03342933456103007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,4,4,balanced,0.03417599946260452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,64,power_law_1.01,0.8764479637145997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,4,4,balanced,0.035760000348091125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,4,4,balanced,0.036746665835380554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,4,4,balanced,0.03751999884843826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,4,4,balanced,0.041840001940727234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,4,4,balanced,0.04363733530044556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,4,4,balanced,0.04574933151404063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,4,4,balanced,0.05194666484991709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,4,balanced,0.059994667768478394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,4,balanced,0.07005333403746287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,4,balanced,0.07642666498819987
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,4,balanced,0.09762666622797649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,8,power_law_1.2,0.08328319787979126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,4,balanced,0.11347732941309611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,4,balanced,0.14777066310246786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,4,balanced,0.1759679913520813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,4,balanced,0.24879467487335205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,4,balanced,0.3139359951019287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,4,balanced,0.3821866512298584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,8,power_law_1.2,0.0846015989780426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,4,balanced,0.5798506736755371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,4,balanced,1.1126666863759358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,8,power_law_1.2,0.08689280152320862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,8,power_law_1.2,0.09082239866256714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,8,power_law_1.2,0.09023360013961793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,64,power_law_1.01,1.5472000122070313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,8,power_law_1.2,0.08687360286712646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,8,power_law_1.2,0.08958719968795777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,8,power_law_1.2,0.0950655996799469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,8,power_law_1.2,0.09553279876708984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,8,power_law_1.2,0.09993600249290466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,8,power_law_1.2,0.11387519836425782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,8,power_law_1.2,0.13230079412460327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,8,power_law_1.2,0.1503872036933899
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,8,power_law_1.2,0.17210880517959595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,8,power_law_1.2,0.20921599864959717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,8,power_law_1.2,0.24293119907379152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,8,power_law_1.2,0.4096320152282715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,8,power_law_1.2,0.45493121147155763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,8,power_law_1.2,0.5898303985595703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,8,power_law_1.2,0.962003231048584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,8,power_law_1.2,1.071673583984375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,8,power_law_1.2,1.9963071823120118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,4,power_law_1.01,0.08908799886703492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,4,power_law_1.01,0.11450239419937133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,8,power_law_1.2,3.9707710266113283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,4,power_law_1.01,0.09351040124893188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,4,power_law_1.01,0.10000640153884888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,4,power_law_1.01,0.1438912034034729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,4,power_law_1.01,0.16974079608917236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,4,power_law_1.01,0.22290558815002443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,4,power_law_1.01,0.22509438991546632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,4,power_law_1.01,0.24739201068878175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,4,power_law_1.01,0.2529344081878662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,4,power_law_1.01,0.24734079837799072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,32,1,power_law_1.01,0.03462400138378143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,32,1,power_law_1.01,0.0347135990858078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,32,1,power_law_1.01,0.03522560000419617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,4,power_law_1.01,0.2628799915313721
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,32,1,power_law_1.01,0.03720960021018982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,32,1,power_law_1.01,0.03722879886627197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,32,1,power_law_1.01,0.038150399923324585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,32,1,power_law_1.01,0.03860479891300202
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,32,1,power_law_1.01,0.039443200826644896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,32,1,power_law_1.01,0.039904001355171206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,4,power_law_1.01,0.2661119937896729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,32,1,power_law_1.01,0.03988479971885681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,32,1,power_law_1.01,0.04142720103263855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,32,1,power_law_1.01,0.043033599853515625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,32,1,power_law_1.01,0.04583039879798889
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,32,1,power_law_1.01,0.0440064013004303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,32,1,power_law_1.01,0.05144320130348205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,32,1,power_law_1.01,0.05264639854431152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,4,power_law_1.01,0.2714495897293091
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,32,1,power_law_1.01,0.05782399773597717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,32,1,power_law_1.01,0.06448000073432922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,32,1,power_law_1.01,0.07320320010185241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,32,1,power_law_1.01,0.09238399863243103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,32,1,power_law_1.01,0.1049280047416687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,32,1,power_law_1.01,0.13908480405807494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,4,power_law_1.01,0.2926464080810547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,32,1,power_law_1.01,0.1675328016281128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,32,1,power_law_1.01,0.23224320411682128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,32,1,power_law_1.01,0.29728639125823975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,32,1,power_law_1.01,0.422649621963501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,4,power_law_1.01,0.2894335985183716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,32,1,power_law_1.01,0.5540160179138184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,32,1,power_law_1.01,0.687775993347168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,32,1,power_law_1.01,1.0515263557434082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,4,power_law_1.01,0.3060415983200073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,32,1,power_law_1.01,2.031353569030762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,4,power_law_1.01,0.3250047922134399
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,4,power_law_1.01,0.3544831991195679
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,4,power_law_1.01,0.41063680648803713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,4,power_law_1.01,0.4470975875854492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,4,power_law_1.01,0.5411839962005616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,4,power_law_1.01,0.6532032012939453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,4,power_law_1.01,0.790828800201416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,4,power_law_1.01,1.0055295944213867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,4,power_law_1.01,1.398745632171631
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,8,8,balanced,0.03160000095764796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,8,8,balanced,0.03385599950949351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,8,8,balanced,0.031285333136717476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,8,8,balanced,0.03345066557327906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,8,8,balanced,0.031514666974544525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,8,8,balanced,0.033200000723203026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,8,8,balanced,0.03322133421897888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,8,8,balanced,0.033733333150545754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,8,8,balanced,0.03344533344109853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,8,8,balanced,0.033344000577926636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,2,8,power_law_1.2,0.04570240080356598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,8,8,balanced,0.033674667278925575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,8,8,balanced,0.03551999976237615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,8,8,balanced,0.03342933456103007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,8,8,balanced,0.035402665535608925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,8,8,balanced,0.0396373321612676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,2,8,power_law_1.2,0.04965119957923889
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,8,8,balanced,0.041562666495641075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,8,8,balanced,0.040133332212766014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,8,8,balanced,0.04569066564242045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,2,8,power_law_1.2,0.04522239863872528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,8,8,balanced,0.046053335070610046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,8,8,balanced,0.06012799839178721
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,2,8,power_law_1.2,0.04415999948978424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,8,8,balanced,0.05881600081920624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,8,8,balanced,0.08675199747085571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,8,8,balanced,0.10544533530871074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,2,8,power_law_1.2,0.044838398694992065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,8,8,balanced,0.13829867045084634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,8,8,balanced,0.16863999764124551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,2,8,power_law_1.2,0.0451200008392334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,8,8,balanced,0.23181867599487305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,8,8,balanced,0.2815200090408325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,2,8,power_law_1.2,0.04596480131149292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,8,8,balanced,0.3399680058161418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,4,power_law_1.01,1.6641855239868164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,2,8,power_law_1.2,0.046009600162506104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,8,8,balanced,0.5137759844462076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,2,8,power_law_1.2,0.049779200553894044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,8,8,balanced,0.9718560377756754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,2,8,power_law_1.2,0.057017600536346434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,2,8,power_law_1.2,0.05688959956169128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,2,8,power_law_1.2,0.06114559769630432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,2,8,power_law_1.2,0.06142079830169678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,2,8,power_law_1.2,0.06636160016059875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,2,8,power_law_1.2,0.07533439993858337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,2,8,power_law_1.2,0.07688959836959838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,2,8,power_law_1.2,0.08733440041542054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,2,8,power_law_1.2,0.10692479610443115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,4,power_law_1.01,2.225760078430176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,2,8,power_law_1.2,0.12021119594573974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,2,8,power_law_1.2,0.1456063985824585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,2,8,power_law_1.2,0.16992640495300293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,4,8,power_law_1.01,0.04744960069656372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,2,8,power_law_1.2,0.22061440944671631
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,4,8,power_law_1.01,0.04762240052223206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,4,8,power_law_1.01,0.04542079865932465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,2,8,power_law_1.2,0.30443520545959474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,4,8,power_law_1.01,0.04469119906425476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,2,8,power_law_1.2,0.42777600288391116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,4,8,power_law_1.01,0.045875200629234315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,4,8,power_law_1.01,0.050342398881912234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,2,8,power_law_1.2,0.5106560230255127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,4,8,power_law_1.01,0.05555840134620667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,4,8,power_law_1.01,0.040908798575401306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,4,8,power_law_1.01,0.046028798818588255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,2,8,power_law_1.2,0.8237695693969727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,4,8,power_law_1.01,0.04824320077896118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,4,power_law_1.01,3.038860893249512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,4,8,power_law_1.01,0.04222719967365265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,2,8,power_law_1.2,0.9676159858703614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,4,8,power_law_1.01,0.0513152003288269
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,4,8,power_law_1.01,0.05539199709892273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,4,8,power_law_1.01,0.057683199644088745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,4,8,power_law_1.01,0.06570240259170532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,2,8,power_law_1.2,1.209881591796875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,4,8,power_law_1.01,0.06561279892921448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,4,8,power_law_1.01,0.07534080147743225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,4,8,power_law_1.01,0.08617600202560424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,4,8,power_law_1.01,0.09248639941215515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,2,8,power_law_1.2,1.9410240173339843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,4,8,power_law_1.01,0.12262400388717651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,4,8,power_law_1.01,0.1388159990310669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,4,8,power_law_1.01,0.18247679471969605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,4,8,power_law_1.01,0.23280000686645508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,4,8,power_law_1.01,0.32603518962860106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,4,8,power_law_1.01,0.4168831825256348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,2,8,power_law_1.2,3.8354751586914064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,4,8,power_law_1.01,0.6147840023040771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,4,8,power_law_1.01,0.7796607971191406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,4,power_law_1.01,5.950368118286133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,4,8,power_law_1.01,1.0535743713378907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,4,8,power_law_1.01,1.553337574005127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,4,8,power_law_1.01,2.838591957092285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,1,1,balanced,0.09079999725023906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,1,1,balanced,0.13827199737230936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,1,1,balanced,0.24331732590993246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,1,1,balanced,0.24125866095225015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,1,1,balanced,0.2376213272412618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,1,1,balanced,0.24093866348266602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,1,1,balanced,0.24230400721232095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,1,1,balanced,0.24515734116236368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,1,1,balanced,0.24700266122817993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,1,1,balanced,0.24674665927886963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,1,1,balanced,0.2505279978116353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,1,1,balanced,0.2633440097173055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,1,1,balanced,0.26663466294606525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,1,1,balanced,0.26984532674153644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,1,1,balanced,0.2926186720530192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,1,1,balanced,0.29958399136861164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,1,1,balanced,0.3012640078862508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,1,1,balanced,0.3721280097961426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,1,1,balanced,0.3760960102081299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,1,1,balanced,0.5112959941228231
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,1,1,balanced,0.5634933312733968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,1,1,balanced,0.7776959737141927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,1,1,balanced,0.9546079635620117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,1,1,balanced,1.3987199465433757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,1,1,balanced,1.8330186208089192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,1,1,balanced,2.7069600423177085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,1,1,balanced,3.5507307052612305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,1,1,balanced,4.423839886983235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,1,1,balanced,7.03875732421875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,1,balanced,0.14758933583895364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,1,balanced,0.24402666091918945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,1,balanced,0.4354986747105916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,8,1,balanced,0.07030400137106578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,1,1,balanced,14.200234731038412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,1,balanced,0.8250666459401449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,8,1,balanced,0.07409599920113881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,8,1,balanced,0.09501866499582927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,8,1,balanced,0.12504532933235168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,1,balanced,1.5973493258158367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,8,1,balanced,0.18254933754603067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,8,1,balanced,0.2144533395767212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,8,1,balanced,0.22131733099619547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,8,1,balanced,0.21548799673716226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,1,balanced,3.140869458516439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,8,1,balanced,0.21804267168045044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,8,1,balanced,0.21824000279108682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,8,1,balanced,0.2228426734606425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,8,1,balanced,0.22691200176874796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,8,1,balanced,0.2293706734975179
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,1,balanced,3.1483147939046225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,8,1,balanced,0.23692800601323447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,8,1,balanced,0.2482773264249166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,8,1,balanced,0.2550986607869466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,8,1,balanced,0.27350932359695435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,8,1,balanced,0.31375465790430707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,1,balanced,3.156592051188151
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,8,1,balanced,0.34141333897908527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,8,1,balanced,0.4206560055414836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,64,balanced,0.05138133466243744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,64,balanced,0.05605866511662801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,8,1,balanced,0.478058656056722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,64,balanced,0.05610666672388712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,64,balanced,0.058143998185793556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,1,balanced,3.159093221028646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,64,balanced,0.06214400132497152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,8,1,balanced,0.6993333498636881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,64,balanced,0.06216000020503998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,64,balanced,0.062021334966023765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,64,balanced,0.06208533545335134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,64,balanced,0.063360000650088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,8,1,balanced,0.8206186294555664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,64,balanced,0.06181333462397257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,64,balanced,0.0624533345301946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,64,balanced,0.06549866497516632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,64,balanced,0.06643199920654297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,1,balanced,3.1652746200561523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,8,1,balanced,1.1847946643829346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,64,balanced,0.06805866460005443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,64,balanced,0.07063466807206471
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,64,balanced,0.072202667593956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,64,balanced,0.07779733339945476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,64,balanced,0.09081600109736125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,64,balanced,0.09927999973297119
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,8,1,balanced,1.509189287821452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,64,balanced,0.11940800150235494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,64,balanced,0.14204266667366028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,1,balanced,3.167248090108236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,64,balanced,0.19770665963490805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,64,balanced,0.2054133415222168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,8,1,balanced,2.222933292388916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,64,balanced,0.2693333427111308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,64,balanced,0.3253759940465291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,1,balanced,3.180090586344401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,64,balanced,0.44564799467722577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,64,balanced,0.5590506792068481
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,8,1,balanced,2.8611040115356445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,4,64,balanced,0.6748053232828776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,1,balanced,3.193984031677246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,4,64,balanced,1.0201013088226318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,8,1,balanced,3.405029296875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,1,balanced,3.216181437174479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,4,64,balanced,2.1614559491475425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,1,balanced,3.232858657836914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,8,1,balanced,5.354330698649089
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,1,balanced,3.2452160517374673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,1,balanced,3.2702506383260093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,1,balanced,3.342560132344564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,8,1,balanced,10.45082155863444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,1,balanced,3.390597343444824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,1,balanced,3.499333381652832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,1,balanced,3.625472068786621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,1,balanced,3.76531187693278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,1,balanced,4.053488095601399
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,1,balanced,4.575471878051758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,1,balanced,4.8436587651570635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,32,power_law_1.01,0.15119999647140503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,1,balanced,6.608320236206055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,32,power_law_1.01,0.15098880529403685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,1,balanced,7.25490125020345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,1,balanced,9.752485275268555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,32,power_law_1.01,0.07783039808273315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,32,power_law_1.01,0.09663360118865967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,4,power_law_1.2,0.11721600294113159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,8,2,balanced,0.03347733368476232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,8,2,balanced,0.03903999924659729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,1,balanced,12.801397959391275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,8,2,balanced,0.057562669118245445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,8,2,balanced,0.05829333265622457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,8,2,balanced,0.05862933397293091
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,8,2,balanced,0.057631999254226685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,8,2,balanced,0.059301331639289856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,8,2,balanced,0.06091733276844025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,8,2,balanced,0.059477334221204124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,8,2,balanced,0.059674665331840515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,8,2,balanced,0.05951466659704844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,8,2,balanced,0.06098666787147522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,8,2,balanced,0.062037333846092224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,8,2,balanced,0.06178133189678192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,8,2,balanced,0.07108266651630402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,32,power_law_1.01,0.11004799604415894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,8,2,balanced,0.07131200035413106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,8,2,balanced,0.07444266478220622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,8,2,balanced,0.08410132924715678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,8,2,balanced,0.08521599570910136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,8,2,balanced,0.10311999917030334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,8,2,balanced,0.11408000191052754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,4,power_law_1.2,0.10424959659576416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,8,2,balanced,0.14417066176732382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,8,2,balanced,0.17172267039616904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,8,2,balanced,0.2299999992052714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,8,2,balanced,0.2674400011698405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,8,2,balanced,0.37537598609924316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,8,2,balanced,0.4763466517130534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,8,2,balanced,0.5803466637929281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,8,2,balanced,0.8903840382893881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,32,power_law_1.01,0.13692159652709962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,8,2,balanced,1.7324479420979817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,4,power_law_1.2,0.10176639556884766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,32,power_law_1.01,0.14388480186462402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,1,balanced,25.59143575032552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,4,power_law_1.2,0.14409600496292113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,32,power_law_1.01,0.14685440063476562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,4,power_law_1.2,0.1992576003074646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,32,power_law_1.01,0.1457856059074402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,4,power_law_1.2,0.3245824098587036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,32,power_law_1.01,0.14174720048904418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,4,power_law_1.2,0.3512063980102539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,32,power_law_1.01,0.14587520360946654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,4,power_law_1.2,0.36457600593566897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,32,power_law_1.01,0.15270400047302246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,4,power_law_1.2,0.38068480491638185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,32,power_law_1.01,0.1534208059310913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,8,power_law_1.01,0.15750399827957154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,4,power_law_1.2,0.39453439712524413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,32,power_law_1.01,0.16074880361557006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,8,power_law_1.01,0.23543040752410888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,4,power_law_1.2,0.41750397682189944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,8,power_law_1.01,0.09812480211257935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,32,power_law_1.01,0.16409599781036377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,2,power_law_1.2,0.06543359756469727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,8,power_law_1.01,0.16996480226516725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,4,power_law_1.2,0.4142591953277588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,2,power_law_1.2,0.07268480062484742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,32,power_law_1.01,0.1726912021636963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,8,power_law_1.01,0.1831231951713562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,2,power_law_1.2,0.07689599990844727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,8,power_law_1.01,0.276198410987854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,4,power_law_1.2,0.4339136123657227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,32,power_law_1.01,0.17623679637908934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,8,power_law_1.01,0.27760000228881837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,2,power_law_1.2,0.0832319974899292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,4,power_law_1.2,0.44855680465698244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,8,power_law_1.01,0.28831360340118406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,32,power_law_1.01,0.20773758888244628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,2,power_law_1.2,0.1072767972946167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,8,power_law_1.01,0.27173120975494386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,4,power_law_1.2,0.4602047920227051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,32,power_law_1.01,0.21235840320587157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,8,power_law_1.01,0.2845632076263428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,2,power_law_1.2,0.12903679609298707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,4,power_law_1.2,0.4813504219055176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,2,power_law_1.2,0.1505280017852783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,32,power_law_1.01,0.27456638813018797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,8,power_law_1.01,0.3060287952423096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,2,power_law_1.2,0.18023040294647216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,4,power_law_1.2,0.4802815914154053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,8,power_law_1.01,0.3040704011917114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,32,power_law_1.01,0.30030078887939454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,2,power_law_1.2,0.18460160493850708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,8,power_law_1.01,0.29531519412994384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,4,power_law_1.2,0.5480127811431885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,32,power_law_1.01,0.37969279289245605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,2,power_law_1.2,0.19260799884796143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,8,power_law_1.01,0.323417592048645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,4,power_law_1.2,0.5896575927734375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,2,power_law_1.2,0.19671039581298827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,8,power_law_1.01,0.325766396522522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,32,power_law_1.01,0.37448959350585936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,8,power_law_1.01,0.33799679279327394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,2,power_law_1.2,0.20165760517120362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,4,power_law_1.2,0.6832640171051025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,8,power_law_1.01,0.3429120063781738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,32,power_law_1.01,0.4855616092681885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,2,power_law_1.2,0.2098304033279419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,4,power_law_1.2,0.791315221786499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,8,power_law_1.01,0.36943359375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,2,power_law_1.2,0.21659519672393798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,8,power_law_1.01,0.4352128028869629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,32,power_law_1.01,0.5976960182189941
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,2,power_law_1.2,0.2344383955001831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,4,power_law_1.2,0.9891903877258301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,8,power_law_1.01,0.5294784069061279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,2,power_law_1.2,0.23532159328460694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,8,power_law_1.01,0.5659840106964111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,4,power_law_1.2,0.9796159744262696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,32,power_law_1.01,0.7604671955108643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,2,power_law_1.2,0.2520639896392822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,8,power_law_1.01,0.6881919860839844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,2,power_law_1.2,0.272979211807251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,8,power_law_1.01,0.6735680103302002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,4,power_law_1.2,1.2029888153076171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,2,power_law_1.2,0.3122688055038452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,32,power_law_1.01,0.9372032165527344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,8,power_law_1.01,0.8603648185729981
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,2,power_law_1.2,0.360915207862854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,8,power_law_1.01,1.0965824127197266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,4,power_law_1.2,1.4924351692199707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,2,power_law_1.2,0.413862419128418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,8,power_law_1.01,1.5268863677978515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,1,32,power_law_1.01,1.1796416282653808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,2,power_law_1.2,0.534771203994751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,4,power_law_1.2,2.1011199951171875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,8,power_law_1.01,1.8227519989013672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,2,power_law_1.2,0.6662271976470947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,8,power_law_1.01,2.2349695205688476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,2,power_law_1.2,0.8562432289123535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,4,power_law_1.2,2.67392635345459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,1,32,power_law_1.01,1.7159296035766602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,2,power_law_1.2,1.1569024085998536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,8,power_law_1.01,3.172480010986328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,2,power_law_1.2,1.5544575691223144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,2,4,power_law_1.2,3.1793792724609373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,2,power_law_1.2,2.003558349609375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,8,power_law_1.01,5.964019012451172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,1,32,power_law_1.01,3.5967681884765623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,2,power_law_1.2,2.552998352050781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,2,4,power_law_1.2,4.4059711456298825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,2,power_law_1.2,3.6099456787109374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,2,power_law_1.2,7.406240081787109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,32,balanced,0.033226666351159416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,32,balanced,0.03105599929889043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,32,balanced,0.0323786661028862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,32,balanced,0.03329599897066752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,32,balanced,0.03543466577927271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,32,balanced,0.034416000048319496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,32,balanced,0.032773333291212715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,2,4,power_law_1.2,9.4793212890625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,32,balanced,0.03440000116825104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,32,balanced,0.034527999659379326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,32,balanced,0.034314667185147606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,32,balanced,0.03454933315515518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,32,balanced,0.034330666065216064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,32,balanced,0.03625066578388214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,32,balanced,0.0365226666132609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,32,balanced,0.03822399924198786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,32,balanced,0.040336000422636666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,32,balanced,0.040063999593257904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,32,balanced,0.042453333735466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,32,balanced,0.041989331444104515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,32,balanced,0.048437332113583885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,32,balanced,0.04837866624196371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,32,balanced,0.05685866872469584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,32,balanced,0.06492800017197926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,32,balanced,0.09244267145792644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,32,balanced,0.10935999949773152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,32,balanced,0.1455573340257009
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,32,balanced,0.1816106637318929
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,2,32,balanced,0.21505600214004517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,2,32,balanced,0.30874667565027875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,2,32,balanced,0.5543626546859741
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,32,4,power_law_1.01,0.04805760085582733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,32,4,power_law_1.01,0.05043839812278748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,32,4,power_law_1.01,0.049772799015045166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,32,4,power_law_1.01,0.05061759948730469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,32,4,power_law_1.01,0.05568000078201294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,32,4,power_law_1.01,0.06328960061073304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,32,4,power_law_1.01,0.07208960056304932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,32,4,power_law_1.01,0.0741823971271515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,32,4,power_law_1.01,0.07643520236015319
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,4,1,power_law_1.01,0.030368000268936157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,32,4,power_law_1.01,0.07631999850273133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,4,1,power_law_1.01,0.03858560025691986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,32,4,power_law_1.01,0.08085119724273682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,4,1,power_law_1.01,0.03850240111351013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,4,1,power_law_1.01,0.0420415997505188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,32,4,power_law_1.01,0.0864960014820099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,4,1,power_law_1.01,0.051648002862930295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,4,1,power_law_1.01,0.07388799786567687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,32,4,power_law_1.01,0.08961920142173767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,4,1,power_law_1.01,0.07747840285301208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,32,4,power_law_1.01,0.09537280201911927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,4,1,power_law_1.01,0.08097280263900757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,4,32,power_law_1.01,0.022944000363349915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,4,1,power_law_1.01,0.08208000063896179
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,4,32,power_law_1.01,0.021849599480628968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,32,4,power_law_1.01,0.10664960145950317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,16,power_law_1.01,0.055622398853302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,4,1,power_law_1.01,0.08437119722366333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,4,32,power_law_1.01,0.02184319943189621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,4,1,power_law_1.01,0.0889024019241333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,32,4,power_law_1.01,0.11329280138015747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,4,32,power_law_1.01,0.02082560062408447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,4,1,power_law_1.01,0.0915008008480072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,4,32,power_law_1.01,0.023046399652957916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,32,4,power_law_1.01,0.1136896014213562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,4,1,power_law_1.01,0.09784319996833801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,4,32,power_law_1.01,0.024198399484157564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,16,power_law_1.01,0.07965440154075623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,4,1,power_law_1.01,0.0979200005531311
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,4,32,power_law_1.01,0.024249599874019624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,32,4,power_law_1.01,0.13096319437026976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,4,1,power_law_1.01,0.10266239643096924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,4,32,power_law_1.01,0.024716800451278685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,4,32,power_law_1.01,0.026848000288009644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,4,1,power_law_1.01,0.1067199945449829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,32,4,power_law_1.01,0.15058560371398927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,16,power_law_1.01,0.05727999806404114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,4,32,power_law_1.01,0.02778880000114441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,4,1,power_law_1.01,0.11775360107421876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,4,32,power_law_1.01,0.029126399755477907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,32,4,power_law_1.01,0.18873599767684937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,4,1,power_law_1.01,0.1332352042198181
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,4,32,power_law_1.01,0.029772800207138062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,1,power_law_1.01,0.1428544044494629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,32,4,power_law_1.01,0.21571199893951415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,16,power_law_1.01,0.05306239724159241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,4,32,power_law_1.01,0.030060800909996032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,1,power_law_1.01,0.17812479734420777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,4,32,power_law_1.01,0.029356798529624938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,1,power_law_1.01,0.1972480058670044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,32,4,power_law_1.01,0.2842112064361572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,4,32,power_law_1.01,0.03187200129032135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,1,power_law_1.01,0.2586560010910034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,16,power_law_1.01,0.05655679702758789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,4,32,power_law_1.01,0.033055999875068666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,32,4,power_law_1.01,0.37550721168518064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,4,32,power_law_1.01,0.03330560028553009
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,1,power_law_1.01,0.3213567972183228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,4,32,power_law_1.01,0.03681919872760773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,1,power_law_1.01,0.44515199661254884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,16,power_law_1.01,0.05948160290718078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,32,4,power_law_1.01,0.5152575969696045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,32,power_law_1.01,0.03859840035438537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,32,power_law_1.01,0.04433279931545257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,1,power_law_1.01,0.5671487808227539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,32,4,power_law_1.01,0.6758975982666016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,32,power_law_1.01,0.045638400316238406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,16,power_law_1.01,0.07041280269622803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,1,power_law_1.01,0.7998208045959473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,32,power_law_1.01,0.05338240265846252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,32,4,power_law_1.01,0.958950424194336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,32,power_law_1.01,0.060192000865936277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,16,power_law_1.01,0.07191680073738098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,1,power_law_1.01,1.0255167961120606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,32,power_law_1.01,0.07978240251541138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,32,power_law_1.01,0.10486400127410889
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,16,power_law_1.01,0.07169280052185059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,32,4,power_law_1.01,1.2672831535339355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,1,power_law_1.01,1.258668804168701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,32,power_law_1.01,0.1447424054145813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,16,power_law_1.01,0.0731328010559082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,32,power_law_1.01,0.1515071988105774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,32,4,power_law_1.01,1.5892671585083007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,1,power_law_1.01,1.9490816116333007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,16,power_law_1.01,0.07567359805107117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,32,power_law_1.01,0.19444479942321777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,32,4,power_law_1.01,2.377222442626953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,16,power_law_1.01,0.07905279994010925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,32,power_law_1.01,0.32032639980316163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,1,power_law_1.01,3.817107009887695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,16,power_law_1.01,0.08138880133628845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,32,power_law_1.01,0.5362688064575195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,16,power_law_1.01,0.08202880024909973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,32,4,power_law_1.01,4.5554241180419925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,16,power_law_1.01,0.09045119881629944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,16,power_law_1.01,0.09557759761810303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,16,power_law_1.01,0.10064640045166015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,16,power_law_1.01,0.11098239421844483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,16,power_law_1.01,0.12008320093154908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,16,power_law_1.01,0.14634239673614502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,16,power_law_1.01,0.18023680448532103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,16,power_law_1.01,0.23708159923553468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,4,balanced,0.05418133238951365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,16,power_law_1.01,0.27191040515899656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,4,balanced,0.054192001620928444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,4,balanced,0.05613866448402405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,4,balanced,0.07032000025113423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,4,balanced,0.09826133648554485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,4,balanced,0.10686933000882466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,4,balanced,0.10748799641927083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,4,balanced,0.10921600461006165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,4,balanced,0.10826133688290913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,4,balanced,0.10977066556612651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,4,balanced,0.11237333218256633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,4,balanced,0.11124799648920695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,4,balanced,0.11425600449244182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,4,balanced,0.11754133303960164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,4,balanced,0.12462400396664937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,4,balanced,0.12593600153923035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,4,balanced,0.13293866316477457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,4,balanced,0.14550399780273438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,4,balanced,0.15401599804560342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,4,balanced,0.1832533280054728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,4,balanced,0.19909334182739258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,4,balanced,0.2663573424021403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,16,power_law_1.01,0.3708224058151245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,4,balanced,0.29894934097925824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,4,balanced,0.4330879847208659
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,32,4,balanced,0.04196799794832865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,4,balanced,0.5281546513239542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,32,4,balanced,0.04386133452256521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,32,4,balanced,0.047775998711586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,32,4,balanced,0.05020800232887268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,4,balanced,0.7588213284810384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,32,4,balanced,0.06213866670926412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,32,4,balanced,0.07820266485214233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,32,4,balanced,0.07913066446781158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,32,4,balanced,0.07444266478220622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,32,4,balanced,0.07675200204054515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,4,balanced,0.9607306321461996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,32,4,balanced,0.07684266567230225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,32,4,balanced,0.07859733204046886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,32,4,balanced,0.07851733267307281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,32,4,balanced,0.08157333234945933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,4,balanced,1.1444053649902344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,32,4,balanced,0.08045866588751475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,32,4,balanced,0.08865066369374593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,32,4,balanced,0.08915199836095174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,32,4,balanced,0.09291733304659526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,32,4,balanced,0.10313066840171814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,32,4,balanced,0.10988266269365947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,32,4,balanced,0.1426400045553843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,4,balanced,1.772330602010091
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,32,4,balanced,0.1623360017935435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,2,power_law_1.01,0.036671999096870425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,32,4,balanced,0.22175467014312744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,16,power_law_1.01,0.45411200523376466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,32,4,balanced,0.24447466929753622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,32,4,balanced,0.36158398787180585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,32,4,balanced,0.426581343015035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,32,4,balanced,0.5834399859110514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,4,balanced,3.4341866175333657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,2,power_law_1.01,0.04411520063877106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,32,4,balanced,0.7482879956563314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,32,4,balanced,0.9216319719950358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,2,power_law_1.01,0.04376960098743439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,32,4,balanced,1.3915947278340657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,2,power_law_1.01,0.06610559821128845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,32,4,balanced,2.643514633178711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,16,power_law_1.01,0.6549056053161622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,2,power_law_1.01,0.08039039969444275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,2,power_law_1.01,0.10976639986038209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,2,power_law_1.01,0.1158336043357849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,16,power_law_1.01,0.8414336204528808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,2,power_law_1.01,0.1179967999458313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,2,power_law_1.01,0.12047359943389893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,2,power_law_1.01,0.12065919637680053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,2,power_law_1.01,0.12541439533233642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,16,power_law_1.01,1.0349184036254884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,2,power_law_1.01,0.13082879781723022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,2,power_law_1.01,0.13383040428161622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,2,power_law_1.01,0.13671679496765138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,2,power_law_1.01,0.14327679872512816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,16,power_law_1.01,1.5541440010070802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,2,power_law_1.01,0.15000959634780883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,2,power_law_1.01,0.1630784034729004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,2,power_law_1.01,0.18103040456771852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,2,power_law_1.01,0.2042304039001465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,2,power_law_1.01,0.2576512098312378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,2,power_law_1.01,0.28558080196380614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,16,power_law_1.01,3.149279975891113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,2,power_law_1.01,0.36029438972473143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,2,power_law_1.01,0.432966423034668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,2,power_law_1.01,0.6278975963592529
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,32,balanced,0.055061335364977516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,32,balanced,0.0440586656332016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,32,balanced,0.04394666850566864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,32,balanced,0.045738667249679565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,32,balanced,0.04715733230113983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,32,balanced,0.05108266572157542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,32,balanced,0.05390933156013489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,32,balanced,0.0525546669960022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,32,balanced,0.051818668842315674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,32,balanced,0.05221866567929586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,32,balanced,0.05436266462008158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,32,balanced,0.05395199855168661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,32,balanced,0.05453866720199585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,32,balanced,0.058229332168896995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,32,balanced,0.062208001812299095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,32,balanced,0.06356266637643178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,2,power_law_1.01,0.7965055942535401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,2,power_law_1.01,0.06451200246810913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,32,balanced,0.0666720022757848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,32,balanced,0.07263466715812683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,32,balanced,0.07903466622034709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,32,balanced,0.0909546713034312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,32,balanced,0.10132799545923869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,32,balanced,0.11744532982508342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,32,balanced,0.13512532909711203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,32,balanced,0.17298134167989096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,32,balanced,0.2067413330078125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,32,balanced,0.2924373348553975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,32,balanced,0.3521866798400879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,32,balanced,0.4378133217493693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,2,power_law_1.01,1.141260814666748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,32,balanced,0.6242719888687134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,2,power_law_1.01,0.07301759719848633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,32,balanced,1.2122080326080322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,2,power_law_1.01,1.3926976203918457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,2,power_law_1.01,0.1019327998161316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,2,power_law_1.01,1.7060543060302735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,2,power_law_1.01,0.09144319891929627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,2,power_law_1.01,0.12791680097579955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,2,power_law_1.01,2.613254356384277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,2,power_law_1.01,0.16491520404815674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,2,power_law_1.01,0.18121600151062012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,64,balanced,0.04112533231576284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,64,balanced,0.040778666734695435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,64,balanced,0.036677333215872444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,64,balanced,0.03721066564321518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,64,balanced,0.039349332451820374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,64,balanced,0.03737066686153412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,64,balanced,0.03804266701141993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,64,balanced,0.037061333656311035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,64,balanced,0.0386613334218661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,2,power_law_1.01,0.21644160747528077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,64,balanced,0.03730666637420654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,2,power_law_1.01,5.565343856811523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,64,balanced,0.03917866696914037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,64,balanced,0.039279999832312264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,64,balanced,0.040762667854626976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,64,balanced,0.03871466716130575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,64,balanced,0.04269866645336151
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,64,balanced,0.04293333490689596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,64,balanced,0.04487466812133789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,64,balanced,0.04647466540336609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,64,balanced,0.047322665651639305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,64,balanced,0.04943466683228811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,64,balanced,0.052426666021347046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,64,balanced,0.06061333417892456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,64,balanced,0.06828799843788147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,64,balanced,0.08718933661778767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,2,power_law_1.01,0.23740160465240479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,64,balanced,0.09594133496284485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,64,balanced,0.1269760032494863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,64,balanced,0.15242666999499002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,64,balanced,0.17847466468811035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,64,balanced,0.25826666752497357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,64,balanced,0.4662880102793376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,2,power_law_1.01,0.23447039127349853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,2,power_law_1.01,0.2474047899246216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,8,power_law_1.01,0.07308160066604615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,8,power_law_1.01,0.10278400182723998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,2,power_law_1.01,0.2554431915283203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,8,power_law_1.01,0.055270397663116456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,8,power_law_1.01,0.07093120217323304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,2,power_law_1.01,0.2650559902191162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,8,power_law_1.01,0.08447999954223633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,8,power_law_1.01,0.10199040174484253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,2,power_law_1.01,0.267577600479126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,8,power_law_1.01,0.10647039413452149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,8,power_law_1.01,0.1087488055229187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,2,power_law_1.01,0.28254079818725586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,8,power_law_1.01,0.11374080181121826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,8,power_law_1.01,0.11169279813766479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,2,power_law_1.01,0.28672640323638915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,8,power_law_1.01,0.11174399852752685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,8,power_law_1.01,0.11582080125808716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,2,power_law_1.01,0.30448639392852783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,8,power_law_1.01,0.11862399578094482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,8,power_law_1.01,0.12347520589828491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,2,balanced,0.036042665441830955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,2,balanced,0.03389866650104523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,2,balanced,0.03886399914820989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,256,power_law_1.2,0.07870720028877258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,2,power_law_1.01,0.3285696029663086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,2,balanced,0.05846933523813883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,2,balanced,0.08237333099047343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,2,balanced,0.08384533723195393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,2,balanced,0.08328533172607422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,2,balanced,0.08630399902661641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,8,power_law_1.01,0.13045120239257812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,2,balanced,0.0855519970258077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,2,balanced,0.08559466401735942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,2,balanced,0.08761599659919739
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,2,balanced,0.08946133653322856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,2,balanced,0.09109866619110107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,2,balanced,0.09399466713269551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,2,balanced,0.10075733065605164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,256,power_law_1.2,0.08449919819831848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,8,power_law_1.01,0.13967360258102418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,2,balanced,0.10506133238474528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,2,balanced,0.10958932836850484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,2,balanced,0.11893332997957866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,2,balanced,0.12903466820716858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,2,power_law_1.01,0.359878396987915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,2,balanced,0.15582399566968283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,2,balanced,0.17023466030756632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,2,balanced,0.22744532426198324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,8,power_law_1.01,0.14230400323867798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,2,balanced,0.26396799087524414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,2,balanced,0.3802186648050944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,256,power_law_1.2,0.08179200291633607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,2,balanced,0.4771253267923991
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,2,balanced,0.6868533293406168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,8,power_law_1.01,0.1674239993095398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,2,balanced,0.8928960164388021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,2,power_law_1.01,0.42188801765441897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,2,balanced,1.1007893085479736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,256,power_law_1.2,0.08138880133628845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,2,balanced,1.7385600407918294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,8,power_law_1.01,0.17689599990844726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,2,balanced,3.3886934916178384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,256,power_law_1.2,0.08061439990997314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,8,power_law_1.01,0.21990399360656737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,2,power_law_1.01,0.4790463924407959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,8,power_law_1.01,0.23633279800415039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,256,power_law_1.2,0.08102399706840516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,8,power_law_1.01,0.2907776117324829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,2,power_law_1.01,0.581766414642334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,256,power_law_1.2,0.07966079711914062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,8,power_law_1.01,0.3622591972351074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,256,power_law_1.2,0.08131200075149536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,2,power_law_1.01,0.7013823986053467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,8,power_law_1.01,0.4857791900634766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,256,power_law_1.2,0.08160640001296997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,8,power_law_1.01,0.6663424015045166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,2,power_law_1.01,0.903775978088379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,256,power_law_1.2,0.08015360236167908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,8,power_law_1.01,0.8367232322692871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,256,power_law_1.2,0.07886719703674316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,2,power_law_1.01,1.1473407745361328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,256,power_law_1.2,0.08017920255661011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,8,power_law_1.01,1.0705344200134277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,256,power_law_1.2,0.07920640110969543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,256,power_law_1.2,0.07585920095443725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,2,power_law_1.01,1.5869888305664062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,8,power_law_1.01,1.2875200271606446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,256,power_law_1.2,0.08117759823799134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,256,power_law_1.2,0.08067839741706848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,8,power_law_1.01,1.9908479690551757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,2,power_law_1.01,2.0568063735961912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,256,power_law_1.2,0.09498879909515381
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,256,power_law_1.2,0.09722880125045777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,1,power_law_1.2,0.0721343994140625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,256,power_law_1.2,0.10885119438171387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,2,power_law_1.01,2.5068735122680663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,1,power_law_1.2,0.08037760257720947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,8,power_law_1.01,3.909734344482422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,256,power_law_1.2,0.14126720428466796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,1,power_law_1.2,0.11787519454956055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,256,power_law_1.2,0.1670016050338745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,1,power_law_1.2,0.16359039545059204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,2,power_law_1.01,3.784537506103516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,256,power_law_1.2,0.21432960033416748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,1,power_law_1.2,0.23196799755096437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,256,power_law_1.2,0.30062720775604246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,1,power_law_1.2,0.42746877670288086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,1,power_law_1.2,0.4747583866119385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,256,power_law_1.2,0.395686411857605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,1,power_law_1.2,0.4973887920379639
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,256,power_law_1.2,0.5146560192108154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,1,power_law_1.2,0.5107840061187744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,2,power_law_1.01,7.554297637939453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,1,power_law_1.2,0.5289472103118896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,256,power_law_1.2,0.7860799789428711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,1,power_law_1.2,0.5504320144653321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,256,power_law_1.2,1.1051008224487304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,1,power_law_1.2,0.5586624145507812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,1,power_law_1.2,0.5670207977294922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,256,power_law_1.2,1.4546879768371581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,1,power_law_1.2,0.5810304164886475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,8,power_law_1.2,0.05102720260620117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,1,power_law_1.2,0.6103551864624024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,8,power_law_1.2,0.05348479747772217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,256,power_law_1.2,2.1632064819335937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,8,power_law_1.2,0.05125759840011597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,1,power_law_1.2,0.6169600009918212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,8,power_law_1.2,0.05424000024795532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,1,power_law_1.2,0.6351168155670166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,8,power_law_1.2,0.0590399980545044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,8,power_law_1.2,0.07374079823493958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,1,power_law_1.2,0.7023615837097168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,8,power_law_1.2,0.07898880243301391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,1,power_law_1.2,0.7457344055175781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,8,power_law_1.2,0.08065919876098633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,256,power_law_1.2,5.10700798034668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,8,power_law_1.2,0.07919999957084656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,1,power_law_1.2,0.8723199844360352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,8,power_law_1.2,0.07918720245361328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,4,8,power_law_1.01,0.037190398573875426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,4,8,power_law_1.01,0.03619840145111084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,8,power_law_1.2,0.08183680176734924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,1,power_law_1.2,0.897913646697998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,4,8,power_law_1.01,0.034560000896453856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,4,8,power_law_1.01,0.03431040048599243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,4,8,power_law_1.01,0.03550080060958862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,8,power_law_1.2,0.08574720025062561
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,4,8,power_law_1.01,0.035718399286270144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,4,8,power_law_1.01,0.03640959858894348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,4,8,power_law_1.01,0.03868159949779511
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,1,power_law_1.2,1.0771967887878418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,4,8,power_law_1.01,0.037363201379776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,8,power_law_1.2,0.08625919818878174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,4,8,power_law_1.01,0.03763200044631958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,4,8,power_law_1.01,0.03775359988212586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,4,8,power_law_1.01,0.041043201088905336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,8,power_law_1.2,0.0899392008781433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,4,8,power_law_1.01,0.04142079949378967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,4,8,power_law_1.01,0.043059200048446655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,4,8,power_law_1.01,0.04919680058956146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,1,power_law_1.2,1.2650879859924316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,4,8,power_law_1.01,0.05253120064735413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,8,power_law_1.2,0.09876480102539062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,4,8,power_law_1.01,0.05512319803237915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,4,8,power_law_1.01,0.06375679969787598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,4,8,power_law_1.01,0.07312639951705932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,8,power_law_1.2,0.1044543981552124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,4,8,power_law_1.01,0.09182720184326172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,4,8,power_law_1.01,0.09975680112838745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,1,power_law_1.2,1.6229375839233398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,4,8,power_law_1.01,0.13178240060806273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,8,power_law_1.2,0.10388480424880982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,4,8,power_law_1.01,0.16059520244598388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,4,8,power_law_1.01,0.21583359241485595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,8,power_law_1.2,0.11787519454956055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,4,8,power_law_1.01,0.2653696060180664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,4,8,power_law_1.01,0.39363200664520265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,1,power_law_1.2,2.073151969909668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,8,power_law_1.2,0.1257151961326599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,4,8,power_law_1.01,0.4802112102508545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,4,8,power_law_1.01,0.587385606765747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,8,power_law_1.2,0.16429439783096314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,4,8,power_law_1.01,0.9086912155151368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,8,power_law_1.2,0.17753599882125853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,1,power_law_1.2,2.77260799407959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,8,power_law_1.2,0.22574079036712646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,4,8,power_law_1.01,1.8229888916015624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,8,power_law_1.2,0.29304959774017336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,1,power_law_1.2,3.5916351318359374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,8,power_law_1.2,0.4056575775146484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,8,power_law_1.2,0.4827455997467041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,8,power_law_1.01,0.07990400195121765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,8,power_law_1.2,0.5693503856658936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,1,power_law_1.2,4.263308715820313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,8,power_law_1.2,1.0101056098937988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,8,power_law_1.01,0.08841599822044373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,8,power_law_1.2,1.0424960136413575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,8,power_law_1.01,0.058246397972106935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,1,power_law_1.2,5.98039665222168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,8,power_law_1.01,0.06730239987373351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,8,power_law_1.2,1.8642816543579102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,8,power_law_1.01,0.08095999956130981
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,32,2,power_law_1.01,0.03335039913654327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,8,power_law_1.2,3.4407680511474608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,32,2,power_law_1.01,0.032742398977279666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,32,2,power_law_1.01,0.03327359855175018
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,32,2,power_law_1.01,0.03383679986000061
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,8,power_law_1.01,0.10005760192871094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,32,2,power_law_1.01,0.034041601419448855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,32,2,power_law_1.01,0.034745600819587705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,32,2,power_law_1.01,0.034944000840187076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,32,2,power_law_1.01,0.03587839901447296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,1,power_law_1.2,11.366758728027344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,32,2,power_law_1.01,0.036831998825073244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,32,2,power_law_1.01,0.037222400307655334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,32,2,power_law_1.01,0.03689599931240082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,32,2,power_law_1.01,0.039603200554847715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,32,2,power_law_1.01,0.03831680119037628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,8,power_law_1.01,0.10374399423599243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,32,2,power_law_1.01,0.03916800022125244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,32,2,power_law_1.01,0.04386560022830963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,32,2,power_law_1.01,0.04792320132255554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,32,2,power_law_1.01,0.04925439953804016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,32,2,power_law_1.01,0.05786240100860596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,32,2,power_law_1.01,0.060812801122665405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,32,2,power_law_1.01,0.07007359862327575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,32,2,power_law_1.01,0.09067519903182983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,8,power_law_1.01,0.10413440465927123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,32,2,power_law_1.01,0.1108672022819519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,32,2,power_law_1.01,0.14351999759674072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,32,2,power_law_1.01,0.18222719430923462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,32,2,power_law_1.01,0.22440319061279296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,32,2,power_law_1.01,0.36497280597686765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,8,power_law_1.01,0.10392960309982299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,32,2,power_law_1.01,0.4141248226165771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,32,2,power_law_1.01,0.5283711910247803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,32,2,power_law_1.01,0.844115161895752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,8,power_law_1.01,0.10613759756088256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,32,2,power_law_1.01,1.6390592575073242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,8,power_law_1.01,0.10782719850540161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,8,power_law_1.01,0.11162240505218506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,1,balanced,0.19561066230138144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,1,balanced,0.32655467589696247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,8,power_law_1.01,0.11352959871292115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,1,balanced,0.43964799245198566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,1,balanced,0.8259039719899496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,1,balanced,1.605445384979248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,1,balanced,3.1358718872070312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,8,power_law_1.01,0.11822719573974609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,1,balanced,4.677599906921387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,1,balanced,4.701375961303711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,8,power_law_1.01,0.12494080066680908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,1,8,power_law_1.01,0.05899519920349121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,1,8,power_law_1.01,0.05811840295791626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,1,balanced,4.692890803019206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,1,8,power_law_1.01,0.053600001335144046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,1,8,power_law_1.01,0.052742397785186766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,1,balanced,4.698554674784343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,1,8,power_law_1.01,0.053932797908782956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,8,power_law_1.01,0.12977919578552247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,1,8,power_law_1.01,0.055929601192474365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,1,8,power_law_1.01,0.05648639798164368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,1,balanced,4.710997263590495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,1,8,power_law_1.01,0.056454402208328244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,1,8,power_law_1.01,0.06222720146179199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,1,balanced,4.727157274881999
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,1,8,power_law_1.01,0.06389759778976441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,8,power_law_1.01,0.13527679443359375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,1,8,power_law_1.01,0.06435840129852295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,16,balanced,0.04578666885693868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,16,balanced,0.04378133515516917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,1,8,power_law_1.01,0.06638079881668091
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,16,balanced,0.047775998711586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,16,balanced,0.04983466863632202
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,16,balanced,0.07051733136177063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,1,balanced,4.73416010538737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,16,balanced,0.09725333253542583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,16,balanced,0.12874666849772134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,16,balanced,0.12289599577585857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,1,8,power_law_1.01,0.07082880139350892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,16,balanced,0.12282133102416992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,16,balanced,0.12449066837628682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,16,balanced,0.12839999794960022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,16,balanced,0.1262933313846588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,16,balanced,0.13007466991742453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,1,8,power_law_1.01,0.0717311978340149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,16,balanced,0.13040533661842346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,16,balanced,0.13537066181500754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,16,balanced,0.13622400164604187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,16,balanced,0.14044800400733948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,16,balanced,0.14829867084821066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,1,balanced,4.7747039794921875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,1,8,power_law_1.01,0.08992000222206116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,16,balanced,0.15356799960136414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,16,balanced,0.16965333620707193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,16,balanced,0.1831093430519104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,8,power_law_1.01,0.15651839971542358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,16,balanced,0.21423999468485513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,1,8,power_law_1.01,0.09166719913482665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,16,balanced,0.24441067377726236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,16,balanced,0.3009759982426961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,16,balanced,0.3651626507441203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,1,8,power_law_1.01,0.09746559858322143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,16,balanced,0.48070398966471356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,1,balanced,4.791760126749675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,16,balanced,0.6029653151830038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,1,8,power_law_1.01,0.11191680431365966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,16,balanced,0.7461439768473307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,1,8,power_law_1.01,0.1405184030532837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,16,balanced,1.0732213656107585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,1,balanced,4.807840029398601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,1,8,power_law_1.01,0.18965760469436646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,16,balanced,2.081567923227946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,8,power_law_1.01,0.17418240308761596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,1,8,power_law_1.01,0.2084671974182129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,1,balanced,4.8271786371866865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,1,8,power_law_1.01,0.27526400089263914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,1,8,power_law_1.01,0.37740159034729004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,1,balanced,4.905466715494792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,1,8,power_law_1.01,0.5221504211425781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,8,power_law_1.01,0.21099519729614258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,1,8,power_law_1.01,0.7184959888458252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,1,balanced,4.961418787638347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,1,8,power_law_1.01,1.0674880027770997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,1,balanced,5.063722610473633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,1,8,power_law_1.01,1.236684799194336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,8,power_law_1.01,0.2564352035522461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,1,8,power_law_1.01,1.8665983200073242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,1,balanced,5.179567972819011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,1,balanced,5.381610870361328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,1,8,power_law_1.01,2.525267219543457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,8,power_law_1.01,0.3216383934020996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,1,balanced,5.6129811604817705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,1,8,power_law_1.01,5.020640182495117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,1,balanced,5.968122482299805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,8,power_law_1.01,0.35905919075012205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,1,balanced,6.45692253112793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,1,balanced,7.107776006062825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,8,power_law_1.01,0.47153921127319337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,128,balanced,0.06326933205127716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,128,balanced,0.06483733157316844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,128,balanced,0.06518400212128957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,128,balanced,0.062080000837643944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,128,balanced,0.06418666740258534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,128,balanced,0.07877866427103679
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,128,balanced,0.06945600112279256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,128,balanced,0.07463466624418895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,1,balanced,9.561653137207031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,128,balanced,0.07283199826876323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,128,balanced,0.07442133128643036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,128,balanced,0.07234133283297221
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,128,balanced,0.07433066765467326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,128,balanced,0.07754133145014445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,128,balanced,0.07874133189519246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,128,balanced,0.08385066191355388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,128,balanced,0.08872532844543457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,128,balanced,0.09488532940546672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,128,balanced,0.09946133693059285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,128,balanced,0.10706667105356853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,128,balanced,0.12691199779510498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,128,balanced,0.14076266686121622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,128,balanced,0.17398399114608765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,128,balanced,0.20512000719706217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,128,balanced,0.2765333255132039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,128,balanced,0.3389066855112712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,8,power_law_1.01,0.6078271865844727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,128,balanced,0.48044268290201825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,128,balanced,0.6103839874267578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,128,balanced,0.7463626861572266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,128,balanced,1.1482453346252441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,1,balanced,12.02243169148763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,128,balanced,2.2519893646240234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,8,power_law_1.01,0.8380800247192383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,1,balanced,14.976912180582682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,8,power_law_1.01,0.9991999626159668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,8,32,power_law_1.2,0.05750399827957153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,1,balanced,28.56507110595703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,8,32,power_law_1.2,0.04896639883518219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,4,8,power_law_1.01,1.2895487785339355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,8,32,power_law_1.2,0.048467200994491574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,8,32,power_law_1.2,0.0488319993019104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,8,32,power_law_1.2,0.04943360090255737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,8,32,power_law_1.2,0.05082880258560181
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,1,power_law_1.2,0.0695743978023529
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,4,8,power_law_1.01,1.9664575576782226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,8,32,power_law_1.2,0.05148800015449524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,1,power_law_1.2,0.11905280351638795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,8,32,power_law_1.2,0.05272960066795349
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,1,power_law_1.2,0.11971839666366577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,8,32,power_law_1.2,0.05295360088348389
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,8,32,power_law_1.2,0.05249919891357422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,1,power_law_1.2,0.1719488024711609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,8,32,power_law_1.2,0.05415679812431336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,1,power_law_1.2,0.3046463966369629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,8,32,power_law_1.2,0.05612800121307373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,1,4,power_law_1.2,0.02971520125865936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,1,power_law_1.2,0.42279682159423826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,8,32,power_law_1.2,0.05602560043334961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,1,4,power_law_1.2,0.0392192006111145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,8,32,power_law_1.2,0.05825279951095581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,1,4,power_law_1.2,0.036499199271202085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,1,power_law_1.2,0.5050496101379395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,4,8,power_law_1.01,3.9809280395507813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,1,4,power_law_1.2,0.040275201201438904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,8,32,power_law_1.2,0.062067198753356936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,1,4,power_law_1.2,0.04797439873218536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,1,power_law_1.2,0.7318336009979248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,8,32,power_law_1.2,0.06376960277557372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,1,4,power_law_1.2,0.06808320283889771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,1,4,power_law_1.2,0.07342079877853394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,8,32,power_law_1.2,0.06826239824295044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,1,power_law_1.2,0.770854377746582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,1,4,power_law_1.2,0.07116159796714783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,8,32,power_law_1.2,0.07972480058670044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,1,4,power_law_1.2,0.07215999960899352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,1,4,power_law_1.2,0.07371519804000855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,1,power_law_1.2,0.8092288017272949
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,8,32,power_law_1.2,0.08664960265159607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,1,4,power_law_1.2,0.07562239766120911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,8,32,power_law_1.2,0.10695040225982666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,1,4,power_law_1.2,0.07453439831733703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,1,power_law_1.2,0.8363455772399903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,1,4,power_law_1.2,0.07473919987678528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,8,32,power_law_1.2,0.11932799816131592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,1,4,power_law_1.2,0.08045439720153809
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,1,4,power_law_1.2,0.08636159896850586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,1,power_law_1.2,0.8966976165771484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,8,32,power_law_1.2,0.1493056058883667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,1,4,power_law_1.2,0.08821759819984436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,1,4,power_law_1.2,0.0900160014629364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,8,32,power_law_1.2,0.18605439662933348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,1,power_law_1.2,0.9174592018127441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,1,4,power_law_1.2,0.09919999837875366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,4,power_law_1.2,0.10810240507125854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,8,32,power_law_1.2,0.2511552095413208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,4,power_law_1.2,0.1266816020011902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,1,power_law_1.2,0.9565567970275879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,4,power_law_1.2,0.13338880538940429
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,8,32,power_law_1.2,0.3327296018600464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,32,balanced,0.049973333875338234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,32,balanced,0.08226666847864787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,4,power_law_1.2,0.1593664050102234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,32,balanced,0.08458133538564046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,32,balanced,0.12350400288899739
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,32,balanced,0.15268266201019287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,32,balanced,0.15634133418401083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,32,balanced,0.15863466262817383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,1,power_law_1.2,1.0121600151062011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,32,balanced,0.15498666961987814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,32,balanced,0.15468266606330872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,32,balanced,0.1569706698258718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,32,balanced,0.15710399548212686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,32,balanced,0.15803200006484985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,4,power_law_1.2,0.21112959384918212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,32,balanced,0.1592693328857422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,32,balanced,0.160261332988739
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,32,balanced,0.16543466846148172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,32,balanced,0.16447466611862183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,8,32,power_law_1.2,0.5157311916351318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,32,balanced,0.17038933436075845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,32,balanced,0.18881599108378092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,32,balanced,0.18552533785502115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,32,balanced,0.23965332905451456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,32,balanced,0.22350933154424033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,4,power_law_1.2,0.2774463891983032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,32,balanced,0.3338079849878947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,32,balanced,0.27941866715749103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,32,balanced,0.3850026528040568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,1,power_law_1.2,1.0359487533569336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,32,balanced,0.3797920147577922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,32,balanced,0.5550986528396606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,32,balanced,0.5799786647160848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,4,power_law_1.2,0.3576064109802246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,1,32,balanced,0.752021312713623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,8,32,power_law_1.2,0.5994431972503662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,1,32,balanced,1.032597303390503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,1,32,balanced,2.067216078440348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,4,power_law_1.2,0.45667200088500975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,1,power_law_1.2,1.1102272033691407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,8,32,power_law_1.2,0.7660096168518067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,4,power_law_1.2,0.7018176078796386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,1,power_law_1.2,1.2281984329223632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,4,power_law_1.2,0.5977727890014648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,8,32,power_law_1.2,1.332147216796875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,1,power_law_1.2,1.3278976440429688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,4,power_law_1.2,1.0540287971496582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,1,power_law_1.2,1.4950400352478028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,8,32,power_law_1.2,2.198297691345215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,1,power_law_1.2,1.519711971282959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,4,power_law_1.2,2.3775423049926756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,1,power_law_1.2,1.7876863479614258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,1,power_law_1.2,2.1038272857666014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,32,2,balanced,0.03581333408753077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,32,2,balanced,0.03765333443880081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,32,2,balanced,0.035616000493367515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,32,2,balanced,0.037530665596326195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,32,2,balanced,0.035530666510264076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,32,2,balanced,0.03756266583998998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,32,2,balanced,0.03584533433119456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,32,2,balanced,0.037578667203585304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,32,2,balanced,0.03748266647259394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,32,2,balanced,0.03781333317359289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,32,2,balanced,0.03781333317359289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,32,2,balanced,0.037503999968369804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,32,2,balanced,0.04001066585381826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,32,2,balanced,0.04182399809360504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,32,2,balanced,0.04428799947102865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,32,2,balanced,0.04580800235271454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,32,2,balanced,0.04784533381462097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,32,2,balanced,0.05398400127887726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,1,power_law_1.2,2.637824058532715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,32,2,balanced,0.056048000852266945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,32,2,balanced,0.06538666784763336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,32,2,balanced,0.0765066643555959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,32,2,balanced,0.09907199939092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,32,2,balanced,0.1111253301302592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,32,2,balanced,0.14917332927385965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,32,2,balanced,0.18147200345993042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,32,2,balanced,0.25729600588480633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,32,2,balanced,0.3314239978790283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,32,2,balanced,0.4074133237202962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,32,2,balanced,0.624074657758077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,32,2,balanced,1.205349365870158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,1,power_law_1.2,2.8305984497070313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,1,power_law_1.2,3.6025985717773437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,8,16,balanced,0.05194666484991709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,8,16,balanced,0.05007466673851013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,8,16,balanced,0.052149335543314614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,8,16,balanced,0.05221866567929586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,8,16,balanced,0.05435733497142792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,8,16,balanced,0.055888002117474876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,8,16,balanced,0.05839466551939646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,8,16,balanced,0.058133333921432495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,8,16,balanced,0.05782933533191681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,8,16,balanced,0.058245331048965454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,8,16,balanced,0.05821866790453593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,8,16,balanced,0.059861332178115845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,64,power_law_1.2,0.04123519957065582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,8,16,balanced,0.05994666616121928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,8,16,balanced,0.06234666705131531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,8,16,balanced,0.06471999982992808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,8,16,balanced,0.06445866823196411
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,8,16,balanced,0.06805866460005443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,8,16,balanced,0.07470400134722392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,8,16,balanced,0.07879466811815898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,8,16,balanced,0.10083733002344768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,1,power_law_1.2,4.6237632751464846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,8,16,balanced,0.1113920013109843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,8,16,balanced,0.14192533493041992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,8,16,balanced,0.15572800238927206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,8,16,balanced,0.20746133724848428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,8,16,balanced,0.2506399949391683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,8,16,balanced,0.34508800506591797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,8,16,balanced,0.4249333143234253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,64,power_law_1.2,0.039052799344062805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,8,16,balanced,0.5068320035934448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,8,16,balanced,0.7782346407572428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,8,16,balanced,1.5131893157958984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,64,power_law_1.2,0.03558399975299835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,64,power_law_1.2,0.03550719916820526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,1,power_law_1.2,5.547353744506836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,64,power_law_1.2,0.03532800078392029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,64,power_law_1.2,0.03530240058898926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,128,power_law_1.2,0.037785598635673524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,64,power_law_1.2,0.036345601081848145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,128,power_law_1.2,0.0383679986000061
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,128,power_law_1.2,0.03711360096931458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,64,power_law_1.2,0.03668479919433594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,1,power_law_1.2,8.032179260253907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,128,power_law_1.2,0.03819519877433777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,64,power_law_1.2,0.035673600435256955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,128,power_law_1.2,0.0380160003900528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,128,power_law_1.2,0.039155200123786926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,64,power_law_1.2,0.03704319894313812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,128,power_law_1.2,0.03930239975452423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,128,power_law_1.2,0.03949440121650696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,64,power_law_1.2,0.03678080141544342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,128,power_law_1.2,0.03994239866733551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,64,power_law_1.2,0.038201600313186646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,128,power_law_1.2,0.03976320028305054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,128,power_law_1.2,0.03968639969825745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,64,power_law_1.2,0.03963519930839539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,128,power_law_1.2,0.04079999923706055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,64,power_law_1.2,0.04033919870853424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,128,power_law_1.2,0.040627199411392215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,8,balanced,0.03370666752258936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,8,balanced,0.029631999631722767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,8,balanced,0.03149333347876867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,8,balanced,0.03173866619666418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,8,balanced,0.03751466671625773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,8,balanced,0.03969600051641464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,128,power_law_1.2,0.04069760143756866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,8,balanced,0.039434666434923805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,64,power_law_1.2,0.04535039961338043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,8,balanced,0.03993066648642222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,8,balanced,0.04062933226426443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,8,balanced,0.03955733279387156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,8,balanced,0.03774400055408478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,8,balanced,0.03940266619126002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,8,balanced,0.04142933338880539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,8,balanced,0.04155733436346054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,1,power_law_1.2,13.951155090332032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,8,balanced,0.043680002291997276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,8,balanced,0.04399466514587402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,8,balanced,0.049728001157442726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,8,balanced,0.050111999114354454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,128,power_law_1.2,0.044896000623703004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,8,balanced,0.05449066559473673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,8,balanced,0.07019199927647908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,8,balanced,0.07698666552702586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,8,balanced,0.0905013382434845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,8,balanced,0.11158933242162068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,64,power_law_1.2,0.045407998561859134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,8,balanced,0.13871999581654867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,8,balanced,0.16827199856440225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,8,balanced,0.22427733739217123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,128,power_law_1.2,0.04503679871559143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,8,balanced,0.2826346755027771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,8,balanced,0.3456159830093384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,8,balanced,0.510480006535848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,128,power_law_1.2,0.046342399716377256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,64,power_law_1.2,0.046649599075317384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,8,balanced,0.9399413267771403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,128,power_law_1.2,0.048403200507164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,64,power_law_1.2,0.049721598625183105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,128,power_law_1.2,0.05251200199127197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,64,power_law_1.2,0.052825599908828735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,128,power_law_1.2,0.05922560095787048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,128,power_law_1.2,0.06787840127944947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,64,power_law_1.2,0.06090880036354065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,128,power_law_1.2,0.07935360074043274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,64,power_law_1.2,0.06883199810981751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,128,power_law_1.2,0.09833599925041199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,64,power_law_1.2,0.09111679792404175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,128,power_law_1.2,0.14022400379180908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,64,power_law_1.2,0.10954879522323609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,128,power_law_1.2,0.1924288034439087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,64,power_law_1.2,0.1796671986579895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,128,power_law_1.2,0.2503743886947632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,128,power_law_1.2,0.33175039291381836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,64,power_law_1.2,0.22316160202026367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,2688,1856,6,128,1,128,power_law_1.2,0.4226367950439453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,64,power_law_1.2,0.35810561180114747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,2688,1856,6,128,1,128,power_law_1.2,0.5992959976196289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,64,power_law_1.2,0.4219007968902588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,64,power_law_1.2,0.4916224002838135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,2688,1856,6,128,1,128,power_law_1.2,1.2869952201843262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,1,2,power_law_1.2,0.09646080136299133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,64,power_law_1.2,0.8502016067504883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,1,2,power_law_1.2,0.10191999673843384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,1,2,power_law_1.2,0.11746560335159302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,4,power_law_1.01,0.04585599899291992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,1,2,power_law_1.2,0.14782079458236694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,1,2,power_law_1.2,0.1478656053543091
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,1,2,power_law_1.2,0.14984320402145385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,1,2,power_law_1.2,0.1504639983177185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,1,2,power_law_1.2,0.15074559450149536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,1,2,power_law_1.2,0.1535040020942688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,1,2,power_law_1.2,0.15510400533676147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,4,power_law_1.01,0.05244160294532776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,1,2,power_law_1.2,0.15587199926376344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,1,2,power_law_1.2,0.1672767996788025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,1,2,power_law_1.2,0.16509439945220947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,1,2,power_law_1.2,0.1752575993537903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,1,2,power_law_1.2,0.18877439498901366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,64,power_law_1.2,1.8950592041015626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,4,power_law_1.01,0.0523967981338501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,1,2,power_law_1.2,0.19875199794769288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,1,2,power_law_1.2,0.213753604888916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,1,2,power_law_1.2,0.23699839115142823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,1,2,power_law_1.2,0.312992000579834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,1,2,power_law_1.2,0.4333888053894043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,4,power_law_1.01,0.05572479963302612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,1,2,power_law_1.2,0.48373122215270997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,1,2,power_law_1.2,0.6987775802612305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,1,2,power_law_1.2,0.9032064437866211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,4,power_law_1.01,0.06881920099258423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,1,2,power_law_1.2,1.1538432121276856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,1,2,power_law_1.2,1.640025520324707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,4,power_law_1.01,0.07901440262794494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,1,2,power_law_1.2,2.3444032669067383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,4,power_law_1.01,0.09468799829483032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,1,2,power_law_1.2,3.2177150726318358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,4,power_law_1.01,0.09548799991607666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,4,power_law_1.01,0.09926400184631348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,1,2,power_law_1.2,3.6873729705810545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,2,2,balanced,0.03830400109291077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,2,2,balanced,0.0749120016892751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,4,power_law_1.01,0.10049279928207397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,2,2,balanced,0.08648533622423808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,2,2,balanced,0.08498133222262065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,2,2,balanced,0.08684800068537395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,2,2,balanced,0.08839466174443562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,2,2,balanced,0.08751466870307922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,2,2,balanced,0.09034132957458496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,2,2,balanced,0.08847999572753906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,2,2,balanced,0.0902239978313446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,2,2,balanced,0.09293333689371745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,2,2,balanced,0.09193600217501323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,2,2,balanced,0.09318932890892029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,2,2,balanced,0.09435733159383138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,2,2,balanced,0.11151466766993205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,4,power_law_1.01,0.10295679569244384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,2,2,balanced,0.11160000165303548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,2,2,balanced,0.11148800452550252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,1,2,power_law_1.2,6.612786865234375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,2,2,balanced,0.1341866652170817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,2,2,balanced,0.1402720014254252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,2,2,balanced,0.17197332779566446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,2,2,balanced,0.19549866517384848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,2,2,balanced,0.2536426583925883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,2,2,balanced,0.3207893371582031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,2,2,balanced,0.4472586711247762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,2,2,balanced,0.5659253199895223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,4,power_law_1.01,0.1062656044960022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,2,2,balanced,0.8246506849924723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,2,2,balanced,1.074677308400472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,2,2,balanced,1.3164746761322021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,4,power_law_1.01,0.11057920455932617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,2,2,balanced,2.0836586952209473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,2,2,balanced,4.067621231079102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,4,power_law_1.01,0.11391359567642212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,1,2,power_law_1.2,10.418386840820313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,4,2,power_law_1.01,0.06929280161857605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,4,power_law_1.01,0.12556159496307373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,4,2,power_law_1.01,0.08275840282440186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,4,2,power_law_1.01,0.09820799827575684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,4,power_law_1.01,0.12969599962234496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,4,2,power_law_1.01,0.12135039567947388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,4,2,power_law_1.01,0.11066880226135253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,4,power_law_1.01,0.13274240493774414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,4,2,power_law_1.01,0.12138240337371826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,4,2,power_law_1.01,0.12746880054473878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,4,2,power_law_1.01,0.1279039978981018
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,4,power_law_1.01,0.15252480506896973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,4,2,power_law_1.01,0.13062399625778198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,4,2,power_law_1.01,0.13086719512939454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,4,2,power_law_1.01,0.1312831997871399
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,4,power_law_1.01,0.168012797832489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,4,2,power_law_1.01,0.13230079412460327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,4,2,power_law_1.01,0.13241599798202514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,4,2,power_law_1.01,0.142848002910614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,4,2,power_law_1.01,0.14700160026550294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,4,power_law_1.01,0.19901440143585206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,4,2,power_law_1.01,0.1596287965774536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,4,2,power_law_1.01,0.17093119621276856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,4,2,power_law_1.01,0.2173504114151001
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,4,power_law_1.01,0.24067840576171876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,4,2,power_law_1.01,0.23719680309295654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,4,2,power_law_1.01,0.2933120012283325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,4,2,power_law_1.01,0.32791039943695066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,4,power_law_1.01,0.30198400020599364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,4,2,power_law_1.01,0.4299007892608643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,4,2,power_law_1.01,0.5714047908782959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,4,power_law_1.01,0.3802880048751831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,4,2,power_law_1.01,0.796563196182251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,4,2,power_law_1.01,1.0976384162902832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,4,power_law_1.01,0.47959041595458984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,4,2,power_law_1.01,1.6104000091552735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,4,2,power_law_1.01,2.0183488845825197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,4,power_law_1.01,0.5947328090667725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,4,2,power_law_1.01,2.25098876953125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,2,2,power_law_1.01,0.07952640056610108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,4,power_law_1.2,0.03731839954853058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,2,2,power_law_1.01,0.09561600089073181
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,4,power_law_1.01,0.8587967872619628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,2,2,power_law_1.01,0.09552639722824097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,2,2,power_law_1.01,0.131660795211792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,4,2,power_law_1.01,3.428607940673828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,2,2,power_law_1.01,0.1315135955810547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,4,power_law_1.2,0.03651840090751648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,2,2,power_law_1.01,0.13690240383148194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,2,2,power_law_1.01,0.14612480401992797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,2,2,power_law_1.01,0.14362879991531372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,4,power_law_1.2,0.03787519931793213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,2,2,power_law_1.01,0.14257919788360596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,2,2,power_law_1.01,0.1447167992591858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,4,power_law_1.01,1.131481647491455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,2,2,power_law_1.01,0.14298239946365357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,4,power_law_1.2,0.042419201135635375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,2,2,power_law_1.01,0.15306240320205688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,2,2,power_law_1.01,0.1536512017250061
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,2,2,power_law_1.01,0.16267520189285278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,4,power_law_1.2,0.05256959795951843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,2,2,power_law_1.01,0.1901695966720581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,2,2,power_law_1.01,0.1934656023979187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,4,power_law_1.2,0.06495360136032105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,2,2,power_law_1.01,0.21064960956573486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,4,2,power_law_1.01,7.712300872802734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,2,2,power_law_1.01,0.2364032030105591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,2,2,power_law_1.01,0.2691904067993164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,4,power_law_1.2,0.06940159797668458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,4,power_law_1.01,1.4164799690246581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,2,2,power_law_1.01,0.3350399971008301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,2,2,power_law_1.01,0.3749119997024536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,4,power_law_1.2,0.07069439888000488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,2,2,power_law_1.01,0.5360511779785156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,4,power_law_1.2,0.0711359977722168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,2,2,power_law_1.01,0.6844607830047608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,2,2,power_law_1.01,0.9597887992858887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,4,power_law_1.2,0.07220479846000671
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,2,2,power_law_1.01,1.268454360961914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,4,power_law_1.2,0.07416319847106934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,4,power_law_1.01,1.9292736053466797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,2,2,power_law_1.01,1.9215616226196288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,4,power_law_1.2,0.0750656008720398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,4,power_law_1.2,0.07744640111923218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,2,2,power_law_1.01,2.303219223022461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,4,power_law_1.2,0.08483200073242188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,4,power_law_1.2,0.09127680063247681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,2,2,power_law_1.01,2.944140815734863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,4,power_law_1.2,0.09527680277824402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,4,power_law_1.2,0.09919360280036926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,2,2,power_law_1.01,4.760255813598633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,4,power_law_1.01,3.872140884399414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,4,power_law_1.2,0.10588159561157226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,4,power_law_1.2,0.1191167950630188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,4,power_law_1.2,0.1275264024734497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,4,power_law_1.2,0.12692480087280272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,4,power_law_1.2,0.14842239618301392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,2,2,power_law_1.01,9.416339111328124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,4,power_law_1.2,0.17530239820480348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,4,power_law_1.2,0.2420543909072876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,4,power_law_1.2,0.2757951974868774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,1,balanced,0.1564906636873881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,1,balanced,0.26074133316675824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,4,power_law_1.2,0.43927679061889646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,1,balanced,0.46811731656392414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,1,balanced,0.8817333380381266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,1,balanced,1.711674690246582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,4,power_law_1.2,0.49484801292419434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,1,balanced,2.1253652572631836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,1,balanced,2.138458728790283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,1,balanced,2.14790932337443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,2,4,power_law_1.2,0.6183936119079589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,1,balanced,2.1580799420674643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,1,balanced,2.156538645426432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,1,balanced,2.17957337697347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,2,4,power_law_1.2,0.899180793762207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,1,balanced,2.18230930964152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,1,balanced,2.1910667419433594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,1,balanced,2.204517364501953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,1,balanced,2.2307039896647134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,1,balanced,2.241290728251139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,2,4,power_law_1.2,1.7542144775390625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,1,balanced,2.263007958730062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,1,balanced,2.2945173581441245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,8,balanced,0.05233600238958994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,8,balanced,0.08456533153851827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,8,balanced,0.09073600172996521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,1,balanced,2.3416479428609214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,8,balanced,0.13529599706331888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,8,balanced,0.20037867625554404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,8,balanced,0.2511199911435445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,8,balanced,0.2521546681722005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,8,balanced,0.2564586599667867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,8,balanced,0.25701334079106647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,8,balanced,0.25837333997090656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,1,balanced,2.9989706675211587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,8,balanced,0.25709333022435504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,8,balanced,0.259770671526591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,8,balanced,0.2621333400408427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,8,balanced,0.26207466920216876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,8,balanced,0.26955199241638184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,1,balanced,2.565626621246338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,8,balanced,0.2722559968630473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,8,balanced,0.2770880063374837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,8,balanced,0.3252799908320109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,8,balanced,0.30511999130249023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,8,balanced,0.4192800124486287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,8,balanced,0.36262933413187665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,8,balanced,0.562666654586792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,1,balanced,3.913146654764811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,8,balanced,0.4662346839904785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,8,balanced,0.7702453136444092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,8,balanced,0.6662720044453939
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,1,balanced,2.9814879099527993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,4,power_law_1.2,0.032467201352119446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,8,balanced,0.9661813577016195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,8,balanced,1.0662293434143066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,2,8,balanced,1.347914695739746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,4,power_law_1.2,0.03118079900741577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,1,balanced,4.863834698994954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,2,8,balanced,1.928378740946452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,4,power_law_1.2,0.03264000117778778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,1,balanced,4.473850568135579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,2,8,balanced,3.904853185017904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,4,power_law_1.2,0.03395200073719025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,4,power_law_1.2,0.03549439907073974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,1,balanced,7.5596160888671875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,4,power_law_1.2,0.04023039937019348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,4,power_law_1.2,0.041715198755264284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,1,balanced,8.02901840209961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,4,power_law_1.2,0.042745599150657655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,4,power_law_1.2,0.04337919950485229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,4,power_law_1.2,0.044064000248909
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,1,balanced,9.035696029663086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,4,power_law_1.2,0.04668799936771393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,4,power_law_1.2,0.049644801020622256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,4,power_law_1.2,0.052083200216293334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,128,power_law_1.01,0.03587839901447296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,4,power_law_1.2,0.05246080160140991
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,1,balanced,14.307914733886719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,128,power_law_1.01,0.036108800768852235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,4,power_law_1.2,0.057894402742385866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,128,power_law_1.01,0.03603839874267578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,4,power_law_1.2,0.06094080209732056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,128,power_law_1.01,0.03622399866580963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,4,power_law_1.2,0.06755200028419495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,128,power_law_1.01,0.036371201276779175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,4,power_law_1.2,0.08296319842338562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,128,power_law_1.01,0.037088000774383546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,4,power_law_1.2,0.08657280206680298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,128,power_law_1.01,0.03743360042572021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,4,power_law_1.2,0.11470719575881957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,128,power_law_1.01,0.037427198886871335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,4,power_law_1.2,0.13287680149078368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,128,power_law_1.01,0.03759360015392303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,4,power_law_1.2,0.17402880191802977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,128,power_law_1.01,0.03804160058498383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,1,balanced,26.89244842529297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,4,8,power_law_1.2,0.037747201323509214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,4,8,power_law_1.2,0.03858560025691986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,4,power_law_1.2,0.2595135927200317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,4,8,power_law_1.2,0.03697279989719391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,128,power_law_1.01,0.03943679928779602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,4,8,power_law_1.2,0.03651840090751648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,4,8,power_law_1.2,0.03768320083618164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,128,power_law_1.01,0.03923200070858002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,4,8,power_law_1.2,0.03776639997959137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,4,power_law_1.2,0.325548791885376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,4,8,power_law_1.2,0.0390720009803772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,4,8,power_law_1.2,0.03819519877433777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,128,power_law_1.01,0.039724799990653994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,4,8,power_law_1.2,0.038867199420928956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,4,8,power_law_1.2,0.0403903990983963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,4,power_law_1.2,0.4306687831878662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,4,8,power_law_1.2,0.04053759872913361
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,128,power_law_1.01,0.04038400053977966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,4,8,power_law_1.2,0.04835200011730194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,4,8,power_law_1.2,0.04707199931144714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,4,8,power_law_1.2,0.05121279954910278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,128,power_law_1.01,0.04309119880199432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,4,8,power_law_1.2,0.0580735981464386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,4,8,power_law_1.2,0.061868798732757566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,4,power_law_1.2,0.6931327819824219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,4,8,power_law_1.2,0.06431999802589417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,128,power_law_1.01,0.04392319917678833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,4,8,power_law_1.2,0.0778432011604309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,4,8,power_law_1.2,0.0907584011554718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,128,power_law_1.01,0.04490880072116852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,4,8,power_law_1.2,0.11002240180969239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,4,8,power_law_1.2,0.12320640087127685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,4,power_law_1.2,0.8631872177124024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,4,8,power_law_1.2,0.15950080156326293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,128,power_law_1.01,0.04913919866085052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,4,8,power_law_1.2,0.2006848096847534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,4,8,power_law_1.2,0.28123519420623777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,128,power_law_1.01,0.05467519760131836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,4,8,power_law_1.2,0.3496511936187744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,128,power_law_1.01,0.060147202014923094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,4,power_law_1.2,1.1315199851989746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,4,8,power_law_1.2,0.4958335876464844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,4,8,power_law_1.2,0.6380671977996826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,128,power_law_1.01,0.06870399713516236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,4,8,power_law_1.2,0.8043711662292481
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,128,power_law_1.01,0.09408640265464782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,4,8,power_law_1.2,1.193843173980713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,4,power_law_1.2,1.9442176818847656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,128,power_law_1.01,0.12139519453048705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,128,power_law_1.01,0.16714880466461182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,4,8,power_law_1.2,2.1718656539916994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,128,power_law_1.01,0.2038719892501831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,128,power_law_1.01,0.2994688034057617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,4,power_law_1.2,4.149875259399414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,128,power_law_1.01,0.37918078899383545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,128,power_law_1.01,0.4479872226715088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,8,power_law_1.2,0.04704639911651611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,128,power_law_1.01,0.7901760101318359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,8,power_law_1.2,0.04425599873065948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,8,power_law_1.2,0.043398401141166686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,8,power_law_1.2,0.04534400105476379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,128,power_law_1.01,1.4524288177490234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,8,power_law_1.2,0.04774399995803833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,8,power_law_1.2,0.05379199981689453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,8,power_law_1.2,0.05566080212593079
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,8,power_law_1.2,0.055936002731323244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,32,power_law_1.2,0.03761920034885406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,32,power_law_1.2,0.03288959860801697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,8,power_law_1.2,0.056467199325561525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,32,power_law_1.2,0.029868799448013305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,32,power_law_1.2,0.029843199253082275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,8,power_law_1.2,0.06044160127639771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,32,power_law_1.2,0.03059839904308319
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,8,16,power_law_1.2,0.054553598165512085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,32,power_law_1.2,0.032025599479675294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,8,power_law_1.2,0.06335999965667724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,8,16,power_law_1.2,0.049414399266242984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,32,power_law_1.2,0.03219839930534363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,32,power_law_1.2,0.03229439854621887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,8,16,power_law_1.2,0.049779200553894044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,8,power_law_1.2,0.06439039707183838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,32,power_law_1.2,0.032953599095344545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,8,16,power_law_1.2,0.0511680006980896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,8,power_law_1.2,0.06555520296096802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,32,power_law_1.2,0.033843201398849485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,8,16,power_law_1.2,0.05148800015449524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,32,power_law_1.2,0.03511680066585541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,8,power_law_1.2,0.0690496027469635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,8,16,power_law_1.2,0.05370240211486817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,32,power_law_1.2,0.03685120046138764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,32,power_law_1.2,0.03823359906673431
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,8,16,power_law_1.2,0.05422080159187317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,8,power_law_1.2,0.07719039916992188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,32,power_law_1.2,0.036025598645210266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,8,16,power_law_1.2,0.05604479908943176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,32,power_law_1.2,0.039904001355171206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,8,power_law_1.2,0.07933440208435058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,8,16,power_law_1.2,0.055936002731323244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,32,power_law_1.2,0.04222080111503601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,64,power_law_1.01,0.12199039459228515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,8,16,power_law_1.2,0.05571200251579285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,8,power_law_1.2,0.0812928020954132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,32,power_law_1.2,0.046291199326515195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,8,16,power_law_1.2,0.057120001316070555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,32,power_law_1.2,0.05186560153961182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,64,power_law_1.01,0.1117184042930603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,8,power_law_1.2,0.09639679789543151
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,32,power_law_1.2,0.054073601961135864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,8,16,power_law_1.2,0.058412802219390866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,32,power_law_1.2,0.0636352002620697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,8,16,power_law_1.2,0.05987840294837952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,8,power_law_1.2,0.11393280029296875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,64,power_law_1.01,0.11643520593643189
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,32,power_law_1.2,0.06972799897193908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,8,16,power_law_1.2,0.062457597255706786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,8,power_law_1.2,0.14194560050964355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,32,power_law_1.2,0.08636800050735474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,8,16,power_law_1.2,0.06702079772949218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,64,power_law_1.01,0.07237759828567505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,32,power_law_1.2,0.11290240287780762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,8,power_law_1.2,0.17209600210189818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,8,16,power_law_1.2,0.07021440267562866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,32,power_law_1.2,0.15360000133514404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,8,16,power_law_1.2,0.07550719976425171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,8,power_law_1.2,0.22204160690307617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,64,power_law_1.01,0.08572800159454345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,32,power_law_1.2,0.18060799837112426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,8,16,power_law_1.2,0.08906880021095276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,32,power_law_1.2,0.23962879180908203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,8,16,power_law_1.2,0.10402560234069824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,8,power_law_1.2,0.2731071949005127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,64,power_law_1.01,0.09318400025367737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,8,16,power_law_1.2,0.12687360048294066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,32,power_law_1.2,0.3226560115814209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,8,power_law_1.2,0.3837631940841675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,8,16,power_law_1.2,0.1436735987663269
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,64,power_law_1.01,0.0951744019985199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,1,32,power_law_1.2,0.39723520278930663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,8,16,power_law_1.2,0.1889407992362976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,8,power_law_1.2,0.5631487846374512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,64,power_law_1.01,0.10191359519958496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,8,16,power_law_1.2,0.22419838905334472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,1,32,power_law_1.2,0.614796781539917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,8,16,power_law_1.2,0.3016063928604126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,8,power_law_1.2,0.7812928199768067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,64,power_law_1.01,0.10854400396347046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,8,16,power_law_1.2,0.3991296052932739
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,1,32,power_law_1.2,1.1924544334411622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,8,power_law_1.2,1.0590784072875976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,64,power_law_1.01,0.11166080236434936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,8,16,power_law_1.2,0.5418240070343018
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,64,power_law_1.01,0.10990079641342163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,8,16,power_law_1.2,0.7651840209960937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,8,power_law_1.2,1.212332820892334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,64,power_law_1.01,0.1047104001045227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,8,16,power_law_1.2,0.8663871765136719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,64,power_law_1.01,0.11509759426116943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,8,power_law_1.2,2.0401727676391603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,8,16,power_law_1.2,1.4782848358154297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,64,power_law_1.01,0.11584000587463379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,64,power_law_1.01,0.12094080448150635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,8,power_law_1.2,3.9494529724121095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,64,power_law_1.01,0.12122880220413208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,8,16,power_law_1.2,3.316787338256836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,64,power_law_1.01,0.12993279695510865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,64,power_law_1.01,0.14042240381240845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,128,balanced,0.058261334896087646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,128,balanced,0.05816533168156942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,128,balanced,0.0580320010582606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,128,balanced,0.05738133192062378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,128,balanced,0.058304001887639366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,128,balanced,0.06238933404286703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,128,balanced,0.06774400174617767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,128,balanced,0.06635733445485432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,64,power_law_1.01,0.1453760027885437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,128,balanced,0.06645866731802623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,128,balanced,0.06629866858323415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,128,balanced,0.06795733173688252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,128,balanced,0.06865066786607106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,128,balanced,0.06625066697597504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,128,balanced,0.06861866513888042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,128,balanced,0.07648533085982005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,128,balanced,0.07518933216730754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,128,balanced,0.08331199983755748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,128,balanced,0.08685866991678874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,128,balanced,0.09291199843088786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,128,balanced,0.10967466235160828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,128,balanced,0.12595733006795248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,128,balanced,0.16153066356976828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,128,balanced,0.1885813275973002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,128,balanced,0.24702932437260947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,128,balanced,0.3077920079231262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,64,power_law_1.01,0.17155200242996216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,128,balanced,0.4235573212305705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,128,balanced,0.541381319363912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,128,balanced,0.6625920136769613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,128,balanced,1.0180693467458088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,128,balanced,1.9921174049377441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,64,power_law_1.01,0.20413439273834227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,64,power_law_1.01,0.265337610244751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,64,power_law_1.01,0.3180799961090088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,64,power_law_1.01,0.4256896018981934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,8,8,power_law_1.01,0.034720000624656674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,8,8,power_law_1.01,0.033670398592948916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,32,8,power_law_1.2,0.030342400074005127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,32,8,power_law_1.2,0.029049599170684816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,8,8,power_law_1.01,0.03459199965000152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,32,8,power_law_1.2,0.029523199796676634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,32,8,power_law_1.2,0.029471999406814574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,8,8,power_law_1.01,0.032876798510551454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,64,power_law_1.01,0.5322303771972656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,32,8,power_law_1.2,0.029337599873542786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,32,8,power_law_1.2,0.029996800422668456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,8,8,power_law_1.01,0.035743999481201175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,32,8,power_law_1.2,0.03059839904308319
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,8,8,power_law_1.01,0.03353599905967712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,32,8,power_law_1.2,0.03080959916114807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,32,8,power_law_1.2,0.030943998694419862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,8,8,power_law_1.01,0.0344895988702774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,32,8,power_law_1.2,0.03139199912548065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,32,8,power_law_1.2,0.031916800141334536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,8,8,power_law_1.01,0.03597440123558045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,32,8,power_law_1.2,0.03333759903907776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,32,8,power_law_1.2,0.033881598711013795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,8,8,power_law_1.01,0.03993600010871887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,64,power_law_1.01,0.7596608161926269
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,32,8,power_law_1.2,0.034297600388526917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,32,8,power_law_1.2,0.04069119989871979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,8,8,power_law_1.01,0.03878400027751923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,32,8,power_law_1.2,0.04259839951992035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,8,8,power_law_1.01,0.039001598954200745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,32,8,power_law_1.2,0.04413439929485321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,32,8,power_law_1.2,0.05329279899597168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,8,8,power_law_1.01,0.042182400822639465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,32,8,power_law_1.2,0.05480319857597351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,8,8,power_law_1.01,0.04102399945259094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,32,8,power_law_1.2,0.06440960168838501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,32,8,power_law_1.2,0.07467520236968994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,8,8,power_law_1.01,0.04348799884319306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,32,8,power_law_1.2,0.10164480209350586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,64,power_law_1.01,1.0077759742736816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,8,8,power_law_1.01,0.0537663996219635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,32,8,power_law_1.2,0.127839994430542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,8,8,power_law_1.01,0.05729280114173889
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,32,8,power_law_1.2,0.18853119611740113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,8,8,power_law_1.01,0.05611519813537598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,32,8,power_law_1.2,0.2290816068649292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,8,8,power_law_1.01,0.07393280267715455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,32,8,power_law_1.2,0.316595196723938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,8,8,power_law_1.01,0.08303359746932984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,32,8,power_law_1.2,0.41265277862548827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,32,8,power_law_1.01,0.04916479885578155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,8,8,power_law_1.01,0.10468480587005616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,32,8,power_law_1.2,0.5327231884002686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,64,power_law_1.01,1.2102720260620117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,8,8,power_law_1.01,0.13345279693603515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,32,8,power_law_1.01,0.048588800430297854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,8,8,power_law_1.01,0.17107839584350587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,32,8,power_law_1.2,0.8107968330383301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,8,8,power_law_1.01,0.20337278842926027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,32,8,power_law_1.01,0.04288640022277832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,8,8,power_law_1.01,0.32182400226593016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,32,8,power_law_1.2,1.5265024185180665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,32,8,power_law_1.01,0.046105599403381346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,8,8,power_law_1.01,0.3881216049194336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,8,8,power_law_1.01,0.5465087890625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,32,8,power_law_1.01,0.04767360091209412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,64,power_law_1.01,1.9139520645141601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,8,8,power_law_1.01,0.6949120044708252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,32,8,power_law_1.01,0.052051198482513425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,8,8,power_law_1.01,0.8279808044433594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,32,8,power_law_1.01,0.0529151976108551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,8,8,power_law_1.01,1.368876838684082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,32,8,power_law_1.01,0.055244797468185426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,32,8,power_law_1.01,0.05374079942703247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,8,8,power_law_1.01,2.53939208984375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,32,8,power_law_1.01,0.05539199709892273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,32,8,power_law_1.01,0.055871999263763426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,64,power_law_1.01,3.8572288513183595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,32,8,power_law_1.01,0.0572160005569458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,32,8,power_law_1.01,0.05978239774703979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,32,8,power_law_1.01,0.058790397644042966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,32,8,power_law_1.01,0.06803839802742004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,32,8,power_law_1.01,0.07206400036811829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,32,8,power_law_1.01,0.0774399995803833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,32,8,power_law_1.01,0.09367039799690247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,32,8,power_law_1.01,0.09619200229644775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,32,8,power_law_1.01,0.11864320039749146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,32,balanced,0.05787733197212219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,32,balanced,0.05199466645717621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,32,8,power_law_1.01,0.14873600006103516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,32,balanced,0.05364799996217092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,32,balanced,0.05216533442338308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,32,balanced,0.06628799935181935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,32,balanced,0.0755626658598582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,32,balanced,0.07621866464614868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,32,balanced,0.07701866825421651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,32,balanced,0.07841599980990092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,32,balanced,0.07719466586907704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,32,balanced,0.07938666641712189
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,32,balanced,0.07859733204046886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,32,balanced,0.07892799874146779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,32,balanced,0.08070399860541026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,32,8,power_law_1.01,0.1838271975517273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,32,balanced,0.08508267005284627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,32,balanced,0.08570133646329244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,32,balanced,0.08820799986521403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,32,balanced,0.09372267127037048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,32,balanced,0.09835199515024821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,32,balanced,0.11167466640472412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,32,balanced,0.12422399719556172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,32,balanced,0.14756799737612405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,32,balanced,0.15781866510709128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,32,balanced,0.20202134052912393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,32,8,power_law_1.01,0.20785920619964598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,32,balanced,0.23513599236806235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,32,balanced,0.33217066526412964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,32,balanced,0.404698650042216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,32,balanced,0.47200532754262287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,32,balanced,0.7162346839904785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,32,8,power_law_1.01,0.28796799182891847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,32,balanced,1.3614880243937175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,32,8,power_law_1.01,0.3626176118850708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,32,8,power_law_1.01,0.4949503898620605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,4,32,balanced,0.03758399933576584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,4,32,balanced,0.03364799916744232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,4,32,balanced,0.033626665671666466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,4,32,balanced,0.03391999999682108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,4,32,balanced,0.03552533437808355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,4,32,balanced,0.037685332198937736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,4,32,balanced,0.03561066587766012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,4,32,balanced,0.035530666510264076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,4,32,balanced,0.03723733375469843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,4,32,balanced,0.03664000084002813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,4,32,balanced,0.0355679988861084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,4,32,balanced,0.03754133234421412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,4,32,balanced,0.03605333218971888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,4,32,balanced,0.039781334499518074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,4,32,balanced,0.040991999208927155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,4,32,balanced,0.04151466737190882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,4,32,balanced,0.041797334949175514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,32,8,power_law_1.01,0.677791976928711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,4,32,balanced,0.04419733087221781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,4,32,balanced,0.04410133262475332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,4,32,balanced,0.04585599899291992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,4,32,balanced,0.04831466575463613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,32,1,power_law_1.01,0.0531391978263855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,4,32,balanced,0.05194133520126343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,4,32,balanced,0.05588266750176748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,4,32,balanced,0.06869866450627644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,4,32,balanced,0.08167999982833862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,4,32,balanced,0.10132799545923869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,4,32,balanced,0.12929600477218628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,4,32,balanced,0.15240533153216043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,4,32,balanced,0.2109066645304362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,4,32,balanced,0.3672586679458618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,32,8,power_law_1.01,0.775545597076416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,32,1,power_law_1.01,0.053260797262191774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,32,8,power_law_1.01,1.242784023284912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,32,1,power_law_1.01,0.06142079830169678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,32,1,power_law_1.01,0.07958400249481201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,32,8,power_law_1.01,2.351968002319336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,32,1,power_law_1.01,0.11019519567489625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,32,1,power_law_1.01,0.13059200048446656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,16,1,power_law_1.2,0.03440000116825104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,16,1,power_law_1.2,0.036320000886917114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,16,1,power_law_1.2,0.03811199963092804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,16,1,power_law_1.2,0.053529602289199826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,32,1,power_law_1.01,0.1626304030418396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,16,1,power_law_1.2,0.05427200198173523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,16,1,power_law_1.2,0.05752320289611816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,16,1,power_law_1.2,0.05978239774703979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,16,1,power_law_1.2,0.05973119735717773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,16,1,power_law_1.2,0.06264320015907288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,16,1,power_law_1.2,0.06249600052833557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,32,1,power_law_1.01,0.1712767958641052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,16,1,power_law_1.2,0.0627135992050171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,16,1,power_law_1.2,0.06743040084838867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,16,1,power_law_1.2,0.06940159797668458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,16,1,power_law_1.2,0.06928640007972717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,16,1,power_law_1.2,0.08076800107955932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,16,1,power_law_1.2,0.08416640162467956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,16,1,power_law_1.2,0.09405440092086792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,32,1,power_law_1.01,0.17965439558029175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,16,1,power_law_1.2,0.11079679727554322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,16,1,power_law_1.2,0.11811200380325318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,16,1,power_law_1.2,0.14625279903411864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,16,1,power_law_1.2,0.17254400253295898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,16,1,power_law_1.2,0.23347840309143067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,16,1,power_law_1.2,0.2852735996246338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,32,1,power_law_1.01,0.1801792025566101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,16,1,power_law_1.2,0.40375680923461915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,16,1,power_law_1.2,0.5145472049713135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,32,2,power_law_1.2,0.05023999810218811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,16,1,power_law_1.2,0.734496021270752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,32,1,power_law_1.01,0.19205119609832763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,16,1,power_law_1.2,0.9520959854125977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,32,2,power_law_1.2,0.05263360142707825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,16,1,power_law_1.2,1.1754048347473145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,32,1,power_law_1.01,0.20067839622497557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,16,1,power_law_1.2,1.8227327346801758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,32,2,power_law_1.2,0.05907840132713318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,32,1,power_law_1.01,0.2087552070617676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,16,1,power_law_1.2,3.5801345825195314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,32,2,power_law_1.2,0.06279680132865906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,1,power_law_1.2,0.047660800814628604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,32,1,power_law_1.01,0.21994240283966066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,1,power_law_1.2,0.047193598747253415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,32,2,power_law_1.2,0.0762112021446228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,1,power_law_1.2,0.050444799661636355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,32,1,power_law_1.01,0.236627197265625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,1,power_law_1.2,0.05639680027961731
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,1,power_law_1.2,0.06235520243644714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,32,2,power_law_1.2,0.09064319729804993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,1,power_law_1.2,0.08770560026168824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,32,1,power_law_1.01,0.24491519927978517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,1,power_law_1.2,0.09283199906349182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,32,2,power_law_1.2,0.10947200059890747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,1,power_law_1.2,0.09630079865455628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,32,1,power_law_1.01,0.2494976043701172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,1,power_law_1.2,0.0959231972694397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,32,2,power_law_1.2,0.12457599639892578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,1,power_law_1.2,0.09923200011253357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,32,1,power_law_1.01,0.2954240083694458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,1,power_law_1.2,0.10169600248336792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,1,power_law_1.2,0.10603519678115844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,32,2,power_law_1.2,0.13248640298843384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,32,1,power_law_1.01,0.3380800008773804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,1,power_law_1.2,0.10890239477157593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,1,power_law_1.2,0.1145792007446289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,32,2,power_law_1.2,0.13470079898834228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,1,power_law_1.2,0.12108800411224366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,32,1,power_law_1.01,0.43778557777404786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,1,power_law_1.2,0.12630399465560913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,32,2,power_law_1.2,0.13738240003585817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,1,power_law_1.2,0.13897600173950195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,32,1,power_law_1.01,0.4823808193206787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,1,power_law_1.2,0.16516480445861817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,32,2,power_law_1.2,0.1463871955871582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,1,power_law_1.2,0.18774399757385254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,16,balanced,0.03779733429352442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,16,balanced,0.03146666785081228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,16,balanced,0.03072533259789149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,16,balanced,0.029658667743206024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,16,balanced,0.03164800008138021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,16,balanced,0.033359999457995095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,32,1,power_law_1.01,0.6379648208618164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,16,balanced,0.035599999129772186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,1,power_law_1.2,0.23952000141143798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,16,balanced,0.03531199942032496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,16,balanced,0.033520000676314034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,16,balanced,0.033530667424201965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,16,balanced,0.03379733363787333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,16,balanced,0.033930666744709015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,16,balanced,0.03547733277082443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,16,balanced,0.035349334279696144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,16,balanced,0.037845333417256675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,16,balanced,0.03941333293914795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,16,balanced,0.037845333417256675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,16,balanced,0.043493335445721946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,16,balanced,0.04009066770474116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,16,balanced,0.045781334241231285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,1,power_law_1.2,0.2875967979431152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,16,balanced,0.04598399996757507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,32,2,power_law_1.2,0.15024640560150146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,16,balanced,0.05264533559481303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,16,balanced,0.05602666735649109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,16,balanced,0.07468800246715546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,16,balanced,0.08261333405971527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,16,balanced,0.10523733496665955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,16,balanced,0.12553067008654276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,4,16,balanced,0.14229866862297058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,4,16,balanced,0.19079999128977457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,4,16,balanced,0.32681600252787274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,1,power_law_1.2,0.39422719478607177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,32,1,power_law_1.01,0.7942783832550049
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,32,2,power_law_1.2,0.15622400045394896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,1,power_law_1.2,0.5005055904388428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,1,power_law_1.2,0.7114111900329589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,32,1,power_law_1.01,1.0807999610900878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,32,2,power_law_1.2,0.16373759508132935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,1,power_law_1.2,0.913548755645752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,32,2,power_law_1.2,0.17547520399093627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,32,1,power_law_1.01,1.3652480125427247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,1,power_law_1.2,1.3242688179016113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,32,2,power_law_1.2,0.18877439498901366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,1,power_law_1.2,1.732467269897461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,32,1,power_law_1.01,1.9634687423706054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,32,2,power_law_1.2,0.2132159948348999
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,1,power_law_1.2,2.142201614379883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,32,2,power_law_1.2,0.24197759628295898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,32,1,power_law_1.01,2.5327552795410155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,32,2,power_law_1.2,0.3066431999206543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,1,power_law_1.2,3.3310718536376953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,32,2,power_law_1.2,0.35272960662841796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,32,1,power_law_1.01,3.111033630371094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,128,balanced,0.05857066810131073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,128,balanced,0.05907199780146281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,128,balanced,0.05638933181762695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,128,balanced,0.05819199979305267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,128,balanced,0.06027733286221822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,128,balanced,0.06047999858856201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,32,2,power_law_1.2,0.45201921463012695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,128,balanced,0.06021333237489065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,128,balanced,0.06020266811052958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,128,balanced,0.060458665092786155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,128,balanced,0.06019733349482218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,128,balanced,0.0603359987338384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,128,balanced,0.06393600006898244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,128,balanced,0.06418666740258534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,128,balanced,0.0642133355140686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,128,balanced,0.06899199883143108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,128,balanced,0.06867200136184692
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,128,balanced,0.07276799778143565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,128,balanced,0.08248533308506012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,128,balanced,0.0888426701227824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,128,balanced,0.11090667049090068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,128,balanced,0.12403200070063274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,128,balanced,0.1689173380533854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,128,balanced,0.16831467549006143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,128,balanced,0.2097866733868917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,1,power_law_1.2,6.551398468017578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,128,balanced,0.23145065704981485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,32,2,power_law_1.2,0.581983995437622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,128,balanced,0.3091893394788106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,128,balanced,0.36580268541971844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,1,128,balanced,0.4360640048980713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,32,1,power_law_1.01,4.8625022888183596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,1,128,balanced,0.633786678314209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,1,128,balanced,1.3706453641255696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,32,2,power_law_1.2,0.782316780090332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,32,2,power_law_1.2,0.8989631652832031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,32,2,power_law_1.2,1.331321620941162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,32,1,power_law_1.01,9.457843017578124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,32,2,power_law_1.2,1.6761920928955079
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,32,2,power_law_1.2,2.0643199920654296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,1,balanced,0.05565866827964783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,1,balanced,0.0780266672372818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,32,2,power_law_1.2,3.4680191040039063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,1,balanced,0.11905599633852641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,1,balanced,0.20626133680343628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,1,balanced,0.37213865915934247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,1,balanced,0.37328000863393146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,1,balanced,0.3758773406346639
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,1,balanced,0.3781333367029826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,1,balanced,0.38236268361409503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,1,balanced,0.3856746753056844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,1,balanced,0.3859626849492391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,1,balanced,0.3930293321609497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,1,balanced,0.39556801319122314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,1,balanced,0.4013173182805379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,1,balanced,0.41234131654103595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,1,balanced,0.4184639851252238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,1,balanced,0.4357973337173462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,1,balanced,0.47280001640319824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,8,power_law_1.2,0.050348800420761106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,1,balanced,0.5131306648254395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,1,balanced,0.566869338353475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,1,balanced,0.6105173428853353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,1,balanced,0.7804213364919027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,1,balanced,0.8728480339050293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,1,balanced,1.2249493598937988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,1,balanced,1.4584372838338215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,1,balanced,2.0865707397460938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,8,power_law_1.2,0.0747648000717163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,1,balanced,2.736053466796875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,1,balanced,3.4066880544026694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,32,2,power_law_1.2,6.29117431640625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,8,power_law_1.2,0.06786559820175171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,1,balanced,5.268896102905273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,8,power_law_1.2,0.058976000547409056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,1,balanced,10.178976058959961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,16,power_law_1.2,0.04813440144062042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,8,power_law_1.2,0.06829439997673034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,8,power_law_1.2,0.08487679958343505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,16,power_law_1.2,0.058316802978515624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,8,power_law_1.2,0.08823040127754211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,16,power_law_1.2,0.048230400681495665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,8,power_law_1.2,0.09742720127105713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,8,power_law_1.2,0.1093951940536499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,16,power_law_1.2,0.04633600115776062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,8,power_law_1.2,0.11274240016937256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,8,power_law_1.2,0.11079039573669433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,16,power_law_1.2,0.0474047988653183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,8,power_law_1.2,0.11672320365905761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,128,power_law_1.2,0.050937598943710326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,16,power_law_1.2,0.05026559829711914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,2,2,power_law_1.01,0.054054397344589236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,8,power_law_1.2,0.12070399522781372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,2,2,power_law_1.01,0.057068800926208495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,2,2,power_law_1.01,0.06951040029525757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,2,2,power_law_1.01,0.0832319974899292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,2,2,power_law_1.01,0.08406400084495544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,2,2,power_law_1.01,0.08295040130615235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,128,power_law_1.2,0.05299199819564819
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,2,2,power_law_1.01,0.08766080141067505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,8,power_law_1.2,0.1235584020614624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,2,2,power_law_1.01,0.09125760197639465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,2,2,power_law_1.01,0.09264640212059021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,16,power_law_1.2,0.052851200103759766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,2,2,power_law_1.01,0.09007999897003174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,2,2,power_law_1.01,0.09182080030441284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,2,2,power_law_1.01,0.09132800102233887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,2,2,power_law_1.01,0.09642879962921143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,2,2,power_law_1.01,0.10860799551010132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,8,power_law_1.2,0.13493119478225707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,2,2,power_law_1.01,0.12151679992675782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,128,power_law_1.2,0.05002880096435547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,2,2,power_law_1.01,0.12513920068740844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,2,2,power_law_1.01,0.1366976022720337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,2,2,power_law_1.01,0.16784000396728516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,2,2,power_law_1.01,0.17368320226669312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,16,power_law_1.2,0.05544319748878479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,2,2,power_law_1.01,0.24361600875854492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,8,power_law_1.2,0.132915198802948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,2,2,power_law_1.01,0.24970240592956544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,2,2,power_law_1.01,0.34879360198974607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,128,power_law_1.2,0.05079039931297302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,2,2,power_law_1.01,0.4862080097198486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,2,2,power_law_1.01,0.6213568210601806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,8,power_law_1.2,0.1492416024208069
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,2,2,power_law_1.01,0.8135359764099122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,16,power_law_1.2,0.057062399387359616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,2,2,power_law_1.01,1.0245120048522949
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,128,power_law_1.2,0.05115519762039185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,8,power_law_1.2,0.17738879919052125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,2,2,power_law_1.01,1.6632896423339845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,2,2,power_law_1.01,2.007673645019531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,16,power_law_1.2,0.057811200618743896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,128,power_law_1.2,0.05256959795951843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,8,power_law_1.2,0.18629759550094604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,2,2,power_law_1.01,2.5920127868652343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,8,power_law_1.2,0.229420804977417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,128,power_law_1.2,0.05256959795951843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,16,power_law_1.2,0.059359997510910034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,2,2,power_law_1.01,5.473523330688477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,8,power_law_1.2,0.24681599140167237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,128,power_law_1.2,0.0536191999912262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,16,power_law_1.2,0.0636672019958496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,8,power_law_1.2,0.33152639865875244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,128,power_law_1.2,0.05351679921150208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,16,power_law_1.2,0.06313599944114685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,8,power_law_1.2,0.40471677780151366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,128,power_law_1.2,0.054636800289154054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,8,power_law_1.2,0.5513663768768311
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,16,power_law_1.2,0.06392319798469544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,128,power_law_1.2,0.054745602607727054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,8,power_law_1.2,0.6527552127838134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,16,power_law_1.2,0.07403519749641418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,128,power_law_1.2,0.05848960280418396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,8,power_law_1.2,0.9449600219726563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,16,power_law_1.2,0.07763199806213379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,128,power_law_1.2,0.060102397203445436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,16,power_law_1.2,0.08702080249786377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,8,power_law_1.2,1.2052736282348633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,128,power_law_1.2,0.0623744010925293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,16,power_law_1.01,0.05541759729385376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,16,power_law_1.01,0.06216319799423218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,16,power_law_1.2,0.10704640150070191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,128,power_law_1.2,0.06800640225410462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,8,power_law_1.2,1.4335935592651368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,16,power_law_1.01,0.048902401328086854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,16,power_law_1.01,0.04948480129241943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,128,power_law_1.2,0.07097600102424621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,16,power_law_1.2,0.11599999666213989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,16,power_law_1.01,0.05260159969329834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,16,power_law_1.01,0.05477759838104248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,16,power_law_1.01,0.05511680245399475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,8,power_law_1.2,2.181043243408203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,128,power_law_1.2,0.0759168028831482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,16,power_law_1.2,0.14325120449066162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,16,power_law_1.01,0.058508801460266116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,16,power_law_1.01,0.056953597068786624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,16,power_law_1.01,0.056595200300216676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,128,power_law_1.2,0.09378560185432434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,16,power_law_1.2,0.16562559604644775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,16,power_law_1.01,0.059443199634552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,16,power_law_1.01,0.05983359813690185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,128,power_law_1.2,0.10217599868774414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,16,power_law_1.01,0.061862397193908694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,8,power_law_1.2,4.708902359008789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,16,power_law_1.2,0.21569280624389647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,16,power_law_1.01,0.061459201574325564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,16,power_law_1.01,0.07228800058364868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,128,power_law_1.2,0.1319167971611023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,16,power_law_1.01,0.07390080094337463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,16,power_law_1.2,0.29088640213012695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,16,power_law_1.01,0.07813119888305664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,16,power_law_1.01,0.09145600199699402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,128,power_law_1.2,0.14913920164108277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,16,power_law_1.01,0.10241919755935669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,16,power_law_1.2,0.38161919116973875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,16,power_law_1.01,0.12480000257492066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,128,power_law_1.2,0.2023616075515747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,16,power_law_1.01,0.12957439422607422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,16,power_law_1.01,0.1728767991065979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,16,power_law_1.2,0.516102409362793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,16,power_law_1.01,0.21074559688568115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,128,power_law_1.2,0.21587200164794923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,16,power_law_1.01,0.2787519931793213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,16,power_law_1.2,0.7216127872467041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,16,power_law_1.01,0.36189439296722414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,128,power_law_1.2,0.29044480323791505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,16,power_law_1.01,0.4663360118865967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,16,power_law_1.01,0.5904511928558349
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,16,power_law_1.2,0.9317119598388672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,256,balanced,0.06105599800745646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,128,power_law_1.2,0.3662911891937256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,256,balanced,0.06217066446940104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,256,balanced,0.06429333488146464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,256,balanced,0.066021333138148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,256,balanced,0.06614399949709575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,256,balanced,0.06825066606203715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,256,balanced,0.06854933500289917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,256,balanced,0.06672533353169759
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,256,balanced,0.06830933193365733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,256,balanced,0.06628799935181935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,256,balanced,0.06841599941253662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,256,balanced,0.07074666519959767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,256,balanced,0.07070399820804596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,16,power_law_1.01,0.7547584056854248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,256,balanced,0.07096000015735626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,256,balanced,0.07462400197982788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,256,balanced,0.07842133442560832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,256,balanced,0.08272533118724823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,256,balanced,0.08989866574605306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,256,balanced,0.09890666604042053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,256,balanced,0.11633066336313884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,256,balanced,0.13380266229311624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,256,balanced,0.17695999145507812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,256,balanced,0.17908799648284912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,256,balanced,0.23361599445343018
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,256,balanced,0.27382399638493854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,256,balanced,0.3702186743418376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,256,balanced,0.4559839963912964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,16,power_law_1.2,1.1407744407653808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,1,256,balanced,0.5368693272272745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,1,256,balanced,0.8109172979990641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,16,power_law_1.01,1.166489601135254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,128,power_law_1.2,0.5157504081726074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,1,256,balanced,1.7256746292114258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,16,power_law_1.2,1.8946111679077149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,128,power_law_1.2,0.6324672222137451
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,16,power_law_1.01,2.6327104568481445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,2,128,power_law_1.2,0.806982421875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,16,power_law_1.2,3.8415550231933593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,2,128,power_law_1.2,1.3745984077453612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,32,8,power_law_1.2,0.049753600358963014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,32,8,power_law_1.2,0.043724799156188966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,32,8,power_law_1.2,0.04428800046443939
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,32,8,power_law_1.2,0.046803200244903566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,32,8,power_law_1.2,0.04715520143508911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,32,8,power_law_1.2,0.05177599787712097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,8,power_law_1.2,0.1477120041847229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,32,8,power_law_1.2,0.05308799743652344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,32,8,power_law_1.2,0.05416960120201111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,32,8,power_law_1.2,0.05489919781684875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,8,power_law_1.2,0.25311999320983886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,32,8,power_law_1.2,0.05569919943809509
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,2,128,power_law_1.2,2.8089920043945313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,32,8,power_law_1.2,0.05576320290565491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,2,power_law_1.2,0.05312640070915222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,8,power_law_1.2,0.18390400409698487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,32,8,power_law_1.2,0.05854719877243042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,32,8,power_law_1.2,0.06272640228271484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,2,power_law_1.2,0.04695680141448975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,8,power_law_1.2,0.14807039499282837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,32,8,power_law_1.2,0.06522240042686463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,2,power_law_1.2,0.06420480012893677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,32,8,power_law_1.2,0.06887680292129517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,8,power_law_1.2,0.1997375965118408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,32,8,power_law_1.2,0.07327359914779663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,2,power_law_1.2,0.0850048005580902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,32,8,power_law_1.2,0.08421760201454162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,8,power_law_1.2,0.2891968011856079
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,2,power_law_1.2,0.10053759813308716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,32,8,power_law_1.2,0.09470720291137695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,32,8,power_law_1.2,0.10316799879074097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,2,power_law_1.2,0.1674623966217041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,8,power_law_1.2,0.30409600734710696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,32,8,power_law_1.2,0.13470720052719115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,2,power_law_1.2,0.18093440532684327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,32,8,power_law_1.2,0.15252480506896973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,8,power_law_1.2,0.39946880340576174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,2,power_law_1.2,0.1834879994392395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,32,8,power_law_1.2,0.19601919651031494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,2,power_law_1.2,0.19165439605712892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,8,power_law_1.2,0.44832639694213866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,32,8,power_law_1.2,0.23894400596618653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,2,power_law_1.2,0.1937407970428467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,32,8,power_law_1.2,0.29832959175109863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,8,power_law_1.2,0.4336832046508789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,2,power_law_1.2,0.19514240026474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,32,8,power_law_1.2,0.41168642044067383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,1,2,balanced,0.08331733445326488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,1,2,balanced,0.1662773291269938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,1,2,balanced,0.2133493423461914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,2,power_law_1.2,0.20992639064788818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,1,2,balanced,0.21450134118398032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,8,power_law_1.2,0.44350719451904297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,1,2,balanced,0.21246933937072754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,1,2,balanced,0.21357333660125732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,1,2,balanced,0.21606934070587158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,32,8,power_law_1.2,0.6741824150085449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,1,2,balanced,0.2178986668586731
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,1,2,balanced,0.21780800819396973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,1,2,balanced,0.21733333667119345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,1,2,balanced,0.22639999787012735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,1,2,balanced,0.2211893399556478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,1,2,balanced,0.2235520084698995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,1,2,balanced,0.23309866587320963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,1,2,balanced,0.2630773385365804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,2,power_law_1.2,0.20364160537719728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,1,2,balanced,0.26447999477386475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,1,2,balanced,0.26136000951131183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,1,2,balanced,0.3354826768239339
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,1,2,balanced,0.33104532957077026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,1,2,balanced,0.4087093273798625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,1,2,balanced,0.4773333470026652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,8,power_law_1.2,0.4770688056945801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,32,8,power_law_1.2,0.7783487796783447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,1,2,balanced,0.6367146571477255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,1,2,balanced,0.7946080366770426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,2,power_law_1.2,0.21905279159545898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,1,2,balanced,1.1270986398061116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,1,2,balanced,1.502629280090332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,2,power_law_1.2,0.2266495943069458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,1,2,balanced,2.199728012084961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,8,power_law_1.2,0.5010816097259522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,32,8,power_law_1.2,0.9088959693908691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,1,2,balanced,2.879455884297689
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,2,power_law_1.2,0.2298367977142334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,1,2,balanced,3.585269292195638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,8,power_law_1.2,0.5049215793609619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,2,power_law_1.2,0.2342207908630371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,32,8,power_law_1.2,1.480185604095459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,1,2,balanced,5.712805430094401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,2,power_law_1.2,0.2655807971954346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,8,power_law_1.2,0.5357312202453614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,2,power_law_1.2,0.2869247913360596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,8,power_law_1.2,0.5616384029388428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,2,power_law_1.2,0.35898239612579347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,1,2,balanced,11.523087819417318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,2,power_law_1.2,0.34019200801849364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,32,8,power_law_1.2,3.031417655944824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,8,power_law_1.2,0.5631167888641357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,2,power_law_1.2,0.42426238059997556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,8,power_law_1.2,0.6706048011779785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,2,power_law_1.2,0.5001152038574219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,2,power_law_1.2,0.6917888164520264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,8,power_law_1.2,0.6894656181335449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,2,power_law_1.2,0.82675199508667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,8,power_law_1.2,0.8129280090332032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,2,power_law_1.2,1.076416015625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,8,power_law_1.2,0.8561920166015625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,2,power_law_1.2,1.5303104400634766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,8,power_law_1.2,1.0105152130126953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,2,power_law_1.2,1.8270336151123048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,8,power_law_1.2,1.1035072326660156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,2,power_law_1.2,2.789606475830078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,8,power_law_1.2,1.358835220336914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,4,power_law_1.01,0.033855998516082765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,4,power_law_1.01,0.030707201361656188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,32,8,power_law_1.2,0.05141760110855102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,8,power_law_1.2,1.4576512336730958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,4,power_law_1.01,0.03047040104866028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,4,power_law_1.01,0.032742398977279666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,32,8,power_law_1.2,0.0500927984714508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,2,power_law_1.2,5.96814079284668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,4,power_law_1.01,0.03472639918327332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,8,power_law_1.2,1.873151969909668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,4,power_law_1.01,0.03672960102558136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,32,8,power_law_1.2,0.04568319916725159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,4,power_law_1.01,0.03842560052871704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,32,8,power_law_1.2,0.046374401450157164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,4,power_law_1.01,0.03889279961585999
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,4,power_law_1.01,0.03924480080604553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,32,8,power_law_1.2,0.048198398947715757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,8,power_law_1.2,2.2486976623535155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,4,power_law_1.01,0.040608000755310056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,4,power_law_1.01,0.04316799938678741
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,32,8,power_law_1.2,0.05260800123214722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,4,power_law_1.01,0.04397439956665039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,4,power_law_1.01,0.047276800870895384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,32,8,power_law_1.2,0.0596671998500824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,4,power_law_1.01,0.04757120013237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,8,power_law_1.2,2.846892738342285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,4,power_law_1.01,0.054553598165512085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,32,8,power_law_1.2,0.06027519702911377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,4,power_law_1.01,0.05644800066947937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,32,8,power_law_1.2,0.06282240152359009
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,4,power_law_1.01,0.06170880198478699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,4,power_law_1.01,0.06919040083885193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,32,8,power_law_1.2,0.06307839751243591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,4,power_law_1.01,0.07620480060577392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,8,power_law_1.2,3.698438262939453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,4,power_law_1.01,0.09536640048027038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,32,8,power_law_1.2,0.0640064001083374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,4,power_law_1.01,0.10936319828033447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,32,8,power_law_1.2,0.06647679805755616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,4,power_law_1.01,0.13519359827041627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,32,8,power_law_1.2,0.06666240096092224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,4,power_law_1.01,0.18135679960250856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,4,power_law_1.01,0.23914239406585694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,8,1,balanced,0.03900266687075297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,32,8,power_law_1.2,0.070278400182724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,8,1,balanced,0.06571199993292491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,8,1,balanced,0.07945600152015686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,8,1,balanced,0.08153066535790761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,8,1,balanced,0.08211199939250946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,8,1,balanced,0.08164266745249431
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,8,1,balanced,0.0813973347345988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,4,power_law_1.01,0.29626879692077634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,8,1,balanced,0.08293866614500682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,8,1,balanced,0.08286933104197185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,32,8,power_law_1.2,0.08127359747886657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,8,1,balanced,0.08346666892369588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,8,1,balanced,0.08505066235860188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,8,1,balanced,0.08661333719889323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,8,1,balanced,0.08828266461690266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,8,1,balanced,0.0941493312517802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,8,1,balanced,0.10570133725802104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,8,1,balanced,0.10591466228167216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,8,1,balanced,0.10744000474611919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,4,power_law_1.01,0.48793601989746094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,8,1,balanced,0.13134400049845377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,8,power_law_1.2,7.567903900146485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,8,1,balanced,0.14038399855295816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,32,8,power_law_1.2,0.08516479730606079
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,8,1,balanced,0.18262932697931925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,8,1,balanced,0.21755200624465942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,8,1,balanced,0.28703999519348145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,8,1,balanced,0.34568532307942706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,8,1,balanced,0.48925864696502686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,8,1,balanced,0.6184320052464803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,4,power_law_1.01,0.5629504203796387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,32,8,power_law_1.2,0.08925439715385437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,8,1,balanced,0.893776019414266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,8,1,balanced,1.1616746584574382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,8,1,balanced,1.4375680287679036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,32,8,power_law_1.2,0.10306559801101685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,8,1,balanced,2.2455466588338218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,4,power_law_1.01,0.7323647975921631
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,8,1,balanced,4.424746513366699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,32,8,power_law_1.2,0.10826239585876465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,4,power_law_1.01,1.052627182006836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,32,8,power_law_1.2,0.13803520202636718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,32,8,power_law_1.2,0.16980479955673217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,32,8,power_law_1.2,0.21016960144042968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,4,power_law_1.01,2.2200895309448243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,32,8,power_law_1.2,0.2506943941116333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,32,8,power_law_1.2,0.34825599193573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,32,8,power_law_1.2,0.42844161987304685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,32,8,power_law_1.2,0.6360383987426758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,32,4,balanced,0.04400533437728882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,32,4,balanced,0.045978665351867676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,32,4,balanced,0.0499839981396993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,32,4,balanced,0.0498986691236496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,32,4,balanced,0.05929600199063619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,32,4,balanced,0.07814933359622955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,32,4,balanced,0.10100799798965454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,32,4,balanced,0.09618133306503296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,32,4,balanced,0.09698133667310078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,32,4,balanced,0.09696533282597859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,32,4,balanced,0.09929066896438599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,32,4,balanced,0.10132799545923869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,32,4,balanced,0.10453333457310994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,32,4,balanced,0.1060693363348643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,32,4,balanced,0.1151039997736613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,32,4,balanced,0.11778133114178975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,32,4,balanced,0.12635733683904013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,32,4,balanced,0.1463520030180613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,32,4,balanced,0.16170666615168253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,32,8,power_law_1.2,0.8832832336425781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,32,4,balanced,0.19169066349665323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,32,4,balanced,0.22185067335764566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,32,4,balanced,0.28153600295384723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,32,4,balanced,0.35918935139973956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,32,4,balanced,0.49460268020629883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,32,4,balanced,0.6619146664937338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,32,4,balanced,0.9403146902720133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,32,4,balanced,1.2131146589914958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,32,4,balanced,1.5119519233703613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,32,8,power_law_1.2,0.9939071655273437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,32,4,balanced,2.307664076487223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,8,power_law_1.01,0.03728640079498291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,32,4,balanced,4.529930750528972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,32,8,power_law_1.2,1.4536191940307617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,8,power_law_1.01,0.030880001187324525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,8,power_law_1.01,0.03084160089492798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,4,8,balanced,0.06866666674613953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,4,8,balanced,0.07195200026035309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,4,8,balanced,0.05481066803137461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,4,8,balanced,0.0563679983218511
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,4,8,balanced,0.05583466589450836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,4,8,balanced,0.05606933434804281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,4,8,balanced,0.05776533484458923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,4,8,balanced,0.055733333031336464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,4,8,balanced,0.055589333176612854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,4,8,balanced,0.05705066521962484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,4,8,balanced,0.05795200169086456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,4,8,balanced,0.058778668443361916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,4,8,balanced,0.05839466551939646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,4,8,balanced,0.06229333579540253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,4,8,balanced,0.06629866858323415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,4,8,balanced,0.06708799799283345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,8,power_law_1.01,0.03116160035133362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,4,8,balanced,0.06990399956703186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,4,8,balanced,0.07744533320267995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,4,8,balanced,0.07710400223731995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,4,8,balanced,0.09296533465385437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,4,8,balanced,0.10283199946085612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,4,8,balanced,0.12191466490427653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,4,8,balanced,0.14711466431617737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,4,8,balanced,0.18557866414388022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,4,8,balanced,0.22852800289789835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,4,8,balanced,0.3190880020459493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,4,8,balanced,0.390122652053833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,8,power_law_1.01,0.03294720053672791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,4,8,balanced,0.4892106850941976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,4,8,balanced,0.7511839866638184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,32,8,power_law_1.2,3.3189247131347654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,4,8,balanced,1.4004054069519043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,8,power_law_1.01,0.03588480055332184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,8,power_law_1.01,0.03532800078392029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,8,power_law_1.01,0.034764799475669864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,8,power_law_1.01,0.03600000143051148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,8,power_law_1.01,0.0351936012506485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,8,power_law_1.01,0.03704319894313812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,8,power_law_1.01,0.03733760118484497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,8,power_law_1.01,0.038047999143600464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,8,power_law_1.01,0.0414463996887207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,8,power_law_1.01,0.04434559941291809
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,8,power_law_1.01,0.04661760032176972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,32,1,power_law_1.01,0.03530240058898926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,8,power_law_1.01,0.04836480021476745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,32,1,power_law_1.01,0.037196800112724304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,32,1,power_law_1.01,0.03672960102558136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,32,1,power_law_1.01,0.05057920217514038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,8,power_law_1.01,0.05569919943809509
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,32,1,power_law_1.01,0.05317760109901428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,32,1,power_law_1.01,0.05289599895477295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,32,1,power_law_1.01,0.052767997980117796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,32,1,power_law_1.01,0.055052798986434934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,8,power_law_1.01,0.06680960059165955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,32,1,power_law_1.01,0.05600000023841858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,32,1,power_law_1.01,0.05899519920349121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,32,1,power_law_1.01,0.05653759837150574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,32,1,power_law_1.01,0.05909119844436646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,8,power_law_1.01,0.07774080038070678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,32,1,power_law_1.01,0.060575997829437254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,32,1,power_law_1.01,0.0671999990940094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,32,1,power_law_1.01,0.07625600099563598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,32,1,power_law_1.01,0.077811199426651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,8,power_law_1.01,0.08065919876098633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,32,1,power_law_1.01,0.0790336012840271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,32,1,power_law_1.01,0.08902400135993957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,32,1,power_law_1.01,0.10572160482406616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,32,1,power_law_1.01,0.1309183955192566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,8,power_law_1.01,0.1089087963104248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,32,1,power_law_1.01,0.1384063959121704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,32,1,power_law_1.01,0.18330880403518676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,32,1,power_law_1.01,0.21811840534210206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,8,power_law_1.01,0.13403520584106446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,32,1,power_law_1.01,0.29928319454193114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,32,1,power_law_1.01,0.38005120754241944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,4,power_law_1.01,0.07502719759941101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,32,1,power_law_1.01,0.543667221069336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,8,power_law_1.01,0.18662400245666505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,4,power_law_1.01,0.08082559704780579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,4,power_law_1.2,0.04795520007610321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,32,1,power_law_1.01,0.6868351936340332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,4,power_law_1.01,0.06662399768829345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,32,1,power_law_1.01,0.8442560195922851
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,8,power_law_1.01,0.2179327964782715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,4,power_law_1.01,0.10084480047225952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,32,1,power_law_1.01,1.3084287643432617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,4,power_law_1.2,0.05320320129394531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,4,power_law_1.01,0.12401920557022095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,8,power_law_1.01,0.33761279582977294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,4,power_law_1.01,0.17293440103530883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,32,1,power_law_1.01,2.5668224334716796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,4,power_law_1.2,0.061273598670959474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,4,power_law_1.01,0.18992639780044557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,8,power_law_1.01,0.48042879104614256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,4,power_law_1.01,0.1905727982521057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,4,power_law_1.01,0.18880000114440917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,4,power_law_1.2,0.05660799741744995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,8,power_law_1.01,0.5183807849884033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,4,power_law_1.01,0.19886720180511475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,4,power_law_1.01,0.20016000270843506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,4,power_law_1.2,0.06865280270576476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,4,power_law_1.01,0.1998080015182495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,8,power_law_1.01,0.7278079986572266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,4,power_law_1.01,0.20183680057525635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,4,power_law_1.2,0.08391039967536926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,4,power_law_1.01,0.20968959331512452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,4,power_law_1.01,0.22014079093933106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,4,power_law_1.2,0.09068160057067871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,4,power_law_1.01,0.22524800300598144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,4,power_law_1.01,0.2273024082183838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,8,power_law_1.01,1.457759952545166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,4,power_law_1.01,0.2428800106048584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,4,power_law_1.2,0.10476160049438477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,4,power_law_1.01,0.2685247898101807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,4,power_law_1.01,0.3026304006576538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,4,power_law_1.2,0.10996479988098144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,4,power_law_1.01,0.3304960012435913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,4,power_law_1.2,0.1125823974609375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,4,power_law_1.01,0.37538559436798097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,4,power_law_1.01,0.43610239028930664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,4,power_law_1.2,0.12168960571289063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,4,power_law_1.01,0.5879744052886963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,2,balanced,0.03345600018898646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,2,balanced,0.03341866781314214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,2,balanced,0.03754133234421412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,2,balanced,0.056608001391092934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,2,balanced,0.0823413332303365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,2,balanced,0.08574400345484416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,2,balanced,0.08463999629020691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,2,balanced,0.08540800213813782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,2,balanced,0.08589866757392883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,2,balanced,0.08776000142097473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,2,balanced,0.08885866403579712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,2,balanced,0.0913813312848409
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,2,balanced,0.09118400017420451
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,2,balanced,0.0958720048268636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,2,balanced,0.10098666946093242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,4,power_law_1.01,0.6884160041809082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,2,balanced,0.10190932949384053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,4,power_law_1.2,0.12573440074920655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,2,balanced,0.10934399565060933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,2,balanced,0.12877866625785828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,2,balanced,0.1288426617781321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,2,balanced,0.1597599983215332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,2,balanced,0.1678666671117147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,2,balanced,0.23111466566721597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,2,balanced,0.26689600944519043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,2,balanced,0.37966398398081463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,2,balanced,0.4793599843978882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,2,balanced,0.6812053521474203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,4,power_law_1.01,1.02357759475708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,2,balanced,0.8596746921539307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,4,power_law_1.2,0.12268160581588745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,2,balanced,1.0680267016092937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,2,balanced,1.6193973223368328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,2,balanced,3.123648007710775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,4,power_law_1.01,1.2772543907165528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,4,power_law_1.2,0.13916800022125245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,4,power_law_1.01,1.4030400276184083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,4,power_law_1.2,0.14269440174102782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,4,power_law_1.2,0.15898879766464233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,4,power_law_1.01,2.394099235534668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,4,1,power_law_1.2,0.045977601408958436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,4,1,power_law_1.2,0.05287039875984192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,4,power_law_1.2,0.16201599836349487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,4,1,power_law_1.2,0.05902720093727112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,4,1,power_law_1.2,0.08102399706840516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,4,1,power_law_1.2,0.08113279938697815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,4,power_law_1.2,0.18143999576568604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,4,1,power_law_1.2,0.08458880186080933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,4,1,power_law_1.2,0.08633599877357483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,4,1,power_law_1.2,0.08629760146141052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,4,1,power_law_1.2,0.08589439988136291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,4,power_law_1.2,0.19688960313796997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,4,power_law_1.01,4.203564834594727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,4,1,power_law_1.2,0.09055359959602356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,4,1,power_law_1.2,0.08856959939002991
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,4,1,power_law_1.2,0.09493119716644287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,4,1,power_law_1.2,0.09591680169105529
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,4,power_law_1.2,0.23651840686798095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,4,1,power_law_1.2,0.09928960204124451
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,4,1,power_law_1.2,0.11007360219955445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,4,1,power_law_1.2,0.11482239961624145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,4,1,power_law_1.2,0.11853439807891845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,4,1,power_law_1.2,0.1383936047554016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,4,power_law_1.2,0.2676991939544678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,4,1,power_law_1.2,0.15058560371398927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,4,1,power_law_1.2,0.1908031940460205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,4,1,power_law_1.2,0.22120959758758546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,4,1,power_law_1.2,0.2939071893692017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,4,power_law_1.2,0.35392639636993406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,4,1,power_law_1.2,0.3554431915283203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,64,power_law_1.01,0.03383040130138397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,4,1,power_law_1.2,0.4872255802154541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,64,power_law_1.01,0.03330560028553009
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,4,1,power_law_1.2,0.6215424060821533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,64,power_law_1.01,0.029440000653266907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,4,power_law_1.2,0.4347519874572754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,4,1,power_law_1.2,0.882579231262207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,64,power_law_1.01,0.03052160143852234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,4,1,power_law_1.2,1.1348928451538085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,64,power_law_1.01,0.030662399530410767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,4,1,power_law_1.2,1.393824005126953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,64,power_law_1.01,0.031353598833084105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,4,power_law_1.2,0.5940351963043213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,64,power_law_1.01,0.03163520097732544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,4,1,power_law_1.2,2.1637311935424806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,64,power_law_1.01,0.03173120021820068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,64,power_law_1.01,0.03189760148525238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,4,power_law_1.2,0.7013760089874268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,64,power_law_1.01,0.03199360072612763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,4,1,power_law_1.2,4.2128959655761715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,64,power_law_1.01,0.03272320032119751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,64,power_law_1.01,0.033504000306129454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,4,power_law_1.2,0.9913472175598145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,64,power_law_1.01,0.03493120074272156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,64,power_law_1.01,0.036313599348068236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,64,power_law_1.01,0.04131839871406555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,64,power_law_1.01,0.042745599150657655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,4,power_law_1.2,1.2560256004333497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,64,power_law_1.01,0.041203200817108154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,64,power_law_1.01,0.04839679896831513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,64,power_law_1.01,0.051571202278137204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,64,power_law_1.01,0.067084801197052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,4,power_law_1.2,1.6230976104736328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,4,power_law_1.01,0.09675520062446594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,64,power_law_1.01,0.06792320013046264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,64,power_law_1.01,0.08387200236320495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,4,power_law_1.01,0.1445312023162842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,64,power_law_1.01,0.11201920509338378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,4,power_law_1.01,0.10080000162124633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,64,power_law_1.01,0.13203200101852416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,4,power_law_1.2,2.3228288650512696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,4,power_law_1.01,0.1275712013244629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,64,power_law_1.01,0.1417407989501953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,64,power_law_1.01,0.20814080238342286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,4,power_law_1.01,0.17800960540771485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,4,power_law_1.01,0.24023680686950682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,64,power_law_1.01,0.27915520668029786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,4,power_law_1.01,0.31617279052734376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,64,power_law_1.01,0.3896320104598999
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,4,power_law_1.2,4.70211181640625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,8,power_law_1.2,0.0377344012260437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,4,power_law_1.01,0.33145599365234374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,64,power_law_1.01,0.5254208087921143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,4,power_law_1.01,0.3503999948501587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,8,power_law_1.2,0.030803200602531434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,4,power_law_1.01,0.34340479373931887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,8,power_law_1.2,0.03086079955101013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,4,power_law_1.01,0.36475520133972167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,8,power_law_1.2,0.03212159872055054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,64,power_law_1.01,1.0728447914123536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,4,power_law_1.01,0.37138559818267824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,8,power_law_1.2,0.033395200967788696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,4,power_law_1.01,0.3684351921081543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,8,power_law_1.2,0.036473599076271054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,4,power_law_1.01,0.38800640106201173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,8,power_law_1.2,0.036748799681663516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,4,power_law_1.01,0.4126143932342529
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,8,power_law_1.2,0.03811199963092804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,4,power_law_1.01,0.4190783977508545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,8,power_law_1.2,0.03834879994392395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,8,power_law_1.2,0.04005120098590851
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,4,power_law_1.01,0.42771201133728026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,8,power_law_1.2,0.04161919951438904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,4,power_law_1.01,0.4725503921508789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,8,power_law_1.2,0.042342400550842284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,4,power_law_1.01,0.478931188583374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,8,power_law_1.2,0.0440447986125946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,4,power_law_1.01,0.5422016143798828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,8,power_law_1.2,0.04782080054283142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,16,power_law_1.2,0.05989760160446167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,4,power_law_1.01,0.5614336013793946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,8,power_law_1.2,0.053414398431777955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,8,power_law_1.2,0.05728639960289002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,4,power_law_1.01,0.648806381225586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,16,power_law_1.2,0.06261119842529297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,8,power_law_1.2,0.05851519703865051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,16,2,power_law_1.01,0.0383103996515274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,16,2,power_law_1.01,0.03751679956912994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,4,power_law_1.01,0.7079999923706055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,16,2,power_law_1.01,0.03823359906673431
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,8,power_law_1.2,0.061510401964187625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,16,2,power_law_1.01,0.05120000243186951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,16,2,power_law_1.01,0.051910400390625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,16,power_law_1.2,0.05916799902915955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,16,2,power_law_1.01,0.05221760272979736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,16,2,power_law_1.01,0.053337597846984865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,16,2,power_law_1.01,0.05756160020828247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,8,power_law_1.2,0.07958400249481201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,16,2,power_law_1.01,0.05628160238265991
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,4,power_law_1.01,0.9094207763671875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,16,2,power_law_1.01,0.056831997632980344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,16,2,power_law_1.01,0.05894399881362915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,16,2,power_law_1.01,0.059596800804138185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,16,power_law_1.2,0.055155199766159055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,8,power_law_1.2,0.09221119880676269
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,16,2,power_law_1.01,0.06218240261077881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,16,2,power_law_1.01,0.06638720035552978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,16,2,power_law_1.01,0.07253119945526124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,16,2,power_law_1.01,0.07694720029830933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,4,power_law_1.01,1.1054335594177247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,16,2,power_law_1.01,0.07602559924125671
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,8,power_law_1.2,0.10310399532318115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,16,2,power_law_1.01,0.09785599708557129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,16,2,power_law_1.01,0.09934719800949096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,16,power_law_1.2,0.05556480288505554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,16,2,power_law_1.01,0.12405760288238525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,16,2,power_law_1.01,0.14547840356826783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,8,power_law_1.2,0.12907520532608033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,16,2,power_law_1.01,0.1959231972694397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,16,2,power_law_1.01,0.2264319896697998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,4,power_law_1.01,1.4111871719360352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,16,2,power_law_1.01,0.3114239931106567
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,16,power_law_1.2,0.0597760021686554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,16,2,power_law_1.01,0.4281599998474121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,8,power_law_1.2,0.15733760595321655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,16,2,power_law_1.01,0.5210048198699951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,16,2,power_law_1.01,0.7091904163360596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,8,power_law_1.2,0.21381120681762694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,16,power_law_1.2,0.06113280057907104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,16,2,power_law_1.01,0.7218431949615478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,4,power_law_1.01,1.8577791213989259
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,16,2,power_law_1.01,1.426809597015381
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,8,power_law_1.2,0.28954880237579345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,16,power_law_1.2,0.06320000290870667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,16,2,power_law_1.01,2.4157888412475588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,4,power_law_1.01,2.1967296600341797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,8,power_law_1.2,0.4814335823059082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,16,power_law_1.2,0.06454399824142457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,8,power_law_1.2,0.5630527973175049
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,16,power_law_1.2,0.06465280055999756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,4,power_law_1.01,3.3027137756347655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,8,power_law_1.2,0.7436543941497803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,16,power_law_1.2,0.06620799899101257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,16,power_law_1.2,0.06880639791488648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,8,power_law_1.2,1.1004927635192872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,16,power_law_1.2,0.06958720088005066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,4,power_law_1.01,6.362438583374024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,16,power_law_1.2,0.07085440158843995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,8,power_law_1.2,2.677510452270508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,16,power_law_1.2,0.07645440101623535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,16,power_law_1.2,0.07838720083236694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,16,power_law_1.2,0.08888959884643555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,1,power_law_1.01,0.09886720180511474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,16,power_law_1.2,0.10107519626617431
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,1,power_law_1.01,0.1765439987182617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,1,power_law_1.01,0.1947711944580078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,16,power_law_1.2,0.11304960250854493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,1,power_law_1.01,0.33267199993133545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,1,power_law_1.01,0.4767615795135498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,16,power_law_1.2,0.1407807946205139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,1,power_law_1.01,0.837337589263916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,16,power_law_1.2,0.16425600051879882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,1,power_law_1.01,0.8913536071777344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,1,power_law_1.01,0.9081727981567382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,8,balanced,0.051967998345692955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,16,power_law_1.2,0.211411190032959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,8,balanced,0.05415999889373779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,8,balanced,0.054469332098960876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,8,balanced,0.05648533503214518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,8,balanced,0.06779199838638306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,8,balanced,0.0946560005346934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,1,power_law_1.01,0.963321590423584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,8,balanced,0.09446932872136433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,8,balanced,0.09757333000500996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,8,balanced,0.09745066364606221
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,8,balanced,0.0979253351688385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,8,balanced,0.0993173321088155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,8,balanced,0.1013866662979126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,8,balanced,0.09966933727264404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,8,balanced,0.10417067011197408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,8,balanced,0.10990400115648906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,8,balanced,0.10921600461006165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,8,balanced,0.11571733156840007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,8,balanced,0.1234773298104604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,1,power_law_1.01,0.9559103965759277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,8,balanced,0.13576533397038779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,8,balanced,0.15524799625078836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,8,balanced,0.17697066068649292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,16,power_law_1.2,0.2565696001052856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,8,balanced,0.20943466822306314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,8,balanced,0.2374080022176107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,8,balanced,0.3200906713803609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,8,balanced,0.38950932025909424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,8,balanced,0.5555359919865926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,1,power_law_1.01,0.9847231864929199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,8,balanced,0.7041707038879395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,64,power_law_1.2,0.07085440158843995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,8,balanced,0.8628213405609131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,8,balanced,1.3190879821777344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,1,power_law_1.01,1.0369152069091796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,16,power_law_1.2,0.3378432035446167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,8,balanced,2.5634826024373374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,64,power_law_1.2,0.07009279727935791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,1,power_law_1.01,1.040595245361328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,1,power_law_1.01,1.0853952407836913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,16,power_law_1.2,0.40663681030273435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,64,power_law_1.2,0.06175360083580017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,1,power_law_1.01,1.1137727737426757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,1,power_law_1.01,1.1345151901245116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,64,power_law_1.2,0.06101120114326477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,16,power_law_1.2,0.6168255805969238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,1,power_law_1.01,1.153555202484131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,1,power_law_1.01,1.258732795715332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,16,power_law_1.2,0.8337216377258301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,64,power_law_1.2,0.0632960021495819
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,1,power_law_1.01,1.3337984085083008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,1,power_law_1.01,1.4904767990112304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,16,power_law_1.2,0.9834367752075195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,64,power_law_1.2,0.06504319906234741
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,1,power_law_1.01,1.697817611694336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,1,power_law_1.01,2.0931264877319338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,64,power_law_1.2,0.06668800115585327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,16,power_law_1.2,1.5733759880065918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,1,power_law_1.01,2.176255989074707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,64,power_law_1.2,0.06741120219230652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,1,power_law_1.01,2.726873588562012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,1,power_law_1.01,0.10069760084152221
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,64,power_law_1.2,0.06808320283889771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,1,power_law_1.01,3.5691009521484376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,16,power_law_1.2,3.363577651977539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,1,power_law_1.01,0.13278080224990846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,64,power_law_1.2,0.06659200191497802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,1,power_law_1.01,0.1525056004524231
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,1,power_law_1.01,4.598175811767578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,64,power_law_1.2,0.06970880031585694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,1,power_law_1.01,0.24416000843048097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,1,power_law_1.01,0.37947518825531007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,1,power_law_1.01,5.776160049438476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,64,power_law_1.2,0.06992639899253845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,1,power_law_1.01,0.49797759056091306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,64,power_law_1.2,0.06802560091018676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,1,power_law_1.01,0.7110079765319824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,1,power_law_1.01,6.784114837646484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,64,power_law_1.2,0.07048959732055664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,1,power_law_1.01,0.7529920101165771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,64,power_law_1.2,0.074099200963974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,1,power_law_1.01,0.7720128059387207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,1,power_law_1.01,0.8268287658691407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,64,power_law_1.2,0.07591040134429931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,1,power_law_1.01,9.657273864746093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,1,power_law_1.01,0.830726432800293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,64,power_law_1.2,0.07838079929351807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,1,power_law_1.01,0.8840191841125489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,16,power_law_1.2,0.03463039994239807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,64,power_law_1.2,0.0877120018005371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,16,power_law_1.2,0.03136639893054962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,1,power_law_1.01,0.8828288078308105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,16,power_law_1.2,0.0304639995098114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,16,power_law_1.2,0.03157120048999786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,64,power_law_1.2,0.10072959661483764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,1,power_law_1.01,0.9128512382507324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,16,power_law_1.2,0.03136000037193298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,16,power_law_1.2,0.033542400598526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,1,power_law_1.01,0.9381759643554688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,64,power_law_1.2,0.12327680587768555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,16,power_law_1.2,0.03372800052165985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,16,power_law_1.2,0.03472639918327332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,1,power_law_1.01,0.9671551704406738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,16,power_law_1.2,0.03498879969120026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,64,power_law_1.2,0.14393600225448608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,1,power_law_1.01,18.164767456054687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,16,power_law_1.2,0.035155200958251955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,1,power_law_1.01,0.9767040252685547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,16,power_law_1.2,0.03620480000972748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,64,power_law_1.2,0.18216960430145263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,16,power_law_1.2,0.03895680010318756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,1,power_law_1.01,1.0830143928527831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,16,power_law_1.2,0.03848319947719574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,16,power_law_1.2,0.03932160139083862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,1,power_law_1.01,1.1292608261108399
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,64,power_law_1.2,0.21841280460357665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,16,power_law_1.2,0.04467200040817261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,16,power_law_1.2,0.045926401019096376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,1,power_law_1.01,1.3003968238830566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,16,power_law_1.2,0.048390400409698484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,64,power_law_1.2,0.31004159450531005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,16,power_law_1.2,0.05518720149993896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,1,power_law_1.01,1.3561792373657227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,16,power_law_1.2,0.06092159748077393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,64,power_law_1.2,0.40584959983825686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,16,power_law_1.2,0.07036160230636597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,1,power_law_1.01,1.5878080368041991
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,16,power_law_1.2,0.09397760033607483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,16,power_law_1.2,0.1354688048362732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,64,power_law_1.2,0.5718272209167481
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,16,power_law_1.2,0.1577855944633484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,1,power_law_1.01,1.8811775207519532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,16,power_law_1.2,0.23498239517211914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,64,power_law_1.2,0.8522687911987304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,16,power_law_1.2,0.3307584047317505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,1,power_law_1.01,2.4237056732177735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,16,power_law_1.2,0.39638400077819824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,16,8,power_law_1.01,0.038201600313186646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,1,power_law_1.01,2.9247936248779296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,64,power_law_1.2,0.9752896308898926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,16,power_law_1.2,0.6244160175323487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,16,8,power_law_1.01,0.03777279853820801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,16,8,power_law_1.01,0.031353598833084105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,16,8,power_law_1.01,0.030399999022483824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,16,8,power_law_1.01,0.031033599376678468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,16,8,power_law_1.01,0.03110400140285492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,16,8,power_law_1.01,0.03224320113658905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,16,power_law_1.2,0.6475456237792969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,16,8,power_law_1.01,0.03322240114212036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,16,8,power_law_1.01,0.03362559974193573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,16,8,power_law_1.01,0.0343423992395401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,16,8,power_law_1.01,0.034534400701522826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,1,power_law_1.01,3.976563262939453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,16,8,power_law_1.01,0.03696640133857727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,16,8,power_law_1.01,0.037350401282310486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,16,8,power_law_1.01,0.0392192006111145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,64,power_law_1.2,1.6359743118286132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,16,8,power_law_1.01,0.04619520008563995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,16,8,power_law_1.01,0.05146239995956421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,16,power_law_1.2,1.209062385559082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,16,8,power_law_1.01,0.05182719826698303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,16,8,power_law_1.01,0.0643392026424408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,16,8,power_law_1.01,0.06778240203857422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,16,8,power_law_1.01,0.08861439824104309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,16,8,power_law_1.01,0.10629119873046874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,16,8,power_law_1.01,0.13586560487747193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,1,power_law_1.01,4.938662338256836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,16,8,power_law_1.01,0.16872960329055786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,16,8,power_law_1.01,0.22354559898376464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,16,8,power_law_1.01,0.234604811668396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,16,8,power_law_1.01,0.4009407997131348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,16,8,power_law_1.01,0.48786559104919436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,16,power_law_1.2,2.8522815704345703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,16,8,power_law_1.01,0.5822463989257812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,64,power_law_1.2,3.150374412536621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,16,8,power_law_1.01,0.9282048225402832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,1,power_law_1.01,5.987052917480469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,16,8,power_law_1.01,1.7099647521972656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,1,power_law_1.01,8.912921905517578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,1,power_law_1.01,16.47254333496094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,64,power_law_1.01,0.0735871970653534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,32,2,power_law_1.2,0.06209279894828797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,32,power_law_1.01,0.09479680061340331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,32,2,power_law_1.2,0.0669376015663147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,64,power_law_1.01,0.07143679857254029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,32,2,power_law_1.2,0.06694399714469909
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,32,2,power_law_1.2,0.07432320117950439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,64,power_law_1.01,0.05815039873123169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,32,2,power_law_1.2,0.08705919981002808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,32,power_law_1.01,0.09399679899215699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,64,power_law_1.01,0.062041598558425906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,32,2,power_law_1.2,0.09904000163078308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,32,2,power_law_1.2,0.12243839502334594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,64,power_law_1.01,0.06431360244750976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,32,power_law_1.01,0.06276479959487916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,32,2,power_law_1.2,0.12857600450515747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,32,2,power_law_1.2,0.12885119915008544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,2,balanced,0.04806933303674062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,2,balanced,0.05141866703828176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,64,power_law_1.01,0.06551679968833923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,2,balanced,0.054010664423306785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,2,balanced,0.06660266717274983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,2,balanced,0.09382933378219604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,2,balanced,0.10705066720644633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,2,balanced,0.11025066177050273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,2,balanced,0.10360532999038696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,2,balanced,0.10523200035095215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,2,balanced,0.10714667042096455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,2,balanced,0.10949866970380147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,32,2,power_law_1.2,0.131660795211792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,2,balanced,0.11053867141405742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,2,balanced,0.11266666650772095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,2,balanced,0.12346667051315308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,2,balanced,0.12915199995040894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,2,balanced,0.13820800185203552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,2,balanced,0.14044800400733948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,2,balanced,0.1648373305797577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,32,power_law_1.01,0.06704000234603882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,2,balanced,0.17685866355895996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,2,balanced,0.24256000916163126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,2,balanced,0.2672800024350484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,2,balanced,0.39023999373118085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,32,2,power_law_1.2,0.13953280448913574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,64,power_law_1.01,0.06812800168991089
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,2,balanced,0.43561065196990967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,2,balanced,0.664682666460673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,2,balanced,0.801520029703776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,2,balanced,1.2050293286641438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,32,2,power_law_1.2,0.1415552020072937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,2,balanced,1.5116106669108074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,2,balanced,1.8110027313232422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,64,power_law_1.01,0.07267199754714966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,32,2,power_law_1.2,0.14483200311660765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,2,balanced,2.839989344278971
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,32,power_law_1.01,0.07938560247421264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,32,2,power_law_1.2,0.1541759967803955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,2,balanced,5.524261474609375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,64,power_law_1.01,0.07483519911766053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,32,2,power_law_1.2,0.16302720308303834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,32,power_law_1.01,0.08427519798278808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,32,2,power_law_1.2,0.16789120435714722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,64,power_law_1.01,0.07575039863586426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,32,2,power_law_1.2,0.1779647946357727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,32,2,power_law_1.2,0.19890559911727906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,64,power_law_1.01,0.07530239820480347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,32,power_law_1.01,0.08481919765472412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,32,2,power_law_1.2,0.23503360748291016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,64,power_law_1.01,0.0772607982158661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,32,2,power_law_1.2,0.28008959293365476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,32,power_law_1.01,0.08616960048675537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,32,2,power_law_1.2,0.3324287891387939
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,64,power_law_1.01,0.07865599989891052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,32,2,power_law_1.2,0.4418496131896973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,64,power_law_1.01,0.0775871992111206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,32,power_law_1.01,0.09230719804763794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,32,2,power_law_1.2,0.5278848171234131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,64,power_law_1.01,0.08568320274353028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,32,2,power_law_1.2,0.7356160163879395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,32,power_law_1.01,0.09286400079727172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,32,2,power_law_1.01,0.033292800188064575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,32,2,power_law_1.2,0.9806912422180176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,32,2,power_law_1.01,0.031046399474143983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,64,power_law_1.01,0.08595839738845826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,32,2,power_law_1.01,0.03200640082359314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,32,2,power_law_1.01,0.032076799869537355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,32,2,power_law_1.01,0.033004799485206605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,32,2,power_law_1.01,0.03446399867534637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,32,2,power_law_1.01,0.03509120047092438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,32,2,power_law_1.01,0.03500800132751465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,32,2,power_law_1.01,0.03578239977359772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,32,2,power_law_1.2,1.3865983963012696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,32,2,power_law_1.01,0.037376001477241516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,64,power_law_1.01,0.09367679953575134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,32,power_law_1.01,0.0914687991142273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,32,2,power_law_1.01,0.03699199855327606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,32,2,power_law_1.01,0.03792639970779419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,32,2,power_law_1.01,0.03963519930839539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,32,2,power_law_1.01,0.04306559860706329
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,32,2,power_law_1.01,0.047788798809051514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,32,2,power_law_1.01,0.052799999713897705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,32,2,power_law_1.01,0.054124802350997925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,32,2,power_law_1.2,1.7543615341186523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,32,2,power_law_1.01,0.07121279835700989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,64,power_law_1.01,0.10829440355300904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,32,2,power_law_1.01,0.08488960266113281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,32,2,power_law_1.01,0.10381439924240113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,32,2,power_law_1.01,0.13586560487747193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,32,power_law_1.01,0.0960319995880127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,32,2,power_law_1.01,0.17603839635849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,32,2,power_law_1.01,0.20753920078277588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,32,2,power_law_1.01,0.2892224073410034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,32,2,power_law_1.01,0.3455615997314453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,32,2,power_law_1.2,2.259993553161621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,64,power_law_1.01,0.12094080448150635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,32,2,power_law_1.01,0.5278783798217773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,32,2,power_law_1.01,0.6717376232147216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,32,2,power_law_1.01,0.7650496006011963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,32,power_law_1.01,0.09457280039787293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,64,power_law_1.01,0.15111039876937865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,32,2,power_law_1.01,1.355583953857422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,32,2,power_law_1.2,3.480767822265625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,32,2,power_law_1.01,2.690118408203125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,32,power_law_1.01,0.09521920084953309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,64,power_law_1.01,0.17294080257415773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,64,power_law_1.01,0.2328255891799927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,32,power_law_1.01,0.1018623948097229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,32,2,power_law_1.2,6.627468872070312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,64,power_law_1.01,0.24968960285186767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,32,power_law_1.01,0.10625920295715333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,64,power_law_1.01,0.32401280403137206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,32,power_law_1.01,0.11400320529937744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,64,power_law_1.01,0.4116991996765137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,32,power_law_1.01,0.1254464030265808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,32,power_law_1.01,0.14035199880599974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,64,power_law_1.01,0.5696383953094483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,4,power_law_1.2,0.14906879663467407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,4,power_law_1.2,0.3089215993881226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,32,power_law_1.01,0.17495039701461793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,4,power_law_1.2,0.13131519556045532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,64,power_law_1.01,0.7410624027252197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,4,power_law_1.2,0.20833919048309327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,8,4,power_law_1.2,0.03953920006752014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,4,power_law_1.2,0.3013056039810181
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,8,4,power_law_1.2,0.03611519932746887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,32,power_law_1.01,0.19612159729003906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,8,4,power_law_1.2,0.03880319893360138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,4,power_law_1.2,0.37754240036010744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,8,4,power_law_1.2,0.04094719886779785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,8,4,power_law_1.2,0.042905598878860474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,4,power_law_1.2,0.5811903953552247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,2,64,power_law_1.01,0.9219327926635742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,8,4,power_law_1.2,0.0511680006980896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,32,power_law_1.01,0.25217280387878416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,4,power_law_1.2,0.5773119926452637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,8,4,power_law_1.2,0.052211201190948485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,8,4,power_law_1.2,0.05265920162200928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,4,power_law_1.2,0.6502912044525146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,8,4,power_law_1.2,0.055276799201965335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,8,4,power_law_1.2,0.05303040146827698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,4,power_law_1.2,0.6599616050720215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,32,power_law_1.01,0.2801919937133789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,8,4,power_law_1.2,0.055340802669525145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,8,4,power_law_1.2,0.05812479853630066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,4,power_law_1.2,0.6736512184143066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,8,4,power_law_1.2,0.05971199870109558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,2,64,power_law_1.01,1.4285375595092773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,8,4,power_law_1.2,0.06414080262184144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,4,power_law_1.2,0.6932159900665283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,8,4,power_law_1.2,0.07152000069618225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,32,power_law_1.01,0.3680063962936401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,4,power_law_1.2,0.7138815879821777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,8,4,power_law_1.2,0.07491840124130249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,8,4,power_law_1.2,0.0788927972316742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,4,power_law_1.2,0.7561088085174561
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,8,4,power_law_1.2,0.09490560293197632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,8,4,power_law_1.2,0.09957119822502136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,4,power_law_1.2,0.7578623771667481
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,8,4,power_law_1.2,0.11700479984283448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,32,power_law_1.01,0.4439040184020996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,8,4,power_law_1.2,0.1455296039581299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,4,power_law_1.2,0.7605567932128906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,8,4,power_law_1.2,0.18030719757080077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,4,power_law_1.2,0.8548159599304199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,8,4,power_law_1.2,0.2371583938598633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,8,4,power_law_1.2,0.32597761154174804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,4,power_law_1.2,0.9262592315673828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,8,4,power_law_1.2,0.3981503963470459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,32,power_law_1.01,0.6150400161743164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,2,64,power_law_1.01,2.790425682067871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,4,power_law_1.2,0.9659328460693359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,8,4,power_law_1.2,0.6387008190155029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,4,power_law_1.2,1.101580810546875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,8,4,power_law_1.2,0.9876607894897461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,4,power_law_1.2,1.0499391555786133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,8,4,power_law_1.2,0.9344832420349121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,32,power_law_1.01,0.7797311782836914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,4,power_law_1.2,1.2354111671447754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,8,4,power_law_1.2,1.7055744171142577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,4,power_law_1.2,1.3773887634277344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,4,power_law_1.2,1.6852096557617187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,1,32,power_law_1.01,1.0463104248046875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,8,4,power_law_1.2,3.80261116027832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,4,power_law_1.2,1.923321533203125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,4,power_law_1.2,2.7456064224243164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,32,2,power_law_1.01,0.04831359982490539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,4,power_law_1.2,3.285676956176758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,1,32,power_law_1.01,1.5466879844665526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,32,2,power_law_1.01,0.05219200253486633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,32,2,power_law_1.01,0.05428479909896851
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,4,power_law_1.2,3.8141761779785157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,32,2,power_law_1.01,0.061388802528381345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,32,2,power_law_1.01,0.07528960108757018
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,32,2,power_law_1.01,0.09168000221252441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,4,power_law_1.2,5.688224029541016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,32,2,power_law_1.01,0.1102463960647583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,32,2,power_law_1.01,0.11671040058135987
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,32,2,power_law_1.01,0.11838079690933227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,8,power_law_1.01,0.07162879705429077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,32,2,power_law_1.01,0.12113280296325683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,1,32,power_law_1.01,2.9157440185546877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,32,2,power_law_1.01,0.1247871994972229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,8,power_law_1.01,0.09303680062294006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,32,2,power_law_1.01,0.1260032057762146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,4,power_law_1.2,10.25551986694336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,8,power_law_1.01,0.07565439939498901
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,32,2,power_law_1.01,0.13232640027999878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,8,power_law_1.01,0.06947199702262878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,32,2,power_law_1.01,0.13715840578079225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,8,power_law_1.01,0.08503040075302123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,32,2,power_law_1.01,0.14536960124969484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,8,power_law_1.01,0.09415040016174317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,32,2,power_law_1.01,0.14855040311813356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,32,2,power_law_1.01,0.15908479690551758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,8,power_law_1.01,0.1197119951248169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,32,2,power_law_1.01,0.18156800270080567
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,8,power_law_1.01,0.12997119426727294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,8,power_law_1.01,0.12817920446395875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,32,2,power_law_1.01,0.20967040061950684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,8,power_law_1.01,0.13030400276184081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,32,2,power_law_1.01,0.26019198894500734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,8,power_law_1.01,0.13514879941940308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,32,2,power_law_1.01,0.3042304039001465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,8,power_law_1.01,0.1416767954826355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,32,2,power_law_1.01,0.39207680225372316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,8,power_law_1.01,0.14280320405960084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,32,2,power_law_1.01,0.4981503963470459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,8,power_law_1.01,0.15025919675827026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,1,power_law_1.2,0.05249279737472534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,32,2,power_law_1.01,0.6925055980682373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,8,power_law_1.01,0.15493760108947754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,8,power_law_1.2,0.05065600275993347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,1,power_law_1.2,0.06555520296096802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,8,power_law_1.01,0.16100480556488037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,32,2,power_law_1.01,0.8270079612731933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,8,power_law_1.2,0.07041280269622803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,8,power_law_1.01,0.16073600053787232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,1,power_law_1.2,0.08842880129814149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,8,power_law_1.2,0.050918400287628174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,32,2,power_law_1.01,1.1476927757263184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,8,power_law_1.01,0.1842687964439392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,1,power_law_1.2,0.1140544056892395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,8,power_law_1.2,0.05632640123367309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,8,power_law_1.01,0.186188805103302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,8,power_law_1.2,0.06298879981040954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,32,2,power_law_1.01,1.533619213104248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,1,power_law_1.2,0.14440959692001343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,8,power_law_1.01,0.21543679237365723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,8,power_law_1.2,0.07498239874839782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,1,power_law_1.2,0.25589759349823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,8,power_law_1.01,0.24487040042877198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,8,power_law_1.2,0.08799999952316284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,32,2,power_law_1.01,1.848966407775879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,1,power_law_1.2,0.2668287992477417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,8,power_law_1.01,0.32042880058288575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,8,power_law_1.2,0.09402880072593689
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,1,power_law_1.2,0.283027195930481
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,8,power_law_1.01,0.3318527936935425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,8,power_law_1.2,0.09595519900321961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,32,2,power_law_1.01,2.873459243774414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,1,power_law_1.2,0.2943552017211914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,8,power_law_1.01,0.4692351818084717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,8,power_law_1.2,0.09759359955787658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,1,power_law_1.2,0.29430398941040037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,8,power_law_1.2,0.09853439927101135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,8,power_law_1.01,0.5438528060913086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,8,power_law_1.2,0.10177919864654542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,1,power_law_1.2,0.3178303956985474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,8,power_law_1.01,0.7383039951324463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,8,power_law_1.2,0.1072383999824524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,1,power_law_1.2,0.3196415901184082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,32,2,power_law_1.01,5.737619018554687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,8,power_law_1.2,0.10757759809494019
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,1,power_law_1.2,0.3309823989868164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,8,power_law_1.01,0.9116095542907715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,8,power_law_1.2,0.11781120300292969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,1,power_law_1.2,0.342303991317749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,8,power_law_1.2,0.12376320362091064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,8,power_law_1.01,1.111680030822754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,1,power_law_1.2,0.35964159965515136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,8,power_law_1.2,0.13319040536880494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,1,power_law_1.2,0.37147519588470457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,8,power_law_1.2,0.14723199605941772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,8,power_law_1.01,1.5505727767944335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,1,power_law_1.2,0.3833728075027466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,8,power_law_1.2,0.17184000015258788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,1,power_law_1.2,0.4401535987854004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,8,power_law_1.2,0.20851199626922606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,1,power_law_1.2,0.4912703990936279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,8,power_law_1.2,0.2271807909011841
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,1,power_law_1.2,0.5953536033630371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,8,power_law_1.01,3.2994815826416017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,8,power_law_1.2,0.28266880512237547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,1,power_law_1.2,0.7116288185119629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,8,power_law_1.2,0.3572351932525635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,1,power_law_1.2,0.9092608451843261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,8,power_law_1.2,0.5055488109588623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,1,power_law_1.2,1.0423551559448243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,8,power_law_1.2,0.5898367881774902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,64,power_law_1.2,0.13894399404525756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,1,power_law_1.2,1.3614784240722657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,64,power_law_1.2,0.07209600210189819
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,8,power_law_1.2,0.8612992286682128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,1,power_law_1.2,1.7533248901367187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,64,power_law_1.2,0.052723199129104614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,8,power_law_1.2,1.2020607948303224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,64,power_law_1.2,0.05665919780731201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,1,power_law_1.2,2.4176639556884765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,64,power_law_1.2,0.05674239993095398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,8,power_law_1.2,1.3677311897277833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,64,power_law_1.2,0.06474239826202392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,1,power_law_1.2,3.0921024322509765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,64,power_law_1.2,0.06947199702262878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,64,power_law_1.2,0.06712960004806519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,8,power_law_1.2,2.1856576919555666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,64,power_law_1.2,0.06970880031585694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,1,power_law_1.2,3.756435012817383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,64,power_law_1.2,0.0723904013633728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,8,power_law_1.01,0.048332801461219786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,64,power_law_1.2,0.07243520021438599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,8,power_law_1.01,0.07093759775161743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,64,power_law_1.2,0.07519360184669495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,8,power_law_1.2,4.31957778930664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,1,power_law_1.2,5.571072006225586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,64,power_law_1.2,0.07562879920005798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,8,power_law_1.01,0.09979519844055176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,64,power_law_1.2,0.07608320116996765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,8,power_law_1.01,0.05596799850463867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,64,power_law_1.2,0.08453119993209839
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,64,power_law_1.2,0.0877951979637146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,8,power_law_1.01,0.06740480065345764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,64,power_law_1.2,0.09047679901123047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,8,power_law_1.01,0.08483200073242188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,64,power_law_1.2,0.10336639881134033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,1,power_law_1.2,10.795142364501952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,8,power_law_1.01,0.09376000165939331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,64,power_law_1.2,0.11333119869232178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,64,power_law_1.2,0.13698559999465942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,8,power_law_1.01,0.10485119819641113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,64,power_law_1.2,0.1527232050895691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,8,power_law_1.01,0.1049407958984375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,64,power_law_1.2,0.2096384048461914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,8,power_law_1.01,0.11102720499038696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,64,power_law_1.2,0.22072319984436034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,8,power_law_1.01,0.11841280460357666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,64,power_law_1.2,0.31427199840545655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,4,2,power_law_1.2,0.07390080094337463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,4,2,power_law_1.2,0.10974080562591552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,4,2,power_law_1.2,0.08672639727592468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,4,2,power_law_1.2,0.12019840478897095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,4,2,power_law_1.2,0.11747200489044189
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,4,2,power_law_1.2,0.11644799709320068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,64,power_law_1.2,0.496889591217041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,8,power_law_1.01,0.11849600076675415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,4,2,power_law_1.2,0.1265727996826172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,4,2,power_law_1.2,0.12497279644012452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,4,2,power_law_1.2,0.1284991979598999
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,4,2,power_law_1.2,0.134278404712677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,4,2,power_law_1.2,0.13055360317230225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,4,2,power_law_1.2,0.13345279693603515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,4,2,power_law_1.2,0.1351359963417053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,8,power_law_1.01,0.11916799545288086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,64,power_law_1.2,0.6900479793548584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,4,2,power_law_1.2,0.14387199878692628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,4,2,power_law_1.2,0.14914560317993164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,4,2,power_law_1.2,0.15445760488510132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,4,2,power_law_1.2,0.16935039758682252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,4,2,power_law_1.2,0.19839359521865846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,4,2,power_law_1.2,0.21870720386505127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,8,power_law_1.01,0.12558720111846924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,4,2,power_law_1.2,0.2941119909286499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,64,power_law_1.2,0.6812096118927002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,4,2,power_law_1.2,0.37230720520019533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,4,2,power_law_1.2,0.455295991897583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,4,2,power_law_1.2,0.6647232055664063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,8,power_law_1.01,0.1364799976348877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,4,2,power_law_1.2,0.7795648097991943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,2,power_law_1.01,0.030099201202392577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,4,2,power_law_1.2,1.0703167915344238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,64,power_law_1.2,0.9054464340209961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,4,2,power_law_1.2,1.5180352210998536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,8,power_law_1.01,0.13905279636383056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,2,power_law_1.01,0.03623040020465851
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,4,2,power_law_1.2,2.495008087158203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,8,power_law_1.01,0.1484992027282715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,64,power_law_1.2,1.458176040649414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,2,power_law_1.01,0.03880319893360138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,4,2,power_law_1.2,2.769478416442871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,8,power_law_1.01,0.16805119514465333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,4,2,power_law_1.2,4.342105484008789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,2,power_law_1.01,0.04472320079803467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,8,power_law_1.01,0.1791808009147644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,2,power_law_1.01,0.050995200872421265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,64,power_law_1.2,2.8011903762817383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,4,2,power_law_1.2,7.254093170166016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,8,power_law_1.01,0.21918718814849852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,2,power_law_1.01,0.06652160286903382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,2,power_law_1.01,0.06888319849967957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,8,power_law_1.01,0.24361600875854492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,2,power_law_1.01,0.06912639737129211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,8,power_law_1.01,0.30947840213775635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,2,power_law_1.01,0.07262719869613647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,2,power_law_1.01,0.0738431990146637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,8,power_law_1.01,0.37742080688476565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,2,power_law_1.01,0.07371519804000855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,2,power_law_1.01,0.07537279725074768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,8,power_law_1.01,0.4957759857177734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,2,power_law_1.01,0.07524480223655701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,2,power_law_1.01,0.07739520072937012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,8,power_law_1.01,0.6360703945159912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,2,power_law_1.01,0.08793600201606751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,2,power_law_1.01,0.09697920083999634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,8,power_law_1.01,0.88919677734375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,2,power_law_1.01,0.09807999730110169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,2,power_law_1.01,0.10805759429931641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,8,power_law_1.01,1.1910719871520996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,2,power_law_1.01,0.11479040384292602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,2,power_law_1.01,0.13768960237503053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,2,power_law_1.01,0.1039423942565918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,2,power_law_1.01,0.14358400106430053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,2,power_law_1.01,0.1813696026802063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,8,power_law_1.01,1.3762880325317384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,32,power_law_1.2,0.11205120086669922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,2,power_law_1.01,0.11950080394744873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,2,power_law_1.01,0.17365119457244874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,2,power_law_1.01,0.20473599433898926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,2,power_law_1.01,0.20519039630889893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,2,power_law_1.01,0.24588799476623535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,32,power_law_1.2,0.09401599764823913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,2,power_law_1.01,0.26333439350128174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,2,power_law_1.01,0.44872322082519533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,8,power_law_1.01,2.138041687011719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,2,power_law_1.01,0.481279993057251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,2,power_law_1.01,0.3267199993133545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,32,power_law_1.2,0.062483197450637816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,2,power_law_1.01,0.4987328052520752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,2,power_law_1.01,0.431276798248291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,2,power_law_1.01,0.4972479820251465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,2,power_law_1.01,0.5044159889221191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,32,power_law_1.2,0.06828160285949707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,2,power_law_1.01,0.5373631954193115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,2,power_law_1.01,0.5262464046478271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,8,power_law_1.01,4.0342144012451175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,2,power_law_1.01,0.536300802230835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,4,2,power_law_1.01,0.6887231826782226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,32,power_law_1.2,0.07124480009078979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,2,power_law_1.01,0.5548416137695312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,2,power_law_1.01,0.5700160026550293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,4,2,power_law_1.01,1.0042688369750976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,32,power_law_1.2,0.0847487986087799
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,2,power_law_1.01,0.5726336002349853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,2,power_law_1.01,0.6106175899505615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,2,power_law_1.01,0.6035776138305664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,32,power_law_1.2,0.08428159952163697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,4,2,power_law_1.01,1.8992448806762696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,2,power_law_1.01,0.6783423900604248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,2,power_law_1.01,0.7073215961456298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,32,power_law_1.2,0.08780800104141236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,2,power_law_1.01,0.8194239616394043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,2,power_law_1.01,0.9053183555603027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,32,power_law_1.2,0.08921599984169007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,2,power_law_1.01,1.078451156616211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,32,power_law_1.2,0.0909056007862091
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,2,power_law_1.01,1.170841598510742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,2,power_law_1.01,1.3479616165161132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,32,power_law_1.2,0.09088640213012696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,2,power_law_1.01,1.8548799514770509
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,2,balanced,0.042410666743914284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,2,balanced,0.04009066770474116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,2,balanced,0.046298667788505554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,2,balanced,0.05234666665395101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,2,balanced,0.07588266829649608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,2,balanced,0.07740800082683563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,2,balanced,0.07920533418655396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,2,balanced,0.0798826664686203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,2,balanced,0.0800799975792567
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,2,balanced,0.08091199894746144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,2,balanced,0.0809333324432373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,32,power_law_1.2,0.09536640048027038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,2,balanced,0.08569066723187764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,2,balanced,0.08553600311279297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,2,balanced,0.08661866188049316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,2,balanced,0.09212266405423482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,2,balanced,0.09474666913350423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,2,balanced,0.09920533498128255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,2,balanced,0.10851732889811198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,2,balanced,0.12043199936548869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,2,balanced,0.14128533005714417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,2,balanced,0.1605226695537567
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,2,power_law_1.01,2.4819135665893555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,2,balanced,0.2188160022099813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,2,balanced,0.2534079949061076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,2,balanced,0.3648000160853068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,2,balanced,0.46042664845784503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,2,balanced,0.6631306807200114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,2,balanced,0.8635413646697998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,2,balanced,1.066853364308675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,32,power_law_1.2,0.09520000219345093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,2,balanced,1.679093360900879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,2,power_law_1.01,2.9544960021972657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,2,balanced,3.241781234741211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,32,power_law_1.2,0.09775360226631165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,2,power_law_1.01,3.594976043701172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,32,power_law_1.2,0.10375039577484131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,2,power_law_1.2,0.06386560201644897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,2,power_law_1.01,4.953164672851562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,32,power_law_1.2,0.10840320587158203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,2,power_law_1.2,0.06945279836654664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,32,power_law_1.2,0.1141759991645813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,2,power_law_1.2,0.07029119729995728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,2,power_law_1.2,0.0795199990272522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,32,power_law_1.2,0.131167995929718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,2,power_law_1.01,9.460550689697266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,2,power_law_1.2,0.08867200016975403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,32,power_law_1.2,0.15025919675827026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,16,8,balanced,0.03766933331886927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,16,8,balanced,0.03372266640265783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,16,8,balanced,0.033402666449546814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,16,8,balanced,0.03398400048414866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,16,8,balanced,0.03342933456103007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,16,8,balanced,0.03349866718053818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,16,8,balanced,0.03537066777547201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,16,8,balanced,0.0354666660229365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,2,power_law_1.2,0.10291199684143067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,16,8,balanced,0.035530666510264076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,16,8,balanced,0.033973333736260734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,16,8,balanced,0.035877334574858345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,16,8,balanced,0.03758399933576584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,16,8,balanced,0.03763733307520548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,16,8,balanced,0.03739733248949051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,16,8,balanced,0.03952533255020777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,16,8,balanced,0.04171733558177948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,16,8,balanced,0.041482667128245033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,16,8,balanced,0.04197866717974345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,16,8,balanced,0.04574400186538696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,16,8,balanced,0.04981866478919983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,32,power_law_1.2,0.18055039644241333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,16,8,balanced,0.0518453319867452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,16,8,balanced,0.062074666221936546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,16,8,balanced,0.07168533404668172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,16,8,balanced,0.09402133027712505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,16,8,balanced,0.1111893355846405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,16,8,balanced,0.14692800243695578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,16,8,balanced,0.17532267173131308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,16,8,balanced,0.21155200401941934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,16,8,balanced,0.3085920015970866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,16,8,balanced,0.5688266754150391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,2,power_law_1.2,0.13240959644317626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,32,power_law_1.2,0.21064319610595703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,2,power_law_1.2,0.13984639644622804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,2,power_law_1.2,0.1398144006729126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,32,power_law_1.2,0.28044159412384034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,2,power_law_1.2,0.14741120338439942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,32,power_law_1.2,0.3011199951171875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,2,power_law_1.2,0.15073280334472655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,2,power_law_1.2,0.15720959901809692
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,32,power_law_1.2,0.3952640056610107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,2,power_law_1.2,0.15568000078201294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,2,power_law_1.2,0.1651584029197693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,256,power_law_1.01,0.06503679752349853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,32,power_law_1.2,0.5140607833862305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,2,power_law_1.2,0.17719039916992188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,256,power_law_1.01,0.06578559875488281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,2,power_law_1.2,0.17829760313034057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,32,power_law_1.2,0.7097919940948486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,256,power_law_1.01,0.06361600160598754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,1,power_law_1.2,0.14393600225448608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,2,power_law_1.2,0.19183360338211058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,1,power_law_1.2,0.24033920764923095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,2,power_law_1.2,0.2246272087097168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,256,power_law_1.01,0.06462079882621766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,32,power_law_1.2,0.9843008041381835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,1,power_law_1.2,0.33895039558410645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,2,power_law_1.2,0.24583039283752442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,256,power_law_1.01,0.06465280055999756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,1,power_law_1.2,0.5620543956756592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,2,power_law_1.2,0.2956928014755249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,256,power_law_1.01,0.06590719819068909
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,1,power_law_1.2,1.0016448020935058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,2,power_law_1.2,0.341919994354248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,1,32,power_law_1.2,1.3477120399475098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,256,power_law_1.01,0.06575999855995178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,2,power_law_1.2,0.44671359062194826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,1,power_law_1.2,1.523097610473633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,256,power_law_1.01,0.06627839803695679
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,2,power_law_1.2,0.5504320144653321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,1,power_law_1.2,1.8296640396118165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,256,power_law_1.01,0.06512640118598938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,2,power_law_1.2,0.7738175868988038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,1,32,power_law_1.2,2.0451967239379885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,1,power_law_1.2,2.7171327590942385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,256,power_law_1.01,0.06600959897041321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,2,power_law_1.2,0.9405632019042969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,256,power_law_1.01,0.06559360027313232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,1,power_law_1.2,2.949977684020996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,2,power_law_1.2,1.387673568725586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,256,power_law_1.01,0.06876800060272217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,1,power_law_1.2,2.9880767822265626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,2,power_law_1.2,1.8309247970581055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,256,power_law_1.01,0.06883839964866638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,1,power_law_1.2,3.2452415466308593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,256,power_law_1.01,0.06922240257263183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,2,power_law_1.2,2.153183937072754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,1,power_law_1.2,3.4080127716064452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,1,32,power_law_1.2,4.36126708984375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,256,power_law_1.01,0.07454720139503479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,1,power_law_1.2,3.4474815368652343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,256,power_law_1.01,0.07546240091323853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,2,power_law_1.2,3.5203136444091796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,256,power_law_1.01,0.07991679906845092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,1,power_law_1.2,3.558745574951172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,256,power_law_1.01,0.09247360229492188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,1,power_law_1.2,3.7634815216064452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,2,power_law_1.2,7.2607872009277346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,256,power_law_1.01,0.10650880336761474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,1,power_law_1.2,3.807865524291992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,256,power_law_1.01,0.12830719947814942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,1,power_law_1.2,3.983379364013672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,256,power_law_1.01,0.15672320127487183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,1,power_law_1.2,4.316019058227539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,256,power_law_1.01,0.2038975954055786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,2,power_law_1.01,0.05320320129394531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,2,power_law_1.01,0.07318400144577027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,1,power_law_1.2,4.63807373046875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,256,power_law_1.01,0.21772160530090331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,2,power_law_1.01,0.09968000054359435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,2,power_law_1.01,0.08166400194168091
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,1,power_law_1.2,5.262271881103516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,256,power_law_1.01,0.28691840171813965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,2,power_law_1.01,0.12391040325164795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,1,4,power_law_1.01,0.08678399920463561
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,1,4,power_law_1.01,0.10039039850234985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,1,power_law_1.2,5.402553558349609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,256,power_law_1.01,0.3601599931716919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,1,4,power_law_1.01,0.0955839991569519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,2,power_law_1.01,0.16197119951248168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,1,4,power_law_1.01,0.09899520277976989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,1,4,power_law_1.01,0.09961599707603455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,1,4,power_law_1.01,0.10041600465774536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,2,power_law_1.01,0.19629440307617188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,1,4,power_law_1.01,0.10041600465774536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,1,4,power_law_1.01,0.10053119659423829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,1,4,power_law_1.01,0.1022528052330017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,1,power_law_1.2,6.299488067626953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,1,4,power_law_1.01,0.10380159616470337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,256,power_law_1.01,0.48706560134887694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,2,power_law_1.01,0.235481595993042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,1,4,power_law_1.01,0.10464639663696289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,1,4,power_law_1.01,0.11424640417099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,1,4,power_law_1.01,0.11676160097122193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,1,4,power_law_1.01,0.12887040376663209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,2,power_law_1.01,0.25333759784698484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,1,4,power_law_1.01,0.15330560207366944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,1,4,power_law_1.01,0.1602687954902649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,1,4,power_law_1.01,0.15882879495620728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,1,power_law_1.2,6.169523239135742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,1,4,power_law_1.01,0.20325119495391847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,2,power_law_1.01,0.2556735992431641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,256,power_law_1.01,0.637721586227417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,1,4,power_law_1.01,0.22953600883483888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,1,4,power_law_1.01,0.2965951919555664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,1,4,power_law_1.01,0.440447998046875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,2,power_law_1.01,0.2680000066757202
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,1,4,power_law_1.01,0.531820821762085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,1,4,power_law_1.01,0.6591807842254639
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,1,4,power_law_1.01,0.9180352210998535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,1,power_law_1.2,7.203558349609375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,2,power_law_1.01,0.2809344053268433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,1,4,power_law_1.01,1.1491328239440919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,1,256,power_law_1.01,0.7455359935760498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,1,4,power_law_1.01,2.349228858947754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,2,power_law_1.01,0.2897599935531616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,1,4,power_law_1.01,2.823859214782715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,2,power_law_1.01,0.30317440032958987
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,1,power_law_1.2,8.423564910888672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,1,4,power_law_1.01,3.191103935241699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,2,power_law_1.01,0.31429760456085204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,1,4,power_law_1.01,5.431436920166016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,2,power_law_1.01,0.3333888053894043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,1,256,power_law_1.01,1.203878402709961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,2,power_law_1.01,0.3448064088821411
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,1,power_law_1.2,10.177369689941406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,1,4,power_law_1.01,8.887967681884765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,2,power_law_1.01,0.3801791906356812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,2,power_law_1.01,0.4067071914672852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,2,power_law_1.01,0.47203841209411623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,1,power_law_1.2,12.125440216064453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,2,power_law_1.01,0.5177663803100586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,1,256,power_law_1.01,2.441196823120117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,2,power_law_1.01,0.6442240238189697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,2,power_law_1.01,0.7825791835784912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,1,power_law_1.2,14.358489990234375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,2,power_law_1.01,1.036076831817627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,2,power_law_1.01,1.2706751823425293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,2,power_law_1.01,1.7055679321289063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,1,power_law_1.2,17.648434448242188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,2,power_law_1.01,2.115430450439453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,64,power_law_1.01,0.08055040240287781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,32,8,power_law_1.01,0.03653759956359863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,32,8,power_law_1.01,0.031200000643730165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,2,power_law_1.01,2.523449516296387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,32,8,power_law_1.01,0.030604800581932066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,64,power_law_1.01,0.07839360237121581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,32,8,power_law_1.01,0.030803200602531434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,32,8,power_law_1.01,0.03146879971027374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,32,8,power_law_1.01,0.0338239997625351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,2,power_law_1.01,3.643417739868164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,32,8,power_law_1.01,0.033964800834655764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,32,8,power_law_1.01,0.03407999873161316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,64,power_law_1.01,0.056454402208328244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,32,8,power_law_1.01,0.03472639918327332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,32,8,power_law_1.01,0.034457600116729735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,32,8,power_law_1.01,0.035123199224472046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,32,8,power_law_1.01,0.03546879887580871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,32,8,power_law_1.01,0.03596799969673157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,1,power_law_1.2,33.4505615234375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,32,8,power_law_1.01,0.03685120046138764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,64,power_law_1.01,0.062028801441192626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,32,8,power_law_1.01,0.04177280068397522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,2,power_law_1.01,6.794585418701172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,32,8,power_law_1.01,0.042854401469230655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,32,8,power_law_1.01,0.04687359929084778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,32,8,power_law_1.01,0.052051198482513425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,64,power_law_1.01,0.06730239987373351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,32,8,power_law_1.01,0.05627520084381103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,32,8,power_law_1.01,0.07057279944419861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,32,8,power_law_1.01,0.07635200023651123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,32,8,power_law_1.01,0.09784319996833801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,64,power_law_1.01,0.0735040009021759
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,32,8,power_law_1.01,0.12718080282211303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,32,8,power_law_1.01,0.17815680503845216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,32,8,power_law_1.01,0.2136064052581787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,64,power_law_1.01,0.07537279725074768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,32,8,power_law_1.01,0.31096959114074707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,32,8,power_law_1.01,0.4525184154510498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,64,power_law_1.01,0.0785215973854065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,32,8,power_law_1.01,0.5898496150970459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,32,8,power_law_1.01,0.8472064018249512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,1,balanced,0.08524800340334575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,1,balanced,0.18127467234929404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,1,balanced,0.2266026735305786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,64,power_law_1.01,0.07733759880065919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,1,balanced,0.4110399881998698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,1,balanced,0.7610666751861572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,1,balanced,0.7519786357879639
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,1,balanced,0.7491947015126547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,1,balanced,0.7677280108133951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,1,balanced,0.7555946509043375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,1,balanced,0.761519988377889
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,1,balanced,0.7784000237782797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,1,balanced,0.7646559874216715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,1,balanced,0.772816022237142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,1,balanced,0.7838559945424398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,1,balanced,0.7763093312581381
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,4,power_law_1.2,0.06056960225105286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,1,balanced,0.8001973628997803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,1,balanced,0.8091893196105957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,32,8,power_law_1.01,1.7426944732666017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,64,power_law_1.01,0.07724159955978394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,1,balanced,0.8343839645385742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,1,balanced,0.8595893383026123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,1,balanced,0.911461353302002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,1,balanced,1.0050933361053467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,1,balanced,1.2415306568145752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,1,balanced,1.2406773567199707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,4,power_law_1.2,0.07106559872627258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,1,balanced,1.7007519404093425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,1,balanced,1.9859840075174968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,64,power_law_1.01,0.07834240198135375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,1,balanced,2.8272905349731445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,4,power_law_1.2,0.07071999907493591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,1,balanced,3.6683200200398765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,1,balanced,4.594335873921712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,4,power_law_1.2,0.0723136007785797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,64,power_law_1.01,0.08229119777679443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,1,balanced,9.299525578816732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,4,power_law_1.2,0.08957440257072449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,64,power_law_1.01,0.08225280046463013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,4,power_law_1.2,0.10456960201263428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,4,power_law_1.2,0.11416319608688355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,1,balanced,19.21294403076172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,64,power_law_1.01,0.08320000171661376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,4,power_law_1.2,0.13400319814682007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,4,1,power_law_1.01,0.05432959794998169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,64,power_law_1.01,0.09142400026321411
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,4,1,power_law_1.01,0.08065919876098633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,4,1,power_law_1.01,0.088646399974823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,4,power_law_1.2,0.14160640239715577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,4,1,power_law_1.01,0.11416319608688355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,4,1,power_law_1.01,0.11607040166854858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,4,1,power_law_1.01,0.11806080341339112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,4,1,power_law_1.01,0.12004480361938477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,4,1,power_law_1.01,0.12027519941329956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,4,1,power_law_1.01,0.12645119428634644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,4,power_law_1.2,0.1469375967979431
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,4,1,power_law_1.01,0.12910720109939575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,4,1,power_law_1.01,0.1276479959487915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,64,power_law_1.01,0.09352959990501404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,4,1,power_law_1.01,0.13207039833068848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,4,1,power_law_1.01,0.1382143974304199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,4,1,power_law_1.01,0.14161920547485352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,4,1,power_law_1.01,0.15391360521316527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,4,1,power_law_1.01,0.16666239500045776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,4,power_law_1.2,0.15303679704666137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,4,1,power_law_1.01,0.1674623966217041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,4,1,power_law_1.01,0.19761279821395875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,4,1,power_law_1.01,0.21854081153869628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,4,1,power_law_1.01,0.2703423976898193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,4,1,power_law_1.01,0.319270396232605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,64,power_law_1.01,0.09915519952774048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,4,1,power_law_1.01,0.4199808120727539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,4,power_law_1.2,0.154367995262146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,4,1,power_law_1.01,0.5155136108398437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,4,1,power_law_1.01,0.7113088130950928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,4,1,power_law_1.01,0.9193216323852539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,4,power_law_1.2,0.16053760051727295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,4,1,power_law_1.01,1.310854434967041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,64,power_law_1.01,0.11884160041809082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,4,1,power_law_1.01,1.6954687118530274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,4,power_law_1.2,0.17258880138397217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,4,1,power_law_1.01,2.086079978942871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,4,power_law_1.2,0.177510404586792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,4,1,power_law_1.01,3.2482112884521483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,4,4,balanced,0.05584000051021576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,4,4,balanced,0.0661599983771642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,64,power_law_1.01,0.1189120054244995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,4,4,balanced,0.08265066643555959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,4,4,balanced,0.07672533392906189
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,4,4,balanced,0.07712000111738841
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,4,4,balanced,0.0767626663049062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,4,4,balanced,0.07738666733105977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,4,4,balanced,0.07826133569081624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,4,4,balanced,0.07812800010045369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,4,4,balanced,0.0776693324247996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,4,4,balanced,0.08051200211048126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,4,4,balanced,0.07886399825414021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,4,4,balanced,0.0805866668621699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,4,4,balanced,0.08448533217112224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,4,4,balanced,0.0927946666876475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,4,4,balanced,0.09305066863695781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,4,4,balanced,0.09690666198730469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,4,4,balanced,0.11020800471305847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,4,4,balanced,0.11971732974052429
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,4,4,balanced,0.14655466874440512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,4,4,balanced,0.15868266423543295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,4,4,balanced,0.19619200627009073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,4,power_law_1.2,0.18062080144882203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,4,4,balanced,0.2320586641629537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,4,4,balanced,0.3160746693611145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,4,4,balanced,0.3784213463465373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,4,4,balanced,0.5565280119578043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,4,4,balanced,0.7053120136260986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,4,1,power_law_1.01,6.3578369140625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,4,4,balanced,0.8544639746348063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,4,4,balanced,1.3484533627827961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,64,power_law_1.01,0.1470080018043518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,4,power_law_1.2,0.19493119716644286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,4,4,balanced,2.5861600240071616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,4,power_law_1.2,0.215231990814209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,64,power_law_1.01,0.1575935959815979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,4,power_law_1.2,0.2307584047317505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,4,power_law_1.2,0.28072319030761717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,64,power_law_1.01,0.2069632053375244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,4,power_law_1.2,0.313971209526062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,4,power_law_1.2,0.4050943851470947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,64,power_law_1.01,0.21573119163513182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,4,power_law_1.2,0.45601282119750974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,64,power_law_1.01,0.29333760738372805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,4,power_law_1.2,0.6231359958648681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,4,power_law_1.2,0.8256896018981934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,16,4,power_law_1.01,0.0369024008512497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,32,8,balanced,0.03774933268626531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,64,power_law_1.01,0.32655360698699953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,32,8,balanced,0.03742400060097376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,32,8,balanced,0.0354720006386439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,16,4,power_law_1.01,0.03705599904060364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,32,8,balanced,0.035887998839219414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,32,8,balanced,0.03573333223660787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,32,8,balanced,0.03557866563399633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,32,8,balanced,0.03552533437808355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,32,8,balanced,0.035605333745479584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,16,4,power_law_1.01,0.03505280017852783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,32,8,balanced,0.03575466573238373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,32,8,balanced,0.03537066777547201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,32,8,balanced,0.03753600021203359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,32,8,balanced,0.03769599894682566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,16,4,power_law_1.01,0.035692799091339114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,32,8,balanced,0.03772266705830892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,32,8,balanced,0.03749333322048187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,32,8,balanced,0.04174399872620901
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,32,8,balanced,0.041738669077555336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,16,4,power_law_1.01,0.035462400317192076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,32,8,balanced,0.04350399971008301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,32,8,balanced,0.04572799801826477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,4,power_law_1.2,1.1448896408081055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,32,8,balanced,0.04970666766166687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,16,4,power_law_1.01,0.03648639917373657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,32,8,balanced,0.051776001850763954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,32,8,balanced,0.05602133274078369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,32,8,balanced,0.07268799841403961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,16,4,power_law_1.01,0.03697920143604279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,32,8,balanced,0.08665066957473755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,32,8,balanced,0.10963199536005656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,32,8,balanced,0.130730668703715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,16,4,power_law_1.01,0.03705599904060364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,32,8,balanced,0.1779786745707194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,32,8,balanced,0.22136000792185465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,16,4,power_law_1.01,0.037651199102401736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,32,8,balanced,0.2624640067418416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,32,8,balanced,0.3816106716791789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,16,4,power_law_1.01,0.037945601344108584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,16,4,power_law_1.01,0.03888640105724335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,32,8,balanced,0.7187999884287516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,16,4,power_law_1.01,0.03961600065231323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,16,4,power_law_1.01,0.041945600509643556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,16,4,power_law_1.01,0.0427264004945755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,16,4,power_law_1.01,0.049184000492095946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,16,4,power_law_1.01,0.04938240051269531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,4,power_law_1.2,1.4328767776489257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,16,4,power_law_1.01,0.05124480128288269
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,64,power_law_1.01,0.4525311946868896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,16,4,power_law_1.01,0.061868798732757566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,16,4,power_law_1.01,0.0686464011669159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,16,4,power_law_1.01,0.08669440150260925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,16,4,power_law_1.01,0.09828479886054993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,16,4,power_law_1.01,0.15170559883117676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,16,4,power_law_1.01,0.190009605884552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,16,4,power_law_1.01,0.22201600074768066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,16,4,power_law_1.01,0.2838912010192871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,4,power_law_1.2,1.7469568252563477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,16,4,power_law_1.01,0.40580477714538576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,16,4,power_law_1.01,0.5625152111053466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,16,4,power_law_1.01,0.7178688049316406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,64,power_law_1.01,0.5513343811035156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,16,4,power_law_1.01,1.1051008224487304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,4,power_law_1.2,2.769267272949219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,16,4,power_law_1.01,1.9903999328613282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,2,64,power_law_1.01,0.644320011138916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,4,power_law_1.2,5.195161437988281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,16,power_law_1.2,0.06283519864082336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,2,64,power_law_1.01,0.965664005279541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,16,power_law_1.2,0.09310719966888428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,16,power_law_1.2,0.08574720025062561
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,16,power_law_1.2,0.059059202671051025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,16,balanced,0.05403199791908264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,16,balanced,0.054117331902186074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,16,balanced,0.054042667150497437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,16,balanced,0.05602133274078369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,16,balanced,0.05786666770776113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,16,balanced,0.06238933404286703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,16,balanced,0.0702453354994456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,16,balanced,0.06840533514817555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,16,balanced,0.06866666674613953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,16,balanced,0.0702400008837382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,16,balanced,0.0703306645154953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,16,balanced,0.07238399982452393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,16,power_law_1.2,0.06286720037460328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,16,balanced,0.07275199890136719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,16,balanced,0.07410133381684621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,16,balanced,0.08082666496435802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,16,balanced,0.07841599980990092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,16,balanced,0.0846560001373291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,16,balanced,0.09492799639701843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,16,balanced,0.09904000163078308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,16,balanced,0.12149332960446675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,16,balanced,0.13826666275660196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,16,balanced,0.17323199907938638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,16,balanced,0.21123733123143515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,16,balanced,0.2712000012397766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,16,balanced,0.34462400277455646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,16,balanced,0.47036266326904297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,16,balanced,0.6071093479792277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,16,power_law_1.2,0.07481600046157837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,16,balanced,0.7513706684112549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,16,balanced,1.1485706965128581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,16,balanced,2.244661331176758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,16,power_law_1.2,0.07764480113983155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,2,64,power_law_1.01,2.073644828796387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,16,power_law_1.2,0.08803200125694274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,16,power_law_1.2,0.060108798742294314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,16,power_law_1.2,0.08681600093841553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,16,power_law_1.2,0.08241279721260071
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,16,power_law_1.2,0.05132160186767578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,16,power_law_1.2,0.08900480270385742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,16,power_law_1.2,0.05551999807357788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,16,power_law_1.2,0.09486079812049866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,16,power_law_1.2,0.055878400802612305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,16,power_law_1.2,0.09430400133132935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,16,power_law_1.2,0.06106879711151123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,16,power_law_1.2,0.06444159746170045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,16,power_law_1.2,0.09637119770050048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,16,power_law_1.2,0.06568319797515869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,16,power_law_1.2,0.10297600030899048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,16,power_law_1.2,0.06661760210990905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,16,power_law_1.2,0.1076416015625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,16,power_law_1.2,0.06643199920654297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,1,balanced,0.07474133372306824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,1,balanced,0.1157919963200887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,1,balanced,0.1329759955406189
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,1,balanced,0.19975467522939047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,1,balanced,0.33297065893809
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,1,balanced,0.4002879858016968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,1,balanced,0.4062560002009074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,1,balanced,0.40775465965270996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,16,power_law_1.2,0.0671231985092163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,1,balanced,0.40857601165771484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,1,balanced,0.41130133469899494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,16,power_law_1.2,0.10931839942932128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,1,balanced,0.4161173502604167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,1,balanced,0.42016534010569256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,1,balanced,0.4243253469467163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,1,balanced,0.4318133195241292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,1,balanced,0.44175465901692706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,1,balanced,0.44699732462565106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,16,power_law_1.2,0.07031040191650391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,1,balanced,0.46932268142700195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,1,balanced,0.5090133349100748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,1,balanced,0.5477973222732544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,1,balanced,0.8231306870778402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,16,power_law_1.2,0.12062079906463623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,1,balanced,0.6911306381225586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,1,balanced,0.9956213633219401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,16,power_law_1.2,0.07328640222549439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,1,balanced,1.0482453505198162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,1,balanced,1.5609173774719238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,1,balanced,1.8237706820170085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,16,power_law_1.2,0.12940160036087037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,16,power_law_1.2,0.07292159795761108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,1,balanced,2.6094133059183755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,1,balanced,3.3604160944620767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,16,power_law_1.2,0.08247680068016053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,16,power_law_1.2,0.14180480241775512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,2,power_law_1.2,0.18529280424118041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,1,balanced,3.9158719380696616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,16,power_law_1.2,0.08584319949150085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,16,power_law_1.2,0.17130240201950073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,1,balanced,6.385429382324219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,16,power_law_1.2,0.09149439930915833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,16,power_law_1.2,0.20079998970031737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,16,power_law_1.2,0.10465279817581177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,16,power_law_1.2,0.12339199781417846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,2,power_law_1.2,0.18335360288619995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,1,balanced,11.931232452392578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,16,power_law_1.2,0.2589119911193848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,16,power_law_1.2,0.1472767949104309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,16,power_law_1.2,0.28679680824279785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,16,power_law_1.2,0.16519039869308472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,2,power_law_1.2,0.2307647943496704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,16,power_law_1.2,0.208076810836792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,16,power_law_1.2,0.39504640102386473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,16,power_law_1.2,0.23513600826263428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,16,power_law_1.2,0.5010560035705567
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,2,power_law_1.2,0.42293758392333985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,16,power_law_1.2,0.3346303939819336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,16,power_law_1.2,0.6715199947357178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,16,power_law_1.2,0.4419712066650391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,2,power_law_1.2,0.5747583866119385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,16,power_law_1.2,0.6094399929046631
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,16,power_law_1.2,0.9381504058837891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,16,power_law_1.2,0.7998208045959473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,16,power_law_1.2,1.1846015930175782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,2,power_law_1.2,1.1069888114929198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,16,power_law_1.2,1.0105600357055664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,4,power_law_1.01,0.04728319942951202
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,4,power_law_1.01,0.05486080050468445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,16,power_law_1.2,1.9638847351074218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,2,power_law_1.2,1.2452287673950195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,16,power_law_1.2,1.474988842010498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,4,power_law_1.01,0.04766719937324524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,4,power_law_1.01,0.05483520030975342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,4,power_law_1.01,0.060102397203445436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,2,power_law_1.2,1.2896703720092773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,4,power_law_1.01,0.06812800168991089
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,16,power_law_1.2,3.04520320892334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,16,power_law_1.2,3.33135986328125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,4,power_law_1.01,0.07962239980697632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,2,power_law_1.2,1.3400575637817382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,4,power_law_1.01,0.08381440043449402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,4,power_law_1.01,0.08269439935684204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,4,power_law_1.01,0.08606079816818238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,2,power_law_1.2,1.3758655548095704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,4,power_law_1.01,0.0907904028892517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,4,power_law_1.01,0.09585279822349549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,4,power_law_1.01,0.09938560128211975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,2,power_law_1.2,1.4260031700134277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,4,power_law_1.01,0.10750080347061157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,4,power_law_1.01,0.11775360107421876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,2,power_law_1.2,1.5110655784606934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,4,power_law_1.01,0.12391040325164795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,4,power_law_1.01,0.12431360483169555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,4,power_law_1.01,0.1422144055366516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,2,power_law_1.2,1.545132827758789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,4,power_law_1.01,0.1655743956565857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,16,2,power_law_1.01,0.03374080061912536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,16,2,power_law_1.01,0.033632001280784606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,16,2,power_law_1.01,0.03359360098838806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,2,power_law_1.01,0.054092800617218016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,4,power_law_1.01,0.20286719799041747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,16,2,power_law_1.01,0.03433600068092346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,16,2,power_law_1.01,0.03446399867534637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,2,power_law_1.01,0.05756800174713135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,16,2,power_law_1.01,0.034995201230049136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,16,2,power_law_1.01,0.03559040129184723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,16,2,power_law_1.01,0.03612160086631775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,4,power_law_1.01,0.23471360206604003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,16,2,power_law_1.01,0.03597440123558045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,2,power_law_1.01,0.0654911994934082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,2,power_law_1.2,1.5961664199829102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,16,2,power_law_1.01,0.037529599666595456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,16,2,power_law_1.01,0.03759360015392303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,16,2,power_law_1.01,0.03912320137023926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,2,power_law_1.01,0.09565439820289612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,16,2,power_law_1.01,0.03994239866733551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,4,power_law_1.01,0.30337278842926024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,16,2,power_law_1.01,0.04077439904212952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,16,2,power_law_1.01,0.0458624005317688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,2,power_law_1.01,0.12001279592514039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,16,2,power_law_1.01,0.04927360117435455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,16,2,power_law_1.01,0.054048001766204834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,16,2,power_law_1.01,0.06136959791183472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,2,power_law_1.01,0.17458560466766357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,16,2,power_law_1.01,0.06213120222091675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,4,power_law_1.01,0.37715198993682864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,16,2,power_law_1.01,0.08419200181961059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,16,2,power_law_1.01,0.09635199904441834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,2,power_law_1.01,0.18584320545196534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,16,2,power_law_1.01,0.11850880384445191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,16,2,power_law_1.01,0.14559999704360962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,2,power_law_1.2,1.6572799682617188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,2,power_law_1.01,0.19020160436630248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,16,2,power_law_1.01,0.19171839952468872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,4,power_law_1.01,0.5149248123168946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,16,2,power_law_1.01,0.24727680683135986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,2,power_law_1.01,0.18972159624099733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,16,2,power_law_1.01,0.4053311824798584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,16,2,power_law_1.01,0.4460480213165283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,2,power_law_1.01,0.19525760412216187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,4,power_law_1.01,0.633350419998169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,16,2,power_law_1.01,0.5580287933349609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,2,power_law_1.01,0.20539519786834717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,16,2,power_law_1.01,0.8413567543029785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,2,power_law_1.01,0.20400640964508057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,2,power_law_1.2,1.6566207885742188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,2,power_law_1.01,0.2122431993484497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,4,power_law_1.01,0.9418944358825684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,16,2,power_law_1.01,2.117670440673828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,2,power_law_1.01,0.21751039028167723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,2,power_law_1.01,0.2331007957458496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,4,power_law_1.01,1.2412351608276366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,2,power_law_1.01,0.24236159324645995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,2,power_law_1.01,0.24085121154785155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,2,power_law_1.2,1.7026687622070313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,2,power_law_1.01,0.2697279930114746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,4,power_law_1.01,1.5085824012756348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,2,power_law_1.01,0.2820096015930176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,2,power_law_1.01,0.33955841064453124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,2,power_law_1.01,0.38462080955505373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,2,power_law_1.2,1.8596672058105468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,2,power_law_1.01,0.4629824161529541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,4,power_law_1.01,2.292947196960449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,2,power_law_1.01,0.5719103813171387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,2,power_law_1.01,0.7847871780395508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,2,power_law_1.2,1.9293888092041016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,2,power_law_1.01,0.9753791809082031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,4,power_law_1.01,4.799795150756836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,2,power_law_1.01,1.3533632278442382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,2,power_law_1.2,2.245088005065918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,2,power_law_1.01,1.6366912841796875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,8,balanced,0.035877334574858345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,8,balanced,0.037658666570981346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,8,balanced,0.03990400085846583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,8,balanced,0.06418666740258534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,8,balanced,0.08483200271924336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,8,balanced,0.08261866867542267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,8,balanced,0.08459200461705525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,8,balanced,0.08426666259765625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,8,balanced,0.0828906645377477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,8,balanced,0.08363200227419536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,8,balanced,0.08412266770998637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,8,balanced,0.08345066507657369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,8,balanced,0.08569600184758504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,8,balanced,0.08955199519793193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,8,balanced,0.09066133697827657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,8,balanced,0.09090133508046468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,8,balanced,0.09179733196894328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,8,balanced,0.09654399752616882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,8,balanced,0.10729599992434184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,8,balanced,0.11690666278203328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,2,power_law_1.01,2.09607048034668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,8,balanced,0.12356799840927124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,8,balanced,0.15078933040301004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,8,balanced,0.15852266550064087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,8,balanced,0.20509332418441772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,8,balanced,0.25432000557581586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,8,balanced,0.3420373201370239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,8,balanced,0.4489599863688151
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,8,balanced,0.5410133202870687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,8,balanced,0.8329280217488607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,2,power_law_1.2,2.3452735900878907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,8,balanced,1.603530724843343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,2,power_law_1.01,3.0657663345336914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,2,power_law_1.2,2.8143104553222655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,2,power_law_1.01,6.220735931396485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,2,power_law_1.2,2.8234560012817385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,32,balanced,0.0561706672112147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,32,balanced,0.05421333511670431
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,32,balanced,0.052000001072883606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,32,balanced,0.05234666665395101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,32,balanced,0.061306665341059365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,32,balanced,0.08252266546090443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,32,balanced,0.08405866225560506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,32,balanced,0.08528000116348267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,32,balanced,0.08598933617273967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,32,balanced,0.08594666918118794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,32,balanced,0.08612799644470215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,32,balanced,0.08843732873598735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,32,balanced,0.08876799543698628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,32,balanced,0.08872000376383464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,32,balanced,0.0956106682618459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,32,balanced,0.09508267045021057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,32,balanced,0.09761599699656169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,32,balanced,0.1046506663163503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,32,balanced,0.10940800110499065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,32,balanced,0.12248532970746358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,32,balanced,0.13580800096193948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,32,balanced,0.15748799840609232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,2,power_law_1.2,3.491219329833984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,32,balanced,0.1829493244489034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,32,balanced,0.23034665981928507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,32,balanced,0.2781920035680135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,32,balanced,0.37829331556955975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,32,balanced,0.4682240088780721
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,32,balanced,0.5804586807886759
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,32,balanced,0.8414666652679443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,32,balanced,1.6348479588826497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,32,power_law_1.2,0.06986240148544312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,2,power_law_1.2,4.782374572753906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,32,power_law_1.2,0.03711360096931458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,32,power_law_1.2,0.03129599988460541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,32,power_law_1.2,0.02998400032520294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,32,power_law_1.2,0.03160319924354553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,32,power_law_1.2,0.033081600069999696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,32,power_law_1.2,0.03284479975700379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,32,power_law_1.2,0.032742398977279666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,2,power_law_1.2,0.041017600893974306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,32,power_law_1.2,0.033024001121521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,2,power_law_1.2,5.978047943115234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,2,power_law_1.2,0.03914879858493805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,32,power_law_1.2,0.033228799700737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,2,power_law_1.2,0.04254080057144165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,32,power_law_1.2,0.03692800104618073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,2,power_law_1.2,0.04911360144615173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,32,power_law_1.2,0.03847039937973022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,2,power_law_1.2,0.05314559936523437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,32,power_law_1.2,0.04165120124816894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,2,power_law_1.2,0.07529600262641907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,32,power_law_1.2,0.042483198642730716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,2,power_law_1.2,0.07628160119056701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,32,power_law_1.2,0.049446401000022885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,2,power_law_1.2,0.07743359804153442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,32,power_law_1.2,0.051737600564956666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,2,power_law_1.2,0.07858560085296631
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,32,power_law_1.2,0.0522816002368927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,2,power_law_1.2,6.593926239013672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,2,power_law_1.2,0.07875199913978577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,32,power_law_1.2,0.0667136013507843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,128,power_law_1.2,0.03248000144958496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,32,power_law_1.2,0.06537600159645081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,2,power_law_1.2,0.08288639783859253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,128,power_law_1.2,0.032601600885391234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,32,power_law_1.2,0.07148799896240235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,2,power_law_1.2,0.0814848005771637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,128,power_law_1.2,0.03225600123405457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,32,power_law_1.2,0.07546240091323853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,2,power_law_1.2,0.08217599987983704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,128,power_law_1.2,0.032671999931335446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,2,power_law_1.2,0.09066240191459655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,32,power_law_1.2,0.09544960260391236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,128,power_law_1.2,0.03308799862861633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,2,power_law_1.2,0.09471359848976135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,32,power_law_1.2,0.11342079639434814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,128,power_law_1.2,0.0342848002910614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,2,power_law_1.2,0.09772160053253173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,32,power_law_1.2,0.14777599573135375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,128,power_law_1.2,0.03406080007553101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,2,power_law_1.2,0.10408320426940917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,128,power_law_1.2,0.03535360097885132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,32,power_law_1.2,0.20060160160064697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,2,power_law_1.2,0.12090239524841309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,2,power_law_1.2,0.13717759847640992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,1,2,power_law_1.2,7.622803497314453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,128,power_law_1.2,0.034329599142074584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,32,power_law_1.2,0.27869439125061035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,2,power_law_1.2,0.15707520246505738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,128,power_law_1.2,0.03412480056285858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,32,power_law_1.2,0.3759488105773926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,2,power_law_1.2,0.18869119882583618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,128,power_law_1.2,0.03543039858341217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,2,power_law_1.2,0.24016640186309815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,1,32,power_law_1.2,0.5007423877716064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,128,power_law_1.2,0.0359935998916626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,2,power_law_1.2,0.307423996925354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,128,power_law_1.2,0.03586559891700745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,128,power_law_1.2,0.036320000886917114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,2,power_law_1.2,0.4008768081665039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,1,32,power_law_1.2,0.7696191787719726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,128,power_law_1.2,0.03928320109844208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,2,power_law_1.2,0.5367872238159179
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,128,power_law_1.2,0.039417600631713866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,128,power_law_1.2,0.04089600145816803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,2,power_law_1.2,0.7812863826751709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,128,power_law_1.2,0.044326400756835936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,1,32,power_law_1.2,1.4755647659301758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,128,power_law_1.2,0.047513601183891294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,2,power_law_1.2,0.9489664077758789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,128,power_law_1.2,0.05392000079154968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,128,power_law_1.2,0.06213759779930115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,1,2,power_law_1.2,12.324422454833984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,2,power_law_1.2,1.1991231918334961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,128,power_law_1.2,0.07676799893379212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,128,power_law_1.2,0.09712640047073365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,128,power_law_1.2,0.1326848030090332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,2,power_law_1.2,2.095648002624512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,128,power_law_1.2,0.1603775978088379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,128,power_law_1.2,0.2392512083053589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,128,power_law_1.2,0.32337279319763185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,2,power_law_1.2,4.37196159362793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,128,power_law_1.2,0.40006399154663086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,128,power_law_1.2,0.6335040092468261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,2,balanced,0.031445334355036415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,2,balanced,0.03375466664632162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,2,balanced,0.035429333647092186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,2,balanced,0.04199466605981191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,2,balanced,0.05188799897829691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,2,balanced,0.051957334081331887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,2,balanced,0.05392000079154968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,2,balanced,0.05649066468079885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,2,balanced,0.058143998185793556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,2,balanced,0.058042665322621666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,2,balanced,0.05593599875768026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,2,balanced,0.060080001751581825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,2,balanced,0.060421332716941833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,2,balanced,0.06211733321348826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,128,power_law_1.2,1.216543960571289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,2,balanced,0.07230933507283528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,2,balanced,0.07402666906515758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,2,balanced,0.08246399958928426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,2,balanced,0.09670933087666829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,2,balanced,0.11594133575757344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,2,balanced,0.13621333241462708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,2,balanced,0.1591253379980723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,2,balanced,0.22479466597239176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,2,balanced,0.280842661857605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,2,balanced,0.40405865510304767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,2,balanced,0.5144693454106649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,2,balanced,0.7578933238983154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,2,balanced,0.9902826944986979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,2,balanced,1.2322879632314045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,1,2,power_law_1.2,22.3444091796875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,2,balanced,1.9227306048075359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,2,balanced,3.7537225087483725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,2,power_law_1.2,0.03772799968719483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,2,power_law_1.2,0.040217599272727965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,2,power_law_1.2,0.04800640046596527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,2,power_law_1.2,0.06360960006713867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,2,power_law_1.2,0.06967039704322815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,2,power_law_1.2,0.10784640312194824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,2,power_law_1.2,0.11338239908218384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,2,power_law_1.2,0.11921919584274292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,2,power_law_1.2,0.1215999960899353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,1,power_law_1.01,0.03371520042419433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,2,power_law_1.2,0.12176640033721924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,1,power_law_1.01,0.04642559885978699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,2,power_law_1.2,0.12629120349884032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,1,power_law_1.01,0.06561279892921448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,2,power_law_1.2,0.13008639812469483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,1,power_law_1.01,0.09037439823150635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,2,power_law_1.2,0.13556480407714844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,1,power_law_1.2,0.039776000380516055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,1,power_law_1.01,0.11333119869232178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,2,power_law_1.2,0.13637759685516357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,1,power_law_1.01,0.17895679473876952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,4,1,power_law_1.01,0.04774399995803833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,2,power_law_1.2,0.14485759735107423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,1,power_law_1.01,0.18264960050582885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,1,power_law_1.2,0.04657920002937317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,4,1,power_law_1.01,0.047328001260757445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,2,power_law_1.2,0.15372159481048583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,1,power_law_1.01,0.1905727982521057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,4,1,power_law_1.01,0.05050879716873169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,1,power_law_1.01,0.19441280364990235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,2,power_law_1.2,0.163046395778656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,1,power_law_1.2,0.06461439728736877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,4,1,power_law_1.01,0.055776000022888184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,1,power_law_1.01,0.19838720560073853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,2,power_law_1.2,0.1841279983520508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,4,1,power_law_1.01,0.06275839805603027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,1,power_law_1.01,0.2045056104660034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,2,power_law_1.2,0.20849919319152832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,1,power_law_1.2,0.0876416027545929
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,4,1,power_law_1.01,0.07749119997024537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,1,power_law_1.01,0.2108544111251831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,2,power_law_1.2,0.2595839977264404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,4,1,power_law_1.01,0.0826367974281311
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,1,power_law_1.01,0.21932799816131593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,1,power_law_1.2,0.10464639663696289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,2,power_law_1.2,0.29234559535980226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,1,power_law_1.01,0.22266879081726074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,4,1,power_law_1.01,0.08090879917144775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,2,power_law_1.2,0.37601919174194337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,1,power_law_1.2,0.17863680124282838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,1,power_law_1.01,0.23121919631958007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,4,1,power_law_1.01,0.08316159844398499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,2,power_law_1.2,0.471673583984375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,1,power_law_1.01,0.2372607946395874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,4,1,power_law_1.01,0.08546559810638428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,1,power_law_1.2,0.1927232027053833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,4,1,power_law_1.01,0.08967040181159973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,1,power_law_1.01,0.25357439517974856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,2,power_law_1.2,0.6459968090057373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,4,1,power_law_1.01,0.09262719750404358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,1,power_law_1.01,0.2863744020462036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,1,power_law_1.2,0.19767680168151855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,4,1,power_law_1.01,0.09512959718704224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,2,power_law_1.2,0.8700799942016602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,1,power_law_1.01,0.3025856018066406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,4,1,power_law_1.01,0.10011520385742187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,1,power_law_1.2,0.20646400451660157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,1,power_law_1.01,0.36412160396575927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,4,1,power_law_1.01,0.10793600082397461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,2,power_law_1.2,1.2499391555786132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,1,power_law_1.01,0.41735038757324217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,1,power_law_1.2,0.20748159885406495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,4,1,power_law_1.01,0.1126911997795105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,4,1,power_law_1.01,0.12471679449081421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,1,power_law_1.01,0.5314303874969483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,1,power_law_1.2,0.21374080181121827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,2,power_law_1.2,1.4703935623168944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,4,1,power_law_1.01,0.14974080324172973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,1,power_law_1.01,0.6281983852386475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,4,1,power_law_1.01,0.17306239604949952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,1,power_law_1.2,0.218720006942749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,1,power_law_1.01,0.8423168182373046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,4,1,power_law_1.01,0.21841919422149658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,2,power_law_1.2,1.8808767318725585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,1,power_law_1.2,0.2270656108856201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,4,1,power_law_1.01,0.2624703884124756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,1,power_law_1.01,1.0446016311645507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,1,power_law_1.2,0.23797760009765626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,4,1,power_law_1.01,0.3596031904220581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,1,power_law_1.2,0.25016961097717283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,4,1,power_law_1.01,0.4601408004760742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,1,power_law_1.01,1.484768009185791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,2,power_law_1.2,2.9092607498168945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,1,power_law_1.2,0.25772159099578856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,4,1,power_law_1.01,0.6643648147583008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,1,power_law_1.01,1.8055807113647462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,1,power_law_1.2,0.26947200298309326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,4,1,power_law_1.01,0.8598848342895508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,1,power_law_1.2,0.3075455904006958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,4,1,power_law_1.01,1.2668736457824707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,1,power_law_1.01,2.219161605834961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,1,power_law_1.2,0.3430720090866089
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,2,power_law_1.2,5.899724960327148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,4,1,power_law_1.01,1.6596864700317382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,1,power_law_1.2,0.4247039794921875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,1,power_law_1.01,3.4305984497070314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,1,power_law_1.2,0.48279681205749514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,4,1,power_law_1.01,2.052729606628418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,1,power_law_1.2,0.6186048030853272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,1,power_law_1.2,0.7536320209503173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,4,1,power_law_1.01,3.209920120239258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,1,power_law_1.2,1.0247232437133789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,1,power_law_1.01,6.6593475341796875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,1,power_law_1.2,1.2644864082336427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,1,power_law_1.2,1.7843967437744142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,4,1,power_law_1.01,6.279283142089843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,1,power_law_1.2,2.304243278503418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,1,power_law_1.2,2.8343488693237306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,32,balanced,0.05719999969005585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,32,balanced,0.052570665876070656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,32,balanced,0.054229333996772766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,32,balanced,0.05495999753475189
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,32,balanced,0.05835733314355215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,32,balanced,0.0603413333495458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,32,balanced,0.06121600170930227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,32,balanced,0.06074133515357971
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,32,balanced,0.06113066772619883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,32,balanced,0.06272533535957336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,32,balanced,0.06074133515357971
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,32,balanced,0.0626453310251236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,32,balanced,0.06265600025653839
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,32,balanced,0.06411199768384297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,32,balanced,0.06741333504517873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,32,balanced,0.06849599877993266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,32,balanced,0.0703413337469101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,32,balanced,0.07652799785137177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,32,balanced,0.08102933565775554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,32,balanced,0.09103467067082723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,32,balanced,0.10507733623186748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,1,power_law_1.2,4.353145599365234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,32,balanced,0.13581333557764688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,32,balanced,0.1553813318411509
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,32,balanced,0.20075732469558716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,32,balanced,0.24310400088628134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,32,balanced,0.3286506732304891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,32,balanced,0.41371198495229083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,32,balanced,0.4941759904225667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,32,balanced,0.7587520281473795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,32,balanced,1.4837759335835774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,1,4,power_law_1.2,0.09963520169258118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,1,4,power_law_1.2,0.12026879787445069
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,1,power_law_1.2,8.369286346435548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,1,4,power_law_1.2,0.11004799604415894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,1,4,power_law_1.2,0.12510720491409302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,1,4,power_law_1.2,0.12629120349884032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,1,4,power_law_1.2,0.1292736053466797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,1,4,power_law_1.2,0.13428479433059692
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,1,4,power_law_1.2,0.13451520204544068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,1,4,power_law_1.2,0.13827199935913087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,1,4,power_law_1.2,0.1542207956314087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,1,4,power_law_1.2,0.1453376054763794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,1,4,power_law_1.2,0.16846079826354982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,64,balanced,0.0335359995563825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,64,balanced,0.035274667044480644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,1,4,power_law_1.2,0.17164160013198854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,64,balanced,0.029450667401154835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,64,balanced,0.029578665892283123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,64,balanced,0.03146666785081228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,64,balanced,0.03148266673088074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,64,balanced,0.03155199935038885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,1,4,power_law_1.2,0.16609920263290406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,64,balanced,0.03150933235883713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,64,balanced,0.03145066648721695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,64,balanced,0.03143999973932902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,64,balanced,0.031701333820819855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,64,balanced,0.03369066615899404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,1,4,power_law_1.2,0.21480960845947267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,64,balanced,0.03342399994532267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,64,balanced,0.03349866718053818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,64,balanced,0.03570133447647095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,64,balanced,0.03571200122435888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,64,balanced,0.037503999968369804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,1,4,power_law_1.2,0.2420351982116699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,64,balanced,0.037477334340413414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,64,balanced,0.03956799954175949
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,64,balanced,0.0407679999868075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,64,balanced,0.0396373321612676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,1,4,power_law_1.2,0.24136319160461425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,64,balanced,0.04398400088151296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,64,balanced,0.048581331968307495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,64,balanced,0.06037333110968272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,64,balanced,0.06823466718196869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,1,4,power_law_1.2,0.28465919494628905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,64,balanced,0.08425600330034892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,64,balanced,0.10194666186968486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,64,balanced,0.11745066444079082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,1,4,power_law_1.2,0.31535360813140867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,64,balanced,0.1509706676006317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,64,balanced,0.25749866167704266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,1,4,power_law_1.2,0.49073281288146975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,1,4,power_law_1.2,0.580460786819458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,1,4,power_law_1.2,0.7335487842559815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,1,4,power_law_1.2,1.060095977783203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,1,4,power_law_1.2,1.251692771911621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,1,4,power_law_1.2,2.3976383209228516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,1,4,power_law_1.2,3.1669567108154295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,1,4,power_law_1.2,4.931135940551758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,2,power_law_1.2,0.041305598616600034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,2,power_law_1.2,0.042099198698997496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,2,power_law_1.2,0.04816640019416809
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,1,4,power_law_1.2,5.460883331298828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,2,power_law_1.2,0.05571200251579285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,2,power_law_1.2,0.06255999803543091
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,1,balanced,0.060959999759991966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,1,balanced,0.06772266825040181
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,1,balanced,0.08146133522192638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,1,balanced,0.11308266719182332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,2,power_law_1.2,0.09043840169906617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,1,balanced,0.17138665914535522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,1,balanced,0.28697067499160767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,1,balanced,0.4040106534957886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,1,balanced,0.4026133219401042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,2,power_law_1.2,0.09528959989547729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,1,balanced,0.40456533432006836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,1,balanced,0.4057919979095459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,1,4,power_law_1.2,9.218029022216797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,1,balanced,0.4081759850184123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,1,balanced,0.41762133439381915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,2,power_law_1.2,0.09844480156898498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,1,balanced,0.420906662940979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,1,balanced,0.4296853144963582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,1,balanced,0.43770134449005127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,1,balanced,0.4420479933420817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,2,power_law_1.2,0.10113919973373413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,1,balanced,0.46320533752441406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,1,balanced,0.5089386701583862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,1,balanced,0.5384586652119955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,2,power_law_1.2,0.10204160213470459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,1,balanced,0.6154453357060751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,1,balanced,0.673904021581014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,1,balanced,0.8200533390045166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,2,power_law_1.2,0.10523519515991211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,1,balanced,1.00163729985555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,1,balanced,1.284991979598999
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,2,balanced,0.052671998739242554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,2,balanced,0.08539199829101562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,2,power_law_1.2,0.1083840012550354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,2,balanced,0.09934932986895244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,2,balanced,0.14685333768526712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,2,balanced,0.24850666522979736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,1,balanced,1.6853866577148438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,2,balanced,0.44199466705322266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,2,balanced,0.6428266763687134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,2,balanced,0.6425226529439291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,2,power_law_1.2,0.11184639930725097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,1,balanced,2.2482186953226724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,2,balanced,0.645034670829773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,2,balanced,0.6448906660079956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,2,balanced,0.6482826471328735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,2,balanced,0.6587146520614624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,2,power_law_1.2,0.11581439971923828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,2,balanced,0.6595093409220377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,1,balanced,3.041733423868815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,2,balanced,0.6671733061472574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,2,balanced,0.6806133588155111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,1,4,power_law_1.2,16.726451110839843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,2,balanced,0.6863413651784261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,2,power_law_1.2,0.12522879838943482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,2,balanced,0.6996106306711832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,1,balanced,3.8018134435017905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,2,balanced,0.7279573281606039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,2,balanced,0.75326935450236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,2,power_law_1.2,0.12621439695358277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,2,balanced,0.809221347173055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,2,balanced,0.8495360215504965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,2,balanced,0.9547519683837891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,2,power_law_1.2,0.1321023941040039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,2,balanced,1.0575093428293865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,1,balanced,5.5779469807942705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,2,balanced,1.2645440101623535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,2,power_law_1.2,0.14949120283126832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,2,balanced,1.5445547103881836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,2,balanced,2.0027146339416504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,2,power_law_1.2,0.16398719549179078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,2,balanced,2.561840057373047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,2,power_law_1.2,0.19326080083847047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,1,balanced,10.599514643351236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,2,balanced,3.217034657796224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,2,power_law_1.2,0.22442240715026857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,2,power_law_1.2,0.2668031930923462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,2,balanced,4.6626081466674805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,2,power_law_1.2,0.32930560111999513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,2,power_law_1.2,0.44776320457458496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,2,balanced,8.87784512837728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,2,power_law_1.2,0.5724671840667724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,2,power_law_1.2,0.7674431800842285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,2,power_law_1.2,0.988038444519043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,2688,1856,6,128,1,2,power_law_1.2,1.3291199684143067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,2688,1856,6,128,1,2,power_law_1.2,2.030240058898926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,2,power_law_1.2,0.1019711971282959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,2688,1856,6,128,1,2,power_law_1.2,3.8920833587646486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,2,power_law_1.2,0.08469120264053345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,16,power_law_1.01,0.036646398901939395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,2,power_law_1.2,0.09549440145492553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,16,power_law_1.01,0.034790399670600894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,4,4,power_law_1.01,0.05596799850463867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,4,4,power_law_1.01,0.06491519808769226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,4,4,power_law_1.01,0.048307201266288756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,4,4,power_law_1.01,0.05496320128440857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,4,4,power_law_1.01,0.055795198678970336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,16,power_law_1.01,0.031641599535942075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,4,4,power_law_1.01,0.056883198022842404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,2,power_law_1.2,0.14791040420532225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,4,4,power_law_1.01,0.05975679755210876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,4,4,power_law_1.01,0.063264000415802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,4,4,power_law_1.01,0.06227200031280518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,4,4,power_law_1.01,0.06321280002593994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,16,power_law_1.01,0.032153600454330446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,4,4,power_law_1.01,0.06533759832382202
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,4,4,power_law_1.01,0.06551039814949036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,4,4,power_law_1.01,0.06557440161705017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,4,4,power_law_1.01,0.0725055992603302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,4,4,power_law_1.01,0.0804095983505249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,2,power_law_1.2,0.19680000543594361
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,16,power_law_1.01,0.033932799100875856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,4,4,power_law_1.01,0.08948479890823365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,4,4,power_law_1.01,0.09381120204925537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,4,4,power_law_1.01,0.11846400499343872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,4,4,power_law_1.01,0.12539520263671874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,16,power_law_1.01,0.03635840117931366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,4,4,power_law_1.01,0.15333759784698486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,4,4,power_law_1.01,0.16844160556793214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,4,4,power_law_1.01,0.22003839015960694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,2,power_law_1.2,0.33390719890594484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,4,4,power_law_1.01,0.311244797706604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,16,power_law_1.01,0.035622400045394895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,4,4,power_law_1.01,0.3833343982696533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,4,4,power_law_1.01,0.548415994644165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,16,power_law_1.01,0.03612160086631775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,4,4,power_law_1.01,0.8581055641174317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,4,4,power_law_1.01,0.9891200065612793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,2,power_law_1.2,0.3695231914520264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,16,power_law_1.01,0.03640959858894348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,4,4,power_law_1.01,1.4974528312683106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,16,power_law_1.01,0.03653120100498199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,4,4,power_law_1.01,2.2367103576660154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,16,power_law_1.01,0.03823359906673431
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,2,power_law_1.2,0.3778111934661865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,16,4,balanced,0.0332640012105306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,16,4,balanced,0.031119999786218006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,16,4,balanced,0.03331200033426285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,16,4,balanced,0.03336533407370249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,16,4,balanced,0.032314665615558624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,16,4,balanced,0.0335359995563825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,16,4,balanced,0.033189333975315094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,16,power_law_1.01,0.04020479917526245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,16,4,balanced,0.03435199956099192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,16,4,balanced,0.03440000116825104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,16,4,balanced,0.03357866654793421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,16,4,balanced,0.033941333492596946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,16,4,balanced,0.0342399999499321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,16,4,balanced,0.03550933301448822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,16,4,balanced,0.035530666510264076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,16,4,balanced,0.04185600082079569
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,16,4,balanced,0.04354666670163473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,16,4,balanced,0.04176533222198486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,16,4,balanced,0.04790933430194855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,4,4,power_law_1.01,4.185760116577148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,16,4,balanced,0.04972266654173533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,16,4,balanced,0.06021333237489065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,16,4,balanced,0.064560001095136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,16,4,balanced,0.09657067060470581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,2,power_law_1.2,0.39140479564666747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,16,4,balanced,0.1179200013478597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,16,power_law_1.01,0.03999359905719757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,16,4,balanced,0.16292267044385275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,16,4,balanced,0.19154665867487589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,16,4,balanced,0.2642666697502136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,16,4,balanced,0.3309119939804077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,16,4,balanced,0.4046880006790161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,16,4,balanced,0.6090986728668213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,16,power_law_1.01,0.04344959855079651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,16,4,balanced,1.167237361272176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,2,power_law_1.2,0.399616003036499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,16,power_law_1.01,0.05082240104675293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,16,power_law_1.01,0.052275198698043826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,16,power_law_1.01,0.0515392005443573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,2,power_law_1.2,0.41651201248168945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,16,power_law_1.01,0.058220797777175905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,16,power_law_1.01,0.06417919993400574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,2,power_law_1.2,0.43858561515808103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,16,power_law_1.01,0.07555840015411378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,16,power_law_1.01,0.0816703975200653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,2,power_law_1.2,0.44423680305480956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,16,power_law_1.01,0.10063999891281128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,16,power_law_1.01,0.1227903962135315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,2,power_law_1.2,0.4508416175842285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,16,power_law_1.01,0.1679360032081604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,16,power_law_1.01,0.1927616000175476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,2,power_law_1.2,0.4772480010986328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,16,power_law_1.01,0.2436288118362427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,64,balanced,0.053898667295773826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,64,balanced,0.05048533280690511
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,64,balanced,0.050240000089009605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,64,balanced,0.07108266651630402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,2,power_law_1.2,0.48712959289550783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,64,balanced,0.07819200058778127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,64,balanced,0.07761066655317943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,64,balanced,0.08035199840863545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,16,power_law_1.01,0.36400001049041747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,64,balanced,0.08043733239173889
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,64,balanced,0.07969066500663757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,64,balanced,0.08082133531570435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,64,balanced,0.0819413314263026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,64,balanced,0.0832533339659373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,64,balanced,0.08431999882062276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,64,balanced,0.0846613347530365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,64,balanced,0.08913600444793701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,64,balanced,0.09051733215649922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,64,balanced,0.09179199735323589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,2,power_law_1.01,0.03284479975700379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,64,balanced,0.10949333508809407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,64,balanced,0.10705600182215373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,64,balanced,0.1301866670449575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,64,balanced,0.13776533802350363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,64,balanced,0.18083200852076212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,64,balanced,0.1731733282407125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,64,balanced,0.2279733419418335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,2,power_law_1.01,0.041280001401901245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,64,balanced,0.23287999629974365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,16,power_law_1.01,0.4028031826019287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,64,balanced,0.31520533561706543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,2,power_law_1.2,0.5033279895782471
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,64,balanced,0.3649119933446248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,2,power_law_1.01,0.042105600237846375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,2,64,balanced,0.43565332889556885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,2,64,balanced,0.6195306777954102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,2,power_law_1.01,0.05601279735565186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,2,64,balanced,1.3357280095418294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,2,power_law_1.01,0.07288320064544677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,16,power_law_1.01,0.5386047840118409
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,2,power_law_1.2,0.5566271781921387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,2,power_law_1.01,0.10430719852447509
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,2,power_law_1.01,0.10885119438171387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,2,power_law_1.01,0.11019519567489625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,2,power_law_1.2,0.6106431961059571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,2,power_law_1.01,0.11308799982070923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,2,power_law_1.01,0.11688319444656373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,16,power_law_1.01,1.3123711585998534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,2,power_law_1.01,0.1179967999458313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,2,power_law_1.01,0.12035200595855713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,2,power_law_1.2,0.7179200172424316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,2,power_law_1.01,0.12291840314865113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,2,power_law_1.01,0.12878079414367677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,2,power_law_1.01,0.1385472059249878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,2,power_law_1.2,0.7874623775482178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,2,power_law_1.01,0.14391679763793946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,2,power_law_1.01,0.14343039989471434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,2,power_law_1.01,0.15141760110855101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,2,power_law_1.2,0.9882816314697266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,2,power_law_1.01,0.1637760043144226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,2,power_law_1.01,0.1899839997291565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,2,power_law_1.01,0.20034561157226563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,2,power_law_1.01,0.24359679222106934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,2,power_law_1.2,1.0986432075500487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,2,power_law_1.01,0.2765887975692749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,2,power_law_1.01,0.3536384105682373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,2,power_law_1.01,0.08940160274505615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,2,power_law_1.01,0.42289280891418457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,2,power_law_1.2,1.422374439239502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,2,power_law_1.01,0.1060096025466919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,2,power_law_1.01,0.603878402709961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,2,power_law_1.01,0.09950079917907714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,64,balanced,0.03395200024048487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,2,power_law_1.01,0.1552448034286499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,2,power_law_1.01,0.7162496089935303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,64,balanced,0.03398933261632919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,64,balanced,0.03297599901755651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,64,balanced,0.03479466587305069
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,64,balanced,0.0353973334034284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,64,balanced,0.03572800010442734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,64,balanced,0.03761066744724909
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,64,balanced,0.03641066700220108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,2,power_law_1.01,0.19964799880981446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,64,balanced,0.03700799991687139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,64,balanced,0.03756266583998998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,2,power_law_1.2,1.7362367630004882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,64,balanced,0.03812800099452337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,64,balanced,0.03709866603215536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,64,balanced,0.0386613334218661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,1,2,power_law_1.01,0.9160191535949707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,64,balanced,0.039061332742373146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,64,balanced,0.04274666806062063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,64,balanced,0.04337066908677419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,2,power_law_1.01,0.28903679847717284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,64,balanced,0.044394666949907936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,64,balanced,0.044591998060544334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,64,balanced,0.04658666749795278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,64,balanced,0.0529013325770696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,64,balanced,0.06323199967543285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,64,balanced,0.08488532900810242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,64,balanced,0.093941330909729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,2,power_law_1.01,0.3228224039077759
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,64,balanced,0.12325333555539449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,64,balanced,0.14812800288200378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,64,balanced,0.19696533679962158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,64,balanced,0.25094934304555255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,1,2,power_law_1.01,1.3348671913146972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,64,balanced,0.3052213390668233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,2,power_law_1.01,0.3295936107635498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,64,balanced,0.4359840154647827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,64,balanced,0.8231093088785807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,2,power_law_1.01,0.3349184036254883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,2,power_law_1.2,2.3601472854614256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,2,power_law_1.01,0.3457344055175781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,2,power_law_1.01,0.351148796081543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,1,2,power_law_1.01,2.5956607818603517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,2,power_law_1.01,0.35939199924468995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,2,power_law_1.01,0.37118079662323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,2,power_law_1.01,0.3745919942855835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,2,power_law_1.2,3.0592575073242188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,2,power_law_1.01,0.39052159786224366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,2,power_law_1.01,0.39616639614105226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,2,power_law_1.01,0.41560959815979004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,2,power_law_1.01,0.4598656177520752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,2,power_law_1.01,0.4946432113647461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,2,power_law_1.01,0.5896192073822022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,2,2,power_law_1.2,3.685951995849609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,2,power_law_1.01,0.6394048213958741
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,2,power_law_1.01,0.7878464221954345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,2,power_law_1.01,0.9740799903869629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,2,power_law_1.01,1.2475456237792968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,4,8,power_law_1.2,0.06535680294036865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,1,power_law_1.2,0.06707199811935424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,2,power_law_1.01,1.567404842376709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,4,8,power_law_1.2,0.055308800935745236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,4,8,power_law_1.2,0.05381119847297668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,4,8,power_law_1.2,0.05371519923210144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,2,2,power_law_1.2,5.843494415283203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,4,8,power_law_1.2,0.053395199775695804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,4,8,power_law_1.2,0.05419520139694214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,4,8,power_law_1.2,0.05692160129547119
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,4,8,power_law_1.2,0.058880001306533813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,1,power_law_1.2,0.07500159740447998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,2,power_law_1.01,2.0262271881103517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,4,8,power_law_1.2,0.06172159910202026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,4,8,power_law_1.2,0.06333439946174621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,4,8,power_law_1.2,0.06223359704017639
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,4,8,power_law_1.2,0.06796159744262695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,4,8,power_law_1.2,0.06836479902267456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,4,8,power_law_1.2,0.07036160230636597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,4,8,power_law_1.2,0.08372480273246766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,1,power_law_1.2,0.09006720185279846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,4,8,power_law_1.2,0.08282880187034607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,4,8,power_law_1.2,0.09566079974174499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,2,power_law_1.01,2.7590784072875976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,4,8,power_law_1.2,0.1250496029853821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,4,8,power_law_1.2,0.13914239406585693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,1,power_law_1.2,0.11068799495697021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,4,8,power_law_1.2,0.17934720516204833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,4,8,power_law_1.2,0.2014336109161377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,4,8,power_law_1.2,0.29348480701446533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,4,8,power_law_1.2,0.3904063940048218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,1,power_law_1.2,0.13657599687576294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,4,8,power_law_1.2,0.5711872100830078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,2,power_law_1.01,3.251814270019531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,4,8,power_law_1.2,0.7099520206451416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,1,power_law_1.2,0.21025280952453612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,4,8,power_law_1.2,1.0978879928588867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,4,8,power_law_1.2,1.4478015899658203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,1,power_law_1.2,0.23271679878234863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,4,8,power_law_1.2,1.6897855758666993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,1,power_law_1.2,0.2407360076904297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,2,power_law_1.01,4.926483154296875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,4,8,power_law_1.2,2.3192319869995117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,1,power_law_1.2,0.2445120096206665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,1,power_law_1.2,0.2515199899673462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,4,8,power_law_1.2,4.862227249145508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,2,2,power_law_1.2,11.019699096679688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,1,power_law_1.2,0.26314239501953124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,1,power_law_1.2,0.2730112075805664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,2,power_law_1.01,9.092294311523437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,1,power_law_1.2,0.2791935920715332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,1,power_law_1.2,0.28626561164855957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,1,power_law_1.2,0.30125439167022705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,1,power_law_1.2,0.3065216064453125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,1,power_law_1.2,0.32554240226745607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,1,power_law_1.2,0.37920000553131106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,1,power_law_1.2,0.4198272228240967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,1,power_law_1.2,0.5167168140411377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,1,power_law_1.2,0.5875199794769287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,1,power_law_1.2,0.7684607982635498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,1,power_law_1.2,0.9309247970581055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,1,power_law_1.2,1.2704128265380858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,16,1,power_law_1.01,0.035622400045394895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,16,1,power_law_1.01,0.035519999265670774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,32,power_law_1.01,0.03438720107078552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,16,1,power_law_1.01,0.03606399893760681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,16,1,power_law_1.01,0.03800959885120392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,16,1,power_law_1.01,0.038899201154708865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,32,power_law_1.01,0.03495680093765259
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,16,1,power_law_1.01,0.03953279852867127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,1,power_law_1.2,1.7572927474975586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,16,1,power_law_1.01,0.03973119854927063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,16,1,power_law_1.01,0.04013440012931824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,32,power_law_1.01,0.028620800375938414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,16,1,power_law_1.01,0.04097920060157776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,16,1,power_law_1.01,0.04230400025844574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,16,1,power_law_1.01,0.04307200014591217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,32,power_law_1.01,0.02858240008354187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,16,1,power_law_1.01,0.045459198951721194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,16,1,power_law_1.01,0.048390400409698484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,16,1,power_law_1.01,0.05088000297546387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,32,power_law_1.01,0.028358399868011475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,16,1,power_law_1.01,0.056505602598190305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,1,power_law_1.2,2.4295616149902344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,16,1,power_law_1.01,0.05726079940795899
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,8,power_law_1.2,0.09513599872589111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,16,1,power_law_1.01,0.06593279838562012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,32,power_law_1.01,0.029625600576400755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,16,1,power_law_1.01,0.073990398645401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,16,1,power_law_1.01,0.08197759985923767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,32,power_law_1.01,0.030112001299858093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,16,1,power_law_1.01,0.10451200008392333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,16,1,power_law_1.01,0.11908479928970336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,16,1,power_law_1.01,0.15058560371398927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,32,power_law_1.01,0.030399999022483824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,8,power_law_1.2,0.16792320013046264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,16,1,power_law_1.01,0.18021759986877442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,16,1,power_law_1.01,0.2539072036743164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,32,power_law_1.01,0.029971200227737426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,1,power_law_1.2,2.9948095321655273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,16,1,power_law_1.01,0.32420480251312256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,32,power_law_1.01,0.029836800694465638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,16,1,power_law_1.01,0.4631552219390869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,8,power_law_1.2,0.14867199659347535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,16,1,power_law_1.01,0.6048831939697266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,32,power_law_1.01,0.029977598786354066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,16,1,power_law_1.01,0.7431359767913819
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,32,power_law_1.01,0.030924800038337707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,8,power_law_1.2,0.09144960045814514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,32,power_law_1.01,0.032204800844192506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,16,1,power_law_1.01,1.14715518951416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,1,power_law_1.2,3.651987075805664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,32,power_law_1.01,0.03221119940280914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,32,power_law_1.01,0.03597440123558045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,16,1,power_law_1.01,2.221990394592285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,8,power_law_1.2,0.12767360210418702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,32,power_law_1.01,0.03635199964046478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,32,power_law_1.01,0.03715839982032776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,8,power_law_1.2,0.16370559930801393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,1,power_law_1.2,5.612153625488281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,32,power_law_1.01,0.04012160003185272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,32,power_law_1.01,0.041331198811531064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,8,power_law_1.2,0.18636800050735475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,32,power_law_1.01,0.04810880124568939
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,32,power_law_1.01,0.049420800805091855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,8,power_law_1.2,0.24191360473632811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,32,power_law_1.01,0.060313600301742556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,32,power_law_1.01,0.0657472014427185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,8,power_law_1.2,0.24968318939208983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,32,power_law_1.01,0.07936000227928161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,32,power_law_1.01,0.09886720180511474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,1,power_law_1.2,10.915666961669922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,8,power_law_1.2,0.2546367883682251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,32,power_law_1.01,0.1331071972846985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,8,power_law_1.2,0.2643712043762207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,32,power_law_1.01,0.15792640447616577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,4,32,power_law_1.01,0.18008320331573485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,8,power_law_1.2,0.2831487894058228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,4,32,power_law_1.01,0.25100159645080566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,8,power_law_1.2,0.2774912118911743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,8,4,balanced,0.038202665746212006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,8,4,balanced,0.03802666564782461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,8,4,balanced,0.03612799942493439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,8,4,balanced,0.035962666074434914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,8,4,balanced,0.03799466788768768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,8,4,balanced,0.037861332297325134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,8,4,balanced,0.03782399992148081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,8,4,balanced,0.03666666646798452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,8,4,balanced,0.037647999823093414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,8,4,balanced,0.03766933331886927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,8,4,balanced,0.037962667644023895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,8,4,balanced,0.03984533250331879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,8,4,balanced,0.03809600075085958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,8,4,balanced,0.03978666663169861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,8,4,balanced,0.04419200122356415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,8,4,balanced,0.04654400050640106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,8,4,balanced,0.04613866905371348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,8,4,balanced,0.05077333251635233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,8,4,balanced,0.05329066514968872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,8,4,balanced,0.06502933303515117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,8,power_law_1.2,0.2878528118133545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,8,4,balanced,0.06634666522343953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,8,4,balanced,0.08907199899355571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,8,4,balanced,0.1079360047976176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,8,4,balanced,0.136543999115626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,8,4,balanced,0.15530666708946228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,4,32,power_law_1.01,0.5057216167449952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,8,4,balanced,0.2143626610438029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,8,4,balanced,0.2669919927914937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,8,4,balanced,0.31804267565409344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,8,4,balanced,0.47674131393432617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,8,4,balanced,0.8871200084686279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,8,power_law_1.2,0.3051455974578857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,8,power_law_1.2,0.31473278999328613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,8,power_law_1.2,0.33634560108184813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,8,power_law_1.2,0.38179199695587157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,8,power_law_1.2,0.4174464225769043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,8,power_law_1.2,0.5080128192901612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,8,power_law_1.2,0.5238272190093994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,16,power_law_1.2,0.12064000368118286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,8,power_law_1.2,0.628652811050415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,4,2,balanced,0.035887998839219414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,8,power_law_1.2,0.7465663909912109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,4,2,balanced,0.06010133524735769
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,4,2,balanced,0.07654933134714763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,4,2,balanced,0.07867200175921123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,4,2,balanced,0.07989866534868877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,4,2,balanced,0.07868800063927968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,4,2,balanced,0.08077333370844524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,4,2,balanced,0.07946133116881053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,4,2,balanced,0.0811359981695811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,4,2,balanced,0.08098133405049641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,4,2,balanced,0.08271466692288716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,4,2,balanced,0.08552533388137817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,16,power_law_1.2,0.10399359464645386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,4,2,balanced,0.08519466718037923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,4,2,balanced,0.09273599584897359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,16,power_law_1.01,0.043424001336097716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,4,2,balanced,0.09886933366457622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,4,2,balanced,0.0999840001265208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,4,2,balanced,0.10834133625030518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,4,2,balanced,0.12158399820327759
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,4,2,balanced,0.1322879989941915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,4,2,balanced,0.15800533692042032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,4,2,balanced,0.1845973332722982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,4,2,balanced,0.24041066567103067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,16,power_law_1.01,0.05723519921302796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,4,2,balanced,0.2972639997800191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,4,2,balanced,0.41204265753428143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,4,2,balanced,0.5264053344726562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,8,power_law_1.2,0.9149696350097656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,4,2,balanced,0.754319985707601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,16,power_law_1.01,0.034745600819587705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,4,2,balanced,0.9789546330769857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,4,2,balanced,1.2004480361938477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,16,power_law_1.01,0.035820800065994265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,4,2,balanced,1.8616426785786946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,16,power_law_1.01,0.03622399866580963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,4,2,balanced,3.6288585662841797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,8,power_law_1.2,1.0489215850830078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,16,power_law_1.01,0.03873920142650604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,16,power_law_1.2,0.07338879704475403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,16,power_law_1.01,0.03850240111351013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,16,power_law_1.01,0.039110401272773744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,16,power_law_1.01,0.03914879858493805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,8,power_law_1.2,1.3576128005981445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,16,power_law_1.01,0.039417600631713866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,16,power_law_1.01,0.04020479917526245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,16,power_law_1.2,0.08784639835357666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,16,power_law_1.01,0.041382399201393125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,16,power_law_1.01,0.04094719886779785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,8,power_law_1.2,1.6985088348388673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,16,power_law_1.01,0.04154880046844482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,16,power_law_1.01,0.04692479968070984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,16,power_law_1.01,0.048691201210021975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,16,power_law_1.2,0.09996799826622009
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,16,power_law_1.01,0.05159040093421936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,8,power_law_1.2,2.050873565673828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,16,power_law_1.01,0.05742080211639404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,16,power_law_1.01,0.0603007972240448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,16,power_law_1.01,0.0716480016708374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,16,power_law_1.01,0.07976959943771363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,16,power_law_1.2,0.13026560544967652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,16,power_law_1.01,0.09288319945335388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,8,power_law_1.2,3.1606847763061525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,16,power_law_1.01,0.1093824028968811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,16,power_law_1.01,0.13824000358581542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,16,power_law_1.01,0.16420480012893676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,16,power_law_1.2,0.1402624011039734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,32,power_law_1.2,0.039136001467704774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,16,power_law_1.01,0.24956159591674804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,32,power_law_1.2,0.03432320058345795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,32,power_law_1.2,0.03439359962940216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,16,power_law_1.01,0.33495678901672366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,32,power_law_1.2,0.0322816014289856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,8,power_law_1.2,6.802828979492188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,16,power_law_1.2,0.14256639480590821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,2688,1856,6,128,1,16,power_law_1.01,0.34514560699462893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,32,power_law_1.2,0.033983999490737916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,32,power_law_1.2,0.034483200311660765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,2688,1856,6,128,1,16,power_law_1.01,0.5843520164489746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,32,power_law_1.2,0.03429119884967804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,16,power_law_1.2,0.14172799587249757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,32,power_law_1.2,0.03441919982433319
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,32,power_law_1.2,0.034764799475669864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,2688,1856,6,128,1,16,power_law_1.01,1.006611156463623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,32,power_law_1.2,0.03542400002479553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,32,power_law_1.2,0.03627519905567169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,16,power_law_1.2,0.14517760276794434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,32,power_law_1.2,0.037363201379776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,32,power_law_1.2,0.03790079951286316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,32,power_law_1.2,0.0381056010723114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,32,power_law_1.2,0.04192639887332916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,16,power_law_1.2,0.14778239727020265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,32,power_law_1.2,0.04276480078697205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,32,power_law_1.2,0.043724799156188966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,32,power_law_1.2,0.04757120013237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,32,power_law_1.2,0.05336959958076477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,16,power_law_1.2,0.15427839756011963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,32,power_law_1.2,0.057868802547454835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,32,power_law_1.2,0.06537600159645081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,32,power_law_1.2,0.07964159846305847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,16,balanced,0.05011733373006185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,16,balanced,0.04975466430187225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,16,balanced,0.04977599779764811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,16,power_law_1.2,0.15495680570602416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,16,balanced,0.05194666484991709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,16,balanced,0.05401599903901418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,16,balanced,0.06413866579532623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,32,power_law_1.2,0.09000319838523865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,16,balanced,0.06632533172766368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,16,balanced,0.07038400073846181
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,16,balanced,0.07006933291753133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,16,balanced,0.07034666836261749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,16,balanced,0.07055466870466869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,16,balanced,0.07275733351707458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,16,balanced,0.07238933444023132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,16,balanced,0.07495466868082683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,16,balanced,0.08255999783674876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,16,balanced,0.08227199812730153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,16,balanced,0.0881226658821106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,16,balanced,0.09485333164532979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,32,power_law_1.2,0.12187520265579224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,16,balanced,0.10242666800816853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,16,balanced,0.11778133114178975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,16,balanced,0.13581333557764688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,16,balanced,0.16106667121251425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,16,balanced,0.18498667081197104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,16,balanced,0.23985600471496582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,16,balanced,0.2937813401222229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,16,balanced,0.40677865346272785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,32,power_law_1.2,0.18943359851837158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,16,balanced,0.5129599968592325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,16,balanced,0.6331146558125814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,16,balanced,0.9586453437805176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,16,power_law_1.2,0.16293760538101196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,16,balanced,1.8618133862813313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,32,power_law_1.2,0.26039040088653564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,4,16,balanced,0.025279998779296875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,4,16,balanced,0.025285333395004272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,4,16,balanced,0.0244159996509552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,4,16,balanced,0.02388266722361247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,4,16,balanced,0.025413334369659424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,4,16,balanced,0.025216000775496166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,4,16,balanced,0.025472000241279602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,4,16,balanced,0.02923733244339625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,4,16,balanced,0.02945599953333537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,4,16,balanced,0.027290667096773785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,4,16,balanced,0.027514666318893433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,4,16,balanced,0.029733332494894665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,4,16,balanced,0.02920000006755193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,4,16,balanced,0.03146666785081228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,4,16,balanced,0.03612799942493439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,32,power_law_1.2,0.363590407371521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,4,16,balanced,0.033488000432650246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,4,16,balanced,0.035642666121323906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,4,16,balanced,0.037690666814645134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,16,balanced,0.03770133356253306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,16,balanced,0.040378667414188385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,16,balanced,0.04171733558177948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,16,balanced,0.04980266590913137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,16,balanced,0.05023466547330221
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,16,balanced,0.0622026671965917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,16,balanced,0.07893866797288258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,16,balanced,0.1020853320757548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,16,balanced,0.12398399909337361
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,16,balanced,0.14477866888046265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,16,balanced,0.197818656762441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,16,balanced,0.35680532455444336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,16,power_law_1.2,0.17314560413360597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,32,power_law_1.2,0.4287680149078369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,32,power_law_1.2,0.8413887977600097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,16,power_law_1.2,0.1698240041732788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,32,power_law_1.2,1.571878433227539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,16,power_law_1.2,0.18309119939804078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,16,power_law_1.2,0.21505279541015626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,16,power_law_1.2,0.2195647954940796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,128,balanced,0.05827199916044871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,128,balanced,0.05708266794681549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,128,balanced,0.05969599882761637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,128,balanced,0.057536001006762184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,128,balanced,0.05840533475081126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,128,balanced,0.06608533362547557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,128,balanced,0.06422399977842967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,128,balanced,0.06634133557478587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,128,balanced,0.06702400247255962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,128,balanced,0.06765333314736684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,128,balanced,0.0680320014556249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,128,balanced,0.0666186660528183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,128,balanced,0.06852266689141591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,128,balanced,0.07242666681607564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,128,balanced,0.07250666618347168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,128,balanced,0.07195200026035309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,128,balanced,0.07693333427111308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,128,balanced,0.08470400174458821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,128,balanced,0.08872000376383464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,128,balanced,0.10300266742706299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,128,balanced,0.11725333333015442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,128,balanced,0.14362133542696634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,128,balanced,0.16614933808644614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,128,balanced,0.2144533395767212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,128,balanced,0.25963733593622845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,16,power_law_1.2,0.2712640047073364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,128,balanced,0.36082132657368976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,128,balanced,0.4527680079142253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,128,balanced,0.55240531762441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,4,1,power_law_1.2,0.04103679955005646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,128,balanced,0.8386560281117758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,4,1,power_law_1.2,0.04419200122356415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,4,1,power_law_1.2,0.05256320238113403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,128,balanced,1.6457974116007488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,4,1,power_law_1.2,0.06718080043792725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,4,1,power_law_1.2,0.0674239993095398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,4,1,power_law_1.2,0.06796159744262695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,4,1,power_law_1.2,0.06851840019226074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,4,1,power_law_1.2,0.06933119893074036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,4,1,power_law_1.2,0.07023360133171082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,4,1,power_law_1.2,0.07023360133171082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,4,1,power_law_1.2,0.07248640060424805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,16,power_law_1.2,0.32646400928497316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,4,1,power_law_1.2,0.074099200963974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,4,1,power_law_1.2,0.07697920203208923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,4,1,power_law_1.2,0.084307199716568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,4,1,power_law_1.2,0.09386240243911743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,4,1,power_law_1.2,0.09587200284004212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,4,1,power_law_1.2,0.09736319780349731
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,4,1,power_law_1.2,0.11344000101089477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,4,1,power_law_1.2,0.12832000255584716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,4,1,power_law_1.2,0.16012799739837646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,4,1,power_law_1.2,0.19000320434570311
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,16,power_law_1.01,0.03370879888534546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,4,1,power_law_1.2,0.2501312017440796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,16,power_law_1.2,0.41973118782043456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,4,1,power_law_1.2,0.29959681034088137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,16,power_law_1.01,0.030828800797462464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,4,1,power_law_1.2,0.4088831901550293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,4,1,power_law_1.2,0.5086463928222656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,16,power_law_1.01,0.030393600463867188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,4,1,power_law_1.2,0.7193088054656982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,16,power_law_1.01,0.030131199955940248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,4,1,power_law_1.2,0.9338432312011719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,16,power_law_1.01,0.0318015992641449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,4,1,power_law_1.2,1.1402112007141114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,16,power_law_1.2,0.42252798080444337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,16,power_law_1.01,0.033504000306129454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,16,power_law_1.01,0.033555200695991515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,4,1,power_law_1.2,1.7802623748779296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,16,power_law_1.01,0.03408640027046204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,16,power_law_1.01,0.0345984011888504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,4,1,power_law_1.2,3.4720958709716796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,16,power_law_1.01,0.03497599959373474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,16,power_law_1.2,0.5223167896270752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,16,power_law_1.01,0.03555200099945068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,16,power_law_1.01,0.03591679930686951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,16,power_law_1.01,0.03665919899940491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,16,power_law_1.01,0.03818880021572113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,16,power_law_1.01,0.04296959936618805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,32,power_law_1.2,0.038924801349639895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,16,power_law_1.01,0.04521600008010864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,16,power_law_1.2,0.7146687984466553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,32,power_law_1.2,0.03466239869594574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,16,power_law_1.01,0.04673919975757599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,32,power_law_1.2,0.03351680040359497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,16,power_law_1.01,0.05058559775352478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,32,power_law_1.2,0.03358719944953918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,16,power_law_1.01,0.05942400097846985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,32,power_law_1.2,0.035123199224472046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,32,power_law_1.2,0.035283198952674864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,16,power_law_1.01,0.0662335991859436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,32,power_law_1.2,0.035519999265670774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,16,power_law_1.01,0.07583360075950622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,16,power_law_1.2,0.9355711936950684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,32,power_law_1.2,0.03587839901447296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,16,power_law_1.01,0.10087039470672607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,32,power_law_1.2,0.035872000455856326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,16,power_law_1.01,0.14099199771881105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,32,power_law_1.2,0.03625600039958954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,16,power_law_1.01,0.18900480270385742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,32,power_law_1.2,0.03687680065631867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,32,power_law_1.2,0.03744640052318573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,16,power_law_1.01,0.2413952112197876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,32,power_law_1.2,0.03940480053424835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,16,power_law_1.01,0.3141439914703369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,32,power_law_1.2,0.04137600064277649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,16,power_law_1.2,1.1314175605773926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,32,power_law_1.2,0.04583039879798889
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,16,power_law_1.01,0.457747220993042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,32,power_law_1.2,0.046854400634765626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,16,power_law_1.2,0.1441151976585388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,32,power_law_1.2,0.04928640127182007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,32,power_law_1.2,0.05614079833030701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,16,power_law_1.01,0.5825727939605713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,16,power_law_1.2,0.2153791904449463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,32,power_law_1.2,0.0639360010623932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,16,power_law_1.2,0.09214079976081849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,32,power_law_1.2,0.07368959784507752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,16,power_law_1.01,0.9076352119445801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,32,power_law_1.2,0.08265600204467774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,16,power_law_1.2,0.10217599868774414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,1,16,power_law_1.2,1.367091178894043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,32,power_law_1.2,0.09857919812202454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,16,power_law_1.2,0.12335360050201416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,32,power_law_1.2,0.13371520042419432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,32,power_law_1.2,0.17304320335388185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,16,power_law_1.2,0.14424320459365844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,16,power_law_1.01,1.5902079582214355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,32,power_law_1.2,0.24839038848876954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,16,power_law_1.2,0.20977280139923096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,32,power_law_1.2,0.3518656015396118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,16,power_law_1.2,0.20198400020599366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,32,power_law_1.2,0.6411007881164551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,16,power_law_1.2,0.2312704086303711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,1,16,power_law_1.2,2.231494331359863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,16,power_law_1.2,0.2090751886367798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,32,power_law_1.2,0.6286143779754638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,16,power_law_1.2,0.2165760040283203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,32,power_law_1.2,0.8852288246154785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,16,power_law_1.2,0.21600639820098877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,16,power_law_1.2,0.22537600994110107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,16,power_law_1.2,0.22576000690460205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,32,power_law_1.2,2.007276725769043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,16,power_law_1.2,0.24291839599609374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,16,power_law_1.2,0.26007680892944335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,16,power_law_1.2,0.2708224058151245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,16,power_law_1.2,0.3175935983657837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,1,16,power_law_1.2,4.841644668579102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,4,1,power_law_1.2,0.030681601166725157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,16,power_law_1.2,0.3219583988189697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,4,1,power_law_1.2,0.03548159897327423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,4,1,power_law_1.2,0.038124799728393555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,16,power_law_1.2,0.3826368093490601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,4,1,power_law_1.2,0.04257920086383819
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,4,1,power_law_1.2,0.047200000286102294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,16,power_law_1.2,0.38156158924102784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,4,1,power_law_1.2,0.0729856014251709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,4,1,power_law_1.2,0.07742080092430115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,16,power_law_1.2,0.45189762115478516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,4,1,power_law_1.2,0.08156800270080566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,4,1,power_law_1.2,0.0828224003314972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,16,power_law_1.2,0.5567359924316406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,4,1,power_law_1.2,0.08540160059928895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,4,1,power_law_1.2,0.09022719860076904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,16,power_law_1.2,0.7138815879821777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,4,1,power_law_1.2,0.09144960045814514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,4,1,power_law_1.2,0.09402239918708802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,16,power_law_1.2,0.8701184272766114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,4,1,power_law_1.2,0.09590399861335755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,4,1,power_law_1.2,0.10410239696502685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,4,1,power_law_1.2,0.10830080509185791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,16,power_law_1.2,1.2017024040222168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,4,1,power_law_1.2,0.11935360431671142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,32,power_law_1.2,0.07519999742507935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,4,1,power_law_1.2,0.13426560163497925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,1,power_law_1.2,0.1475648045539856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,16,power_law_1.2,1.3573375701904298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,32,power_law_1.2,0.07989760041236878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,1,power_law_1.2,0.17880959510803224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,1,power_law_1.2,0.20016639232635497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,1,power_law_1.2,0.2624768018722534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,32,power_law_1.2,0.06858879923820496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,16,power_law_1.2,1.8320640563964843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,1,power_law_1.2,0.32232959270477296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,1,power_law_1.2,0.44666237831115724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,32,power_law_1.2,0.05978239774703979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,1,power_law_1.2,0.5613311767578125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,16,power_law_1.2,2.690278434753418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,32,power_law_1.2,0.06232320070266724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,1,power_law_1.01,0.05345919728279114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,1,power_law_1.2,0.7972415924072266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,32,power_law_1.2,0.06686080098152161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,1,power_law_1.2,1.0201919555664063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,32,power_law_1.2,0.06753919720649719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,1,power_law_1.2,1.2524864196777343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,1,power_law_1.01,0.05632640123367309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,16,power_law_1.2,5.4279296875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,32,power_law_1.2,0.07139840126037597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,1,power_law_1.2,1.9404672622680663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,32,power_law_1.2,0.07415040135383606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,1,power_law_1.01,0.06488320231437683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,32,power_law_1.2,0.07204480171203613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,1,power_law_1.01,0.08149759769439698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,1,power_law_1.2,3.803801727294922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,32,power_law_1.2,0.07251840233802795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,32,power_law_1.2,0.07611520290374756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,1,power_law_1.01,0.11758719682693482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,32,power_law_1.2,0.07674239873886109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,1,power_law_1.01,0.14721920490264892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,32,power_law_1.2,0.08060160279273987
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,32,power_law_1.2,0.08481919765472412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,1,power_law_1.01,0.16748160123825073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,32,power_law_1.2,0.08677759766578674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,8,power_law_1.2,0.03535360097885132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,32,power_law_1.2,0.0905344009399414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,1,power_law_1.01,0.2004159927368164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,8,power_law_1.2,0.037088000774383546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,32,power_law_1.2,0.1018496036529541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,8,power_law_1.2,0.034764799475669864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,8,power_law_1.2,0.03826560080051422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,1,power_law_1.01,0.21358718872070312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,32,power_law_1.2,0.11660159826278686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,8,power_law_1.2,0.03715839982032776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,8,power_law_1.2,0.04993920028209686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,8,power_law_1.2,0.05256320238113403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,32,power_law_1.2,0.14074239730834961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,8,power_law_1.2,0.04445439875125885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,1,power_law_1.01,0.21840639114379884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,8,power_law_1.2,0.05425919890403748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,8,power_law_1.2,0.03638400137424469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,8,power_law_1.2,0.05324800014495849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,32,power_law_1.2,0.16273280382156372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,8,power_law_1.2,0.03460479974746704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,8,power_law_1.2,0.054630398750305176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,8,power_law_1.2,0.03863680064678192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,1,power_law_1.01,0.23235199451446534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,8,power_law_1.2,0.055769598484039305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,8,power_law_1.2,0.040934398770332336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,8,power_law_1.2,0.05583999752998352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,32,power_law_1.2,0.20972158908843994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,8,power_law_1.2,0.04492799937725067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,8,power_law_1.2,0.05850239992141724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,8,power_law_1.2,0.04564479887485504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,8,power_law_1.2,0.05809919834136963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,1,power_law_1.01,0.24646399021148682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,8,power_law_1.2,0.048089599609375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,32,power_law_1.2,0.24725759029388428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,8,power_law_1.2,0.06755840182304382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,8,power_law_1.2,0.047968000173568726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,8,power_law_1.2,0.06839680075645446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,8,power_law_1.2,0.04806399941444397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,8,power_law_1.2,0.07002239823341369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,32,power_law_1.2,0.3338239908218384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,8,power_law_1.2,0.04805760085582733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,1,power_law_1.01,0.2576128005981445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,8,power_law_1.2,0.08112639784812928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,8,power_law_1.2,0.04893440008163452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,8,power_law_1.2,0.07929599881172181
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,8,power_law_1.2,0.0506816029548645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,8,power_law_1.2,0.09327359795570374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,32,power_law_1.2,0.43344640731811523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,8,power_law_1.2,0.05351679921150208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,8,power_law_1.2,0.09991040229797363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,1,power_law_1.01,0.2742975950241089
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,8,power_law_1.2,0.060038399696350095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,8,power_law_1.2,0.12120959758758545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,8,power_law_1.2,0.06109439730644226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,8,power_law_1.2,0.15586559772491454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,32,power_law_1.2,0.6051519870758056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,8,power_law_1.2,0.06587520241737366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,8,power_law_1.2,0.1949056029319763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,1,power_law_1.01,0.29616000652313235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,8,power_law_1.2,0.07714560031890869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,8,power_law_1.2,0.07992960214614868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,8,power_law_1.2,0.2543999910354614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,8,power_law_1.2,0.09704319834709167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,32,power_law_1.2,0.7689727783203125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,8,power_law_1.2,0.3842175960540771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,8,power_law_1.2,0.10993280410766601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,1,power_law_1.01,0.3107968091964722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,8,power_law_1.2,0.1287551999092102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,8,power_law_1.2,0.391379189491272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,8,power_law_1.2,0.18053120374679565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,32,power_law_1.2,1.0290176391601562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,1,8,power_law_1.2,0.511084794998169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,1,power_law_1.01,0.3122879981994629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,8,power_law_1.2,0.20793600082397462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,8,power_law_1.2,0.262444806098938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,1,8,power_law_1.2,0.8456831932067871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,8,power_law_1.2,0.35815041065216063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,1,power_law_1.01,0.36855039596557615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,32,power_law_1.2,1.6939008712768555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,8,power_law_1.2,0.5330368041992187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,1,8,power_law_1.2,1.613542366027832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,1,power_law_1.01,0.40694398880004884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,2688,1856,6,128,1,8,power_law_1.2,0.6834239959716797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,2688,1856,6,128,1,8,power_law_1.2,0.9230719566345215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,1,power_law_1.01,0.5053631782531738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,32,power_law_1.2,3.1265920639038085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,2688,1856,6,128,1,8,power_law_1.2,1.922265625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,1,power_law_1.01,0.5951871871948242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,1,power_law_1.01,0.7722752094268799
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,1,power_law_1.01,0.9489791870117188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,1,power_law_1.01,1.2989503860473632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,1,power_law_1.01,1.6152191162109375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,4,1,balanced,0.04808533191680908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,4,1,balanced,0.0663679987192154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,4,1,balanced,0.09083200494448344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,4,1,balanced,0.09057066837946574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,4,1,balanced,0.09211732943852742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,4,1,balanced,0.09294933080673218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,4,1,balanced,0.09431999921798706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,4,1,balanced,0.09266133109728496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,4,1,balanced,0.09410132964452107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,4,1,balanced,0.094842662413915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,4,1,balanced,0.09494933485984802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,4,1,balanced,0.09689066807428996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,4,1,balanced,0.09759466846783955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,4,1,balanced,0.10633599758148193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,4,1,balanced,0.11477333307266235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,4,1,balanced,0.11668800314267476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,4,1,balanced,0.12098133563995361
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,4,1,balanced,0.14270400007565817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,4,1,balanced,0.16192000110944113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,4,1,balanced,0.20276800791422525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,4,1,balanced,0.22209600607554117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,4,1,balanced,0.28544000784556073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,4,1,balanced,0.3538186550140381
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,1,power_law_1.01,2.2948415756225584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,4,1,balanced,0.49537599086761475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,4,1,balanced,0.63427734375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,4,1,balanced,0.9170293013254801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,4,1,balanced,1.1956053574879963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,4,1,balanced,1.471402645111084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,4,1,balanced,2.3108693758646646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,16,balanced,0.046069333950678505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,16,balanced,0.049925332268079124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,16,balanced,0.049269333481788635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,16,balanced,0.061903998255729675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,16,balanced,0.08058133224646251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,16,balanced,0.08303999900817871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,16,balanced,0.08292266726493835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,16,balanced,0.08468266328175862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,16,balanced,0.0862559974193573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,16,balanced,0.08396266897519429
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,16,balanced,0.08747733632723491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,4,1,balanced,4.548288027445476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,16,balanced,0.08947199583053589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,16,balanced,0.08874133229255676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,16,balanced,0.0936959981918335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,1,power_law_1.01,2.969875144958496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,16,balanced,0.09548800190289815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,16,balanced,0.10310932993888855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,16,balanced,0.10378666718800862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,16,balanced,0.11528533697128296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,16,power_law_1.01,0.04028800129890442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,16,balanced,0.12743999560674033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,16,balanced,0.15573867162068686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,16,balanced,0.1728000044822693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,16,balanced,0.23616532484690347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,16,balanced,0.23385600248972574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,16,balanced,0.328384002049764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,16,power_law_1.01,0.0385919988155365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,16,balanced,0.37539732456207275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,16,balanced,0.5291093190511068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,16,balanced,0.6409226655960083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,16,power_law_1.01,0.035097599029541016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,4,16,balanced,0.7897013028462728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,4,16,balanced,1.1680959860483806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,16,power_law_1.01,0.036083200573921205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,4,16,balanced,2.3997012774149575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,16,power_law_1.01,0.036883199214935304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,1,power_law_1.01,3.6537216186523436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,16,power_law_1.01,0.03919999897480011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,16,power_law_1.01,0.03983359932899475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,16,power_law_1.01,0.039852800965309146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,16,power_law_1.01,0.03996799886226654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,16,power_law_1.01,0.040057599544525146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,16,power_law_1.01,0.04090240001678467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,1,power_law_1.01,5.795347213745117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,16,power_law_1.01,0.04184960126876831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,16,power_law_1.01,0.042630401253700254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,16,power_law_1.01,0.044863998889923096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,16,power_law_1.01,0.050444799661636355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,16,power_law_1.01,0.052153599262237546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,16,power_law_1.01,0.05436800122261047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,16,power_law_1.01,0.06757760047912598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,16,power_law_1.01,0.06651520133018493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,16,power_law_1.01,0.08216320276260376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,16,balanced,0.050944000482559204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,16,balanced,0.05379733443260193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,16,balanced,0.0728959987560908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,16,balanced,0.0981226662794749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,16,balanced,0.14648000399271646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,16,power_law_1.01,0.09388160109519958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,16,balanced,0.24544533093770346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,16,balanced,0.34030401706695557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,16,balanced,0.3466453154881795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,16,balanced,0.34674668312072754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,16,balanced,0.34812267621358234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,1,power_law_1.01,11.042540740966796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,16,balanced,0.34815998872121173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,16,balanced,0.3482079903284709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,16,power_law_1.01,0.12607359886169434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,16,balanced,0.34833598136901855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,16,balanced,0.3482026656468709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,16,balanced,0.35524264971415204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,16,balanced,0.355402668317159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,16,balanced,0.36211200555165607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,16,balanced,0.3700000047683716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,16,balanced,0.376362681388855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,16,balanced,0.39393067359924316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,16,power_law_1.01,0.13723520040512086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,16,balanced,0.41078933080037433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,16,balanced,0.44339199860890705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,16,balanced,0.4789493481318156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,16,balanced,0.5365066528320312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,16,balanced,0.6339519818623861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,16,power_law_1.01,0.18933119773864746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,16,balanced,0.7436319986979166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,16,balanced,0.9589280287424723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,16,balanced,1.1634079615275066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,16,power_law_1.01,0.24021759033203124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,16,balanced,1.5676587422688801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,16,balanced,2.9943040211995444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,16,power_law_1.01,0.3006975889205933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,16,power_law_1.01,0.4843264102935791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,16,power_law_1.01,0.49541120529174804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,1,power_law_1.2,0.09191039800643921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,16,power_law_1.01,0.7201663970947265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,1,power_law_1.2,0.11456639766693115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,1,power_law_1.2,0.12534400224685668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,1,power_law_1.2,0.1640895962715149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,16,power_law_1.01,1.5753024101257325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,1,power_law_1.2,0.24651520252227782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,4,power_law_1.01,0.03968639969825745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,1,power_law_1.2,0.3287487983703613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,4,power_law_1.01,0.04325760006904602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,1,power_law_1.2,0.3749504089355469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,4,power_law_1.01,0.03804160058498383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,32,power_law_1.2,0.061913597583770755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,32,power_law_1.2,0.04206080138683319
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,1,power_law_1.2,0.5182464122772217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,4,power_law_1.01,0.046419200301170346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,32,power_law_1.2,0.03715200126171112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,32,power_law_1.2,0.036364799737930296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,4,power_law_1.01,0.05511040091514587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,1,power_law_1.2,0.5367807865142822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,32,power_law_1.2,0.041382399201393125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,4,power_law_1.01,0.06648319959640503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,32,power_law_1.2,0.049184000492095946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,32,power_law_1.2,0.05301759839057922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,1,power_law_1.2,0.5712831974029541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,4,power_law_1.01,0.0681984007358551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,32,power_law_1.2,0.05338240265846252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,32,power_law_1.2,0.05403519868850708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,4,power_law_1.01,0.06895359754562377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,1,power_law_1.2,0.6053247928619385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,32,power_law_1.2,0.05414400100708008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,4,power_law_1.01,0.07178239822387696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,32,power_law_1.2,0.056383997201919556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,4,power_law_1.01,0.07212799787521362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,32,power_law_1.2,0.05730559825897217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,1,power_law_1.2,0.6325952053070069
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,32,power_law_1.2,0.05666559934616089
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,4,power_law_1.01,0.07212799787521362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,32,power_law_1.2,0.060134398937225345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,4,power_law_1.2,0.10030720233917237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,4,power_law_1.01,0.0716480016708374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,32,power_law_1.2,0.06823679804801941
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,1,power_law_1.2,0.6519807815551758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,4,power_law_1.2,0.08023040294647217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,4,power_law_1.01,0.07484800219535828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,32,power_law_1.2,0.07022719979286193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,32,power_law_1.2,0.0718720018863678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,4,power_law_1.2,0.09467520117759705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,4,power_law_1.01,0.08154879808425904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,1,power_law_1.2,0.6825664043426514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,32,power_law_1.2,0.0796992003917694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,4,power_law_1.2,0.1363136053085327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,4,power_law_1.01,0.08855040073394775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,32,power_law_1.2,0.08803840279579163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,4,power_law_1.2,0.16293120384216309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,4,power_law_1.01,0.09431040287017822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,32,power_law_1.2,0.10899840593338013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,1,power_law_1.2,0.7232831954956055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,4,power_law_1.2,0.2541568040847778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,32,power_law_1.2,0.11040639877319336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,4,power_law_1.01,0.09447680115699768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,4,power_law_1.2,0.2629184007644653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,32,power_law_1.2,0.15682560205459595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,1,power_law_1.2,0.7472320079803467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,4,power_law_1.01,0.10224640369415283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,4,power_law_1.2,0.26695680618286133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,32,power_law_1.2,0.17787519693374634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,4,power_law_1.01,0.11345920562744141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,4,power_law_1.2,0.287827205657959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,32,power_law_1.2,0.28256640434265134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,1,power_law_1.2,0.7839168071746826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,4,power_law_1.01,0.12929919958114625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,4,power_law_1.2,0.29839999675750734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,32,power_law_1.2,0.36808319091796876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,4,power_law_1.2,0.2987967967987061
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,4,power_law_1.01,0.12517759799957276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,1,power_law_1.2,0.8798527717590332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,32,power_law_1.2,0.488588809967041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,4,power_law_1.2,0.3019776105880737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,4,power_law_1.01,0.14466559886932373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,4,power_law_1.2,0.3001471996307373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,1,power_law_1.2,0.9559616088867188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,32,power_law_1.2,1.0513600349426269
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,4,power_law_1.01,0.1820736050605774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,4,power_law_1.2,0.3180927991867065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,4,power_law_1.2,0.34406399726867676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,32,power_law_1.2,1.048038387298584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,4,power_law_1.01,0.21909759044647217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,1,power_law_1.2,1.1154751777648926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,4,power_law_1.2,0.3538431882858276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,4,power_law_1.01,0.2646591901779175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,4,power_law_1.2,0.35329279899597166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,1,power_law_1.2,1.200921630859375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,32,power_law_1.2,1.5954496383666992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,4,power_law_1.2,0.39673600196838377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,4,power_law_1.01,0.34291839599609375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,4,power_law_1.2,0.4439295768737793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,1,power_law_1.2,1.4237824440002442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,4,power_law_1.01,0.43445758819580077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,4,power_law_1.2,0.5269887924194336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,4,power_law_1.2,0.6058559894561768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,2,4,power_law_1.01,0.506105613708496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,32,power_law_1.2,3.7374847412109373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,1,power_law_1.2,1.727097511291504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,4,power_law_1.2,0.7456319808959961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,4,power_law_1.2,0.8214207649230957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,2,4,power_law_1.01,0.7905792236328125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,1,power_law_1.2,2.1707008361816404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,4,power_law_1.2,0.9935680389404297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,4,power_law_1.2,1.2963647842407227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,2,4,power_law_1.01,1.464838409423828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,1,power_law_1.2,2.6372095108032227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,4,power_law_1.2,2.06616325378418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,1,power_law_1.2,3.4589889526367186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,4,power_law_1.2,2.2587968826293947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,4,power_law_1.2,2.737209510803223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,1,power_law_1.2,4.265862274169922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,4,power_law_1.2,4.119558334350586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,32,2,power_law_1.01,0.046828800439834596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,1,power_law_1.2,5.123603057861328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,32,2,power_law_1.01,0.049644801020622256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,32,2,power_law_1.01,0.05303679704666138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,32,2,power_law_1.01,0.05877760052680969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,32,2,power_law_1.01,0.06730239987373351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,4,power_law_1.2,8.542240142822266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,32,2,power_law_1.01,0.08983680009841918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,4,1,power_law_1.01,0.07708160281181335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,1,power_law_1.2,7.687117004394532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,32,2,power_law_1.01,0.09496960043907166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,4,1,power_law_1.01,0.09496319890022278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,4,1,power_law_1.01,0.10766079425811767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,32,2,power_law_1.01,0.09739519953727722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,4,1,power_law_1.01,0.20493440628051757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,4,1,power_law_1.01,0.20724480152130126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,32,2,power_law_1.01,0.0986240029335022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,4,1,power_law_1.01,0.20764799118041993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,4,1,power_law_1.01,0.20529279708862305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,32,2,power_law_1.01,0.10047359466552734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,4,1,power_law_1.01,0.2135103940963745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,32,2,power_law_1.01,0.10296319723129273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,4,1,power_law_1.01,0.21534719467163085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,4,1,power_law_1.01,0.20879359245300294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,32,2,power_law_1.01,0.10544639825820923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,4,1,power_law_1.01,0.21822719573974608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,4,1,power_law_1.01,0.22062718868255615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,32,2,power_law_1.01,0.10661120414733886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,4,1,power_law_1.01,0.2285952091217041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,4,1,power_law_1.01,0.23547520637512206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,32,2,power_law_1.01,0.11144319772720337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,4,1,power_law_1.01,0.2504447937011719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,4,1,power_law_1.01,0.26077439785003664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,32,2,power_law_1.01,0.12188800573348998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,4,1,power_law_1.01,0.27393920421600343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,4,1,power_law_1.01,0.31880960464477537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,32,2,power_law_1.01,0.12936320304870605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,4,1,power_law_1.01,0.3446592092514038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,1,power_law_1.2,14.325375366210938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,4,1,power_law_1.01,0.41861758232116697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,32,2,power_law_1.01,0.13686399459838866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,4,1,power_law_1.01,0.4880256175994873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,4,1,power_law_1.01,0.6459775924682617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,32,2,power_law_1.01,0.1575935959815979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,4,1,power_law_1.01,0.781657600402832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,32,2,power_law_1.01,0.1879487991333008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,4,1,power_law_1.01,1.0847104072570801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,32,2,power_law_1.01,0.23948800563812256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,4,1,power_law_1.01,1.4103232383728028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,32,2,power_law_1.01,0.2983871936798096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,4,1,power_law_1.01,2.0535295486450194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,32,2,power_law_1.01,0.3749567985534668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,8,4,balanced,0.0359199990828832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,8,4,balanced,0.033717334270477295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,4,1,power_law_1.01,2.6733503341674805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,8,4,balanced,0.037615999579429626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,8,4,balanced,0.035391998787721
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,8,4,balanced,0.035743998984495796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,8,4,balanced,0.03600533306598663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,8,4,balanced,0.03772799919048945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,8,4,balanced,0.03779733429352442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,8,4,balanced,0.03743999948104223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,8,4,balanced,0.03946666667858759
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,8,4,balanced,0.0415786678592364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,8,4,balanced,0.04167466859022776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,8,4,balanced,0.04158399999141693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,8,4,balanced,0.043712000052134194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,8,4,balanced,0.053904001911481224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,8,4,balanced,0.05426666637261709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,32,2,power_law_1.01,0.4432064056396484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,8,4,balanced,0.05524266759554545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,8,4,balanced,0.06850133339564006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,8,4,balanced,0.07995200157165527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,8,4,balanced,0.0936906635761261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,8,4,balanced,0.10727999607721965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,8,4,balanced,0.14018133282661438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,8,4,balanced,0.16869332393010458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,4,1,power_law_1.01,3.329075241088867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,8,4,balanced,0.2283359964688619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,8,4,balanced,0.28735466798146564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,8,4,balanced,0.40356798966725665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,8,4,balanced,0.5036906798680624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,32,2,power_law_1.01,0.574944019317627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,8,4,balanced,0.6223306655883789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,8,4,balanced,0.947925329208374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,8,4,balanced,1.82149871190389
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,32,2,power_law_1.01,0.8005696296691894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,4,1,power_law_1.01,5.02215690612793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,32,2,power_law_1.01,1.0449983596801757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,32,2,power_law_1.01,1.3851519584655763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,4,1,power_law_1.01,9.78784637451172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,32,2,power_law_1.01,1.7577983856201171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,8,power_law_1.2,0.1027519941329956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,8,power_law_1.2,0.08196480274200439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,8,power_law_1.2,0.07342079877853394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,32,2,power_law_1.01,2.6527999877929687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,8,power_law_1.2,0.09868159890174866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,8,power_law_1.2,0.11100159883499146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,8,power_law_1.2,0.14862719774246216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,8,power_law_1.2,0.15895040035247804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,8,power_law_1.2,0.16234879493713378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,32,2,power_law_1.01,4.964019012451172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,8,power_law_1.2,0.1569983959197998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,8,power_law_1.2,0.16815359592437745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,8,power_law_1.2,0.16783360242843628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,1,power_law_1.01,0.10160640478134156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,8,power_law_1.2,0.17824000120162964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,1,power_law_1.01,0.14289920330047606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,8,power_law_1.2,0.18298879861831666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,1,power_law_1.01,0.20225920677185058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,8,power_law_1.2,0.18424960374832153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,1,power_law_1.01,0.33251841068267823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,8,2,power_law_1.01,0.037478399276733396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,8,power_law_1.2,0.1960319995880127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,1,power_law_1.01,0.46943359375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,8,2,power_law_1.01,0.03736959993839264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,8,2,power_law_1.01,0.038022398948669434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,8,power_law_1.2,0.20611200332641602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,1,power_law_1.01,0.825875186920166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,8,2,power_law_1.01,0.03728640079498291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,8,power_law_1.2,0.210316801071167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,8,2,power_law_1.01,0.037567999958992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,1,power_law_1.01,0.8585536003112793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,8,2,power_law_1.01,0.037747201323509214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,8,power_law_1.2,0.25852160453796386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,8,2,power_law_1.01,0.03798399865627289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,1,power_law_1.01,0.8878911972045899
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,8,2,power_law_1.01,0.038675200939178464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,8,power_law_1.2,0.26988160610198975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,8,2,power_law_1.01,0.040300801396369934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,8,2,power_law_1.01,0.040531200170516965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,1,power_law_1.01,0.9351424217224121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,8,2,power_law_1.01,0.0424703985452652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,8,power_law_1.2,0.30855679512023926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,8,2,power_law_1.01,0.04415999948978424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,1,power_law_1.01,0.9735360145568848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,8,2,power_law_1.01,0.044064000248909
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,8,power_law_1.2,0.32417280673980714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,8,2,power_law_1.01,0.047968000173568726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,8,2,power_law_1.01,0.057171201705932616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,1,power_law_1.01,0.9876416206359864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,8,power_law_1.2,0.4195072174072266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,8,2,power_law_1.01,0.05864319801330566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,8,2,power_law_1.01,0.06238080263137817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,1,power_law_1.01,1.0165184020996094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,8,2,power_law_1.01,0.06829439997673034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,8,power_law_1.2,0.479750394821167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,8,2,power_law_1.01,0.08241279721260071
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,8,2,power_law_1.01,0.09704319834709167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,1,power_law_1.01,1.045625591278076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,8,2,power_law_1.01,0.11200640201568604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,8,power_law_1.2,0.5673471927642822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,8,2,power_law_1.01,0.14913920164108277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,32,power_law_1.2,0.04910080134868622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,1,power_law_1.01,1.074400043487549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,8,2,power_law_1.01,0.18556159734725952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,8,power_law_1.2,0.7551360130310059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,32,power_law_1.2,0.04467839896678925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,8,2,power_law_1.01,0.23029758930206298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,1,power_law_1.01,1.1105728149414062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,8,2,power_law_1.01,0.2816384077072144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,32,power_law_1.2,0.044435200095176694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,8,2,power_law_1.01,0.4166912078857422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,8,power_law_1.2,1.0623231887817384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,1,power_law_1.01,1.1434111595153809
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,32,power_law_1.2,0.04392319917678833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,8,2,power_law_1.01,0.6481791973114014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,32,power_law_1.2,0.04586879909038544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,8,2,power_law_1.01,0.6792064189910889
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,1,power_law_1.01,1.1532223701477051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,8,power_law_1.2,1.3284992218017577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,32,power_law_1.2,0.04702720046043396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,8,2,power_law_1.01,1.073862361907959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,1,power_law_1.01,1.2477503776550294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,32,power_law_1.2,0.04800640046596527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,32,power_law_1.2,0.0491456001996994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,1,power_law_1.01,1.3064191818237305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,8,power_law_1.2,1.5722368240356446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,8,2,power_law_1.01,2.0622400283813476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,32,power_law_1.2,0.04925439953804016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,1,power_law_1.01,1.4520319938659667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,32,power_law_1.2,0.05067520141601563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,32,power_law_1.2,0.05236480236053467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,1,power_law_1.01,1.636422348022461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,8,power_law_1.2,2.395910453796387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,32,power_law_1.2,0.054611200094223024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,1,power_law_1.01,1.9925056457519532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,32,power_law_1.2,0.0585919976234436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,32,power_law_1.2,0.06096640229225159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,1,power_law_1.01,2.0168127059936523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,32,power_law_1.2,0.06543999910354614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,32,power_law_1.2,0.06897280216217042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,8,power_law_1.2,4.580249786376953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,1,power_law_1.01,2.405209541320801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,32,power_law_1.2,0.07422720193862915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,32,power_law_1.2,0.09054080247879029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,1,power_law_1.01,3.2624256134033205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,32,power_law_1.2,0.10097279548645019
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,32,power_law_1.2,0.1298815965652466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,32,power_law_1.2,0.1502784013748169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,1,power_law_1.01,4.1447807312011715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,32,power_law_1.2,0.19015040397644042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,32,power_law_1.2,0.2219327926635742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,32,power_law_1.2,0.33152639865875244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,2,balanced,0.03565866748491923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,1,power_law_1.01,5.153792190551758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,2,balanced,0.04934933284918467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,2,balanced,0.06667733192443848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,2,balanced,0.107205331325531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,2,balanced,0.16537599762280783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,2,balanced,0.21394133567810059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,2,balanced,0.21453867355982462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,2,balanced,0.21370132764180502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,2,balanced,0.21353065967559814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,32,power_law_1.2,0.39531519412994387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,2,balanced,0.21492266654968262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,2,balanced,0.21801066398620605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,2,balanced,0.21801066398620605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,2,balanced,0.21764800945917764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,2,balanced,0.2196213404337565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,2,balanced,0.22700266043345133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,2,balanced,0.22961600621541342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,2,balanced,0.23133333524068198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,2,balanced,0.25496000051498413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,2,balanced,0.24322134256362915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,32,power_law_1.2,0.6283391952514649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,2,balanced,0.29014400641123456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,2,balanced,0.27510933081309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,2,balanced,0.41393065452575684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,2,balanced,0.343669335047404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,2,balanced,0.5102133353551229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,1,power_law_1.01,6.181817626953125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,2,balanced,0.5208213329315186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,2,balanced,0.7800107002258301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,2,balanced,0.7674187024434408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,32,power_law_1.2,0.8097023963928223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,1,2,balanced,0.9473973115285238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,1,2,balanced,1.4105119705200195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,1,2,balanced,2.7107626597086587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,32,power_law_1.2,1.037235164642334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,8,balanced,0.03012799968322118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,8,balanced,0.029743999242782593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,8,balanced,0.031850665807724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,8,balanced,0.03390933324893316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,8,balanced,0.035786665976047516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,8,balanced,0.038047999143600464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,8,balanced,0.03932799895604452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,8,balanced,0.0382080003619194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,8,balanced,0.039706667264302574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,1,power_law_1.01,8.72732162475586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,8,balanced,0.03803733239571253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,8,balanced,0.03875733415285746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,8,balanced,0.04004266609748205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,8,balanced,0.03953066716591517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,8,balanced,0.03988266736268997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,8,balanced,0.04781333108743032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,8,balanced,0.04887466629346212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,8,balanced,0.04861866434415182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,8,balanced,0.05221333106358846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,32,power_law_1.2,1.6129472732543946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,8,balanced,0.05981333553791046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,8,balanced,0.06862933437029521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,8,balanced,0.0769760012626648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,8,balanced,0.09855467081069946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,8,balanced,0.11533866326014201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,8,balanced,0.14858133594195047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,8,balanced,0.18409067392349243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,8,balanced,0.251418670018514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,8,balanced,0.3152586619059245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,8,balanced,0.3864373366038005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,8,balanced,0.5959146817525228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,8,balanced,1.122442642847697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,32,power_law_1.2,3.019443130493164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,1,power_law_1.01,16.167674255371093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,1,power_law_1.01,0.04640640020370483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,1,power_law_1.01,0.05331839919090271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,1,power_law_1.01,0.06096000075340271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,1,power_law_1.01,0.07562879920005798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,8,balanced,0.032933334509531655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,8,balanced,0.030389333764712017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,8,balanced,0.03197866678237915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,8,balanced,0.034341332813103996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,8,balanced,0.037461332976818085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,8,balanced,0.03740799923737844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,8,balanced,0.03863999992609024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,8,balanced,0.03809600075085958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,1,power_law_1.01,0.09437440037727356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,8,balanced,0.039962666730086006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,8,balanced,0.0382080003619194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,8,balanced,0.04018666595220566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,8,balanced,0.039850667119026184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,8,balanced,0.04181866844495138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,8,balanced,0.04195733368396759
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,8,balanced,0.048063998421033226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,8,balanced,0.04795733094215393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,8,balanced,0.048010667165120445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,8,balanced,0.05253866811593374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,8,balanced,0.05852266649405161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,8,balanced,0.0666133314371109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,8,balanced,0.07558399935563405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,8,balanced,0.10195733110109965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,8,balanced,0.13544533650080362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,1,power_law_1.01,0.12572799921035765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,8,balanced,0.18236800034840903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,8,balanced,0.2250666618347168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,8,balanced,0.318448007106781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,8,balanced,0.40358400344848633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,8,balanced,0.5008159875869751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,8,balanced,0.7574293613433838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,1,power_law_1.01,0.13582079410552977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,8,balanced,1.4753546714782715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,2,power_law_1.2,0.03297280073165894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,2,power_law_1.2,0.030643200874328612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,1,power_law_1.01,0.13808640241622924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,2,power_law_1.2,0.03297280073165894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,1,power_law_1.01,0.14558720588684082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,2,power_law_1.2,0.03315199911594391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,2,power_law_1.2,0.03578239977359772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,1,power_law_1.01,0.14899840354919433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,1,power_law_1.01,0.05307520031929016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,2,power_law_1.2,0.04249599874019623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,2,power_law_1.2,0.04294399917125702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,1,power_law_1.01,0.15573760271072387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,2,power_law_1.2,0.04544639885425568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,1,power_law_1.01,0.05603839755058289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,1,power_law_1.01,0.16718080043792724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,2,power_law_1.2,0.04780159890651703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,2,power_law_1.2,0.04787839949131012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,1,power_law_1.01,0.17651840448379516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,1,power_law_1.01,0.07784960269927979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,2,power_law_1.2,0.05108479857444763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,2,power_law_1.2,0.055052798986434934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,1,power_law_1.01,0.17377920150756837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,2,power_law_1.2,0.055199998617172244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,1,power_law_1.01,0.09065600037574768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,1,power_law_1.01,0.1911296010017395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,2,power_law_1.2,0.05780479907989502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,2,power_law_1.2,0.06543999910354614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,1,power_law_1.01,0.1995967984199524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,1,power_law_1.01,0.13877760171890258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,2,power_law_1.2,0.0686016023159027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,2,power_law_1.2,0.07152640223503112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,1,power_law_1.01,0.21777279376983644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,2,power_law_1.2,0.08640000224113464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,1,power_law_1.01,0.18625919818878173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,1,power_law_1.01,0.26122241020202636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,2,power_law_1.2,0.09576320052146911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,64,balanced,0.03242133309443792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,2,power_law_1.2,0.1235967993736267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,64,balanced,0.029535998900731403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,64,balanced,0.02941333254178365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,64,balanced,0.02977599948644638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,64,balanced,0.03183999905983607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,1,power_law_1.01,0.2160128116607666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,1,power_law_1.01,0.3035648107528687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,64,balanced,0.031925333042939506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,64,balanced,0.03178133318821589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,64,balanced,0.031925333042939506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,64,balanced,0.03177600105603536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,64,balanced,0.031914666295051575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,64,balanced,0.03188266605138779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,64,balanced,0.03376533339420954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,2,power_law_1.2,0.1425536036491394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,64,balanced,0.03379199902216593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,64,balanced,0.03577066709597906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,64,balanced,0.03591466695070267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,64,balanced,0.037530665596326195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,64,balanced,0.037834666669368744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,64,balanced,0.03803733239571253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,64,balanced,0.03967999915281931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,64,balanced,0.041893333196640015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,64,balanced,0.04437333345413208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,64,balanced,0.04578666885693868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,64,balanced,0.05016533533732096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,2,power_law_1.2,0.19351680278778077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,64,balanced,0.06633066634337108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,1,power_law_1.01,0.3887167930603027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,64,balanced,0.08362666765848796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,64,balanced,0.11892267068227132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,64,balanced,0.14775466918945312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,64,balanced,0.178874671459198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,1,power_law_1.01,0.2781951904296875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,64,balanced,0.2510133385658264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,64,balanced,0.4453333218892415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,2,power_law_1.2,0.22673280239105226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,1,power_law_1.01,0.4614208221435547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,2,power_law_1.2,0.34374399185180665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,1,power_law_1.01,0.28736000061035155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,1,power_law_1.01,0.6131072044372559
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,2,power_law_1.2,0.49320321083068847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,1,power_law_1.01,0.2935551881790161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,2,power_law_1.2,0.6580031871795654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,1,power_law_1.01,0.8034815788269043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,1,power_law_1.01,0.3155904054641724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,2,power_law_1.2,0.9329343795776367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,1,power_law_1.01,1.096403217315674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,1,power_law_1.01,0.32778880596160886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,2,power_law_1.2,1.158086395263672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,1,power_law_1.01,1.4412991523742675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,1,power_law_1.01,0.33819520473480225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,1,power_law_1.01,2.0230079650878907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,2,power_law_1.2,1.8624639511108398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,1,power_law_1.01,0.3464191913604736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,1,power_law_1.01,2.581452751159668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,1,power_law_1.01,0.364134407043457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,2,power_law_1.2,3.7501697540283203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,1,power_law_1.01,0.3714495897293091
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,1,balanced,0.07349333167076111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,1,balanced,0.0993173321088155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,1,power_law_1.01,3.1708095550537108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,1,balanced,0.15068266789118448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,1,balanced,0.2448319991429647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,1,balanced,0.4453973372777303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,1,balanced,0.8423306941986084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,1,balanced,1.2312266826629639
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,1,power_law_1.01,0.3924736022949219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,1,balanced,1.2364213466644287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,1,balanced,1.2433119614919026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,1,balanced,1.2465866406758626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,1,balanced,1.249679962793986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,1,balanced,1.2600853443145752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,1,balanced,1.2633439699808757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,1,power_law_1.01,0.45141119956970216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,1,balanced,1.2745227018992107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,1,balanced,1.2904053529103596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,1,power_law_1.01,4.818483352661133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,1,balanced,1.297866662343343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,1,balanced,1.329973300298055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,1,balanced,1.3762292861938477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,1,power_law_1.01,0.5032447814941406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,1,balanced,1.4243040084838867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,1,balanced,1.531285285949707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,1,balanced,1.589845339457194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,1,balanced,1.7907253901163738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,1,power_law_1.01,0.5992576122283936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,1,balanced,1.9160586992899578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,1,balanced,2.2600266138712564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,1,balanced,2.7125867207845054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,1,power_law_1.01,0.671884822845459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,1,balanced,3.3735574086507163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,1,balanced,4.501472155253093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,1,power_law_1.01,9.461529541015626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,1,power_law_1.01,0.8430208206176758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,1,balanced,5.8624693552653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,1,power_law_1.01,1.0323519706726074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,1,balanced,8.06329600016276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,128,power_law_1.01,0.038764798641204835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,128,power_law_1.01,0.03769600093364715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,1,power_law_1.01,1.360326385498047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,128,power_law_1.01,0.04028159976005554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,128,power_law_1.01,0.03818880021572113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,1,balanced,15.279834747314453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,128,power_law_1.01,0.03858560025691986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,1,power_law_1.01,1.7682367324829102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,128,power_law_1.01,0.03976320028305054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,128,power_law_1.01,0.03960959911346436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,128,power_law_1.01,0.040191999077796935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,128,power_law_1.01,0.04063999950885773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,1,power_law_1.01,2.401913642883301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,128,power_law_1.01,0.03989120125770569
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,128,power_law_1.01,0.040761598944664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,128,power_law_1.01,0.04085760116577149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,128,power_law_1.01,0.04142079949378967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,1,power_law_1.01,2.9623231887817383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,128,power_law_1.01,0.04089600145816803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,128,power_law_1.01,0.044998401403427125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,2,8,balanced,0.04213866591453552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,2,8,balanced,0.042261332273483276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,2,8,balanced,0.04191466669241587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,2,8,balanced,0.04180799921353658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,2,8,balanced,0.041797334949175514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,128,power_law_1.01,0.04769920110702515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,2,8,balanced,0.04192000130812327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,2,8,balanced,0.04154666761557261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,2,8,balanced,0.04208533465862274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,2,8,balanced,0.04228800038496653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,2,8,balanced,0.04196799794832865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,2,8,balanced,0.042453333735466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,2,8,balanced,0.043338666359583534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,2,8,balanced,0.04457066456476847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,128,power_law_1.01,0.047065600752830505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,2,8,balanced,0.044346665342648826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,2,8,balanced,0.05092266698678335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,2,8,balanced,0.052330667773882546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,2,8,balanced,0.049733335773150124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,2,8,balanced,0.05878399809201559
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,2,8,balanced,0.06142933170000712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,2,8,balanced,0.08086933195590973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,128,power_law_1.01,0.05181440114974976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,2,8,balanced,0.08167999982833862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,2,8,balanced,0.09875200192133586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,2,8,balanced,0.11369599898656209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,1,power_law_1.01,3.6130367279052735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,2,8,balanced,0.146506667137146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,2,8,balanced,0.16516266266504923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,2,8,balanced,0.22268267472585043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,128,power_law_1.01,0.057817602157592775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,2,8,balanced,0.2807573278745015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,2,8,balanced,0.3375306526819865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,2,8,balanced,0.511247992515564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,128,power_law_1.01,0.06689280271530151
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,2,8,balanced,0.9751573403676351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,128,power_law_1.01,0.07797120213508606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,128,power_law_1.01,0.10509439706802368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,128,power_law_1.01,0.13660800457000732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,1,power_law_1.01,5.587091064453125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,32,8,balanced,0.0315733328461647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,128,power_law_1.01,0.18487679958343506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,32,8,balanced,0.031397332747777305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,32,8,balanced,0.03133333226044973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,32,8,balanced,0.030426666140556335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,32,8,balanced,0.03153600047032038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,32,8,balanced,0.031514666974544525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,32,8,balanced,0.03339199970165888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,32,8,balanced,0.03160000095764796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,32,8,balanced,0.03142933299144109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,32,8,balanced,0.03336533407370249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,32,8,balanced,0.03346666693687439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,32,8,balanced,0.03359466542800268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,32,8,balanced,0.033402666449546814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,32,8,balanced,0.03446933378775915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,32,8,balanced,0.03976533313592275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,32,8,balanced,0.04159999887148539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,32,8,balanced,0.041519999504089355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,128,power_law_1.01,0.2307584047317505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,32,8,balanced,0.043765331308046974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,32,8,balanced,0.04769066472848257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,32,8,balanced,0.055589333176612854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,32,8,balanced,0.06643733382225037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,32,8,balanced,0.09278933207194011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,32,8,balanced,0.11585600177447002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,32,8,balanced,0.16148799657821655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,32,8,balanced,0.1954186757405599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,32,8,balanced,0.27369600534439087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,32,8,balanced,0.3497920036315918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,32,8,balanced,0.43331201871236164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,128,power_law_1.01,0.3165760040283203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,32,8,balanced,0.6436266501744589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,32,8,balanced,1.2354133129119873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,128,power_law_1.01,0.4111167907714844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,128,power_law_1.01,0.5561215877532959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,1,power_law_1.01,10.612576293945313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,128,power_law_1.01,0.9149760246276856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,128,power_law_1.01,1.7312192916870117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,32,power_law_1.01,0.03278720080852508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,32,power_law_1.01,0.028224000334739686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,32,power_law_1.01,0.029478400945663452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,32,power_law_1.01,0.028908801078796387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,32,power_law_1.01,0.02956799864768982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,32,power_law_1.01,0.030720001459121703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,32,power_law_1.01,0.030905601382255555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,4,power_law_1.01,0.0722815990447998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,32,power_law_1.01,0.031353598833084105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,32,power_law_1.01,0.031123200058937074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,4,power_law_1.01,0.09505280256271362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,32,power_law_1.01,0.03226880133152008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,32,power_law_1.01,0.0322816014289856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,4,power_law_1.01,0.14360320568084717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,32,power_law_1.01,0.03271040022373199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,32,power_law_1.01,0.03412480056285858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,4,power_law_1.01,0.08090879917144775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,32,4,power_law_1.2,0.04708479940891266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,32,power_law_1.01,0.03441919982433319
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,4,power_law_1.01,0.13189120292663575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,32,power_law_1.01,0.03788160085678101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,64,balanced,0.04200533529122671
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,64,balanced,0.03756266583998998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,64,balanced,0.03522666543722153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,64,balanced,0.03544000039498011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,64,balanced,0.03777066618204117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,64,balanced,0.036720000207424164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,64,balanced,0.03751466671625773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,32,4,power_law_1.2,0.05132160186767578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,64,balanced,0.03545066714286804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,64,balanced,0.03748800108830134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,32,power_law_1.01,0.03916800022125244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,64,balanced,0.037503999968369804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,64,balanced,0.03565866748491923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,64,balanced,0.03748800108830134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,64,balanced,0.03766400118668874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,64,balanced,0.03974399964014689
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,64,balanced,0.04163199911514918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,64,balanced,0.041663999358812966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,64,balanced,0.043621331453323364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,64,balanced,0.04358399907747904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,64,balanced,0.047007997830708824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,4,power_law_1.01,0.1670464038848877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,64,balanced,0.045754666129748024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,64,balanced,0.04991999765237173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,64,balanced,0.05634133517742157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,64,balanced,0.057861333092053734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,32,power_law_1.01,0.040166398882865904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,64,balanced,0.07685866455237071
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,64,balanced,0.08806399504343669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,64,balanced,0.11162133018175761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,64,balanced,0.12589866916338602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,2688,1856,6,128,1,64,balanced,0.14544533689816794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,2688,1856,6,128,1,64,balanced,0.2081600030263265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,32,4,power_law_1.2,0.05406079888343811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,2688,1856,6,128,1,64,balanced,0.365503986676534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,32,power_law_1.01,0.04301440119743347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,4,power_law_1.01,0.18915200233459473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,32,power_law_1.01,0.04523519873619079
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,32,4,power_law_1.2,0.055622398853302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,32,power_law_1.01,0.05050240159034729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,4,power_law_1.01,0.24157440662384033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,32,power_law_1.01,0.05674880146980286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,32,4,power_law_1.2,0.06105599999427795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,4,power_law_1.01,0.24566400051116943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,32,power_law_1.01,0.06810879707336426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,32,4,power_law_1.2,0.07089920043945312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,32,power_law_1.01,0.07953280210494995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,4,power_law_1.01,0.26109440326690675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,32,power_law_1.01,0.10227839946746826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,32,4,power_law_1.2,0.07505919933319091
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,4,power_law_1.01,0.2783168077468872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,32,power_law_1.01,0.1468287944793701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,32,4,power_law_1.2,0.08942720293998718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,32,power_law_1.01,0.18686720132827758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,4,power_law_1.01,0.28254079818725586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,32,4,power_law_1.2,0.09650560021400452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,32,power_law_1.01,0.25899519920349123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,4,power_law_1.01,0.28501760959625244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,32,4,power_law_1.2,0.09779199957847595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,32,power_law_1.01,0.30499839782714844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,4,power_law_1.01,0.2865216016769409
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,32,4,power_law_1.2,0.10020480155944825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,4,power_law_1.01,0.3128576040267944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,2,2,balanced,0.042362665136655174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,2,2,balanced,0.051872000098228455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,2,2,balanced,0.0703359991312027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,32,power_law_1.01,0.4807744026184082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,2,2,balanced,0.07101866602897644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,2,2,balanced,0.07042133311430614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,2,2,balanced,0.0724373310804367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,2,2,balanced,0.07094933092594147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,2,2,balanced,0.0727893312772115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,2,2,balanced,0.07278400162855785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,2,2,balanced,0.0724533349275589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,2,2,balanced,0.0734986662864685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,2,2,balanced,0.0740533322095871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,2,2,balanced,0.07635733485221863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,2,2,balanced,0.08177066842714946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,2,2,balanced,0.08556800087292989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,2,2,balanced,0.09133866429328918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,32,4,power_law_1.2,0.10503040552139283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,2,2,balanced,0.08954133590062459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,2,2,balanced,0.10681066910425822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,4,power_law_1.01,0.3178112030029297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,2,2,balanced,0.11405866344769795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,2,2,balanced,0.13987200458844504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,2,2,balanced,0.1634773313999176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,2,2,balanced,0.21251199642817178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,2,2,balanced,0.2635200023651123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,2,2,balanced,0.3540266752243042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,2,2,balanced,0.42576531569163006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,2,2,balanced,0.6039626598358154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,2,2,balanced,0.7858239809672037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,2,2,balanced,0.9683519999186198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,32,4,power_law_1.2,0.10757759809494019
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,4,power_law_1.01,0.32245121002197263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,2,2,balanced,1.495888074239095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,32,power_law_1.01,0.9613759994506836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,2,2,balanced,2.917146682739258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,32,4,power_law_1.2,0.11592320203781128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,4,power_law_1.01,0.354367995262146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,32,4,power_law_1.2,0.121670401096344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,4,power_law_1.01,0.38607358932495117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,32,4,power_law_1.2,0.12715519666671754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,4,power_law_1.01,0.43811841011047364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,32,4,power_law_1.2,0.13098880052566528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,4,power_law_1.01,0.4467904090881348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,32,4,power_law_1.2,0.14613759517669678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,4,power_law_1.01,0.544543981552124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,32,4,power_law_1.2,0.16297600269317628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,4,power_law_1.01,0.6304192066192627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,32,4,power_law_1.2,0.2044543981552124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,4,power_law_1.01,0.8154303550720214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,32,4,power_law_1.2,0.22684159278869628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,4,power_law_1.01,0.9246080398559571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,32,4,power_law_1.2,0.30631680488586427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,128,power_law_1.01,0.03247359991073608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,1,8,power_law_1.2,0.08529279828071594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,128,power_law_1.01,0.032793599367141726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,1,8,power_law_1.2,0.08739839792251587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,1,8,power_law_1.2,0.08627840280532836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,128,power_law_1.01,0.032332798838615416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,4,power_law_1.01,1.2783871650695802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,32,4,power_law_1.2,0.3962496042251587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,1,8,power_law_1.2,0.08503040075302123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,128,power_law_1.01,0.03278720080852508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,1,8,power_law_1.2,0.08384640216827392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,128,power_law_1.01,0.033036801218986514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,1,8,power_law_1.2,0.08733440041542054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,1,8,power_law_1.2,0.09969919919967651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,128,power_law_1.01,0.034041601419448855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,32,4,power_law_1.2,0.5114431858062745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,1,8,power_law_1.2,0.1155392050743103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,4,power_law_1.01,1.5477503776550292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,128,power_law_1.01,0.034272000193595886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,1,8,power_law_1.2,0.14440319538116456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,1,8,power_law_1.2,0.15361920595169068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,128,power_law_1.01,0.03445119857788086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,1,8,power_law_1.2,0.15393279790878295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,128,power_law_1.01,0.03427839875221252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,1,8,power_law_1.2,0.1758463978767395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,32,4,power_law_1.2,0.6419648170471192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,1,8,power_law_1.2,0.18873599767684937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,128,power_law_1.01,0.03406080007553101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,1,8,power_law_1.2,0.1299839973449707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,4,power_law_1.01,1.8983680725097656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,128,power_law_1.01,0.03477759957313538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,1,8,power_law_1.2,0.15176960229873657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,1,8,power_law_1.2,0.16970239877700805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,128,power_law_1.01,0.03558399975299835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,1,8,power_law_1.2,0.16942720413208007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,32,4,power_law_1.2,0.9347711563110351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,128,power_law_1.01,0.03633280098438263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,1,8,power_law_1.2,0.23117439746856688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,1,8,power_law_1.2,0.272377610206604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,128,power_law_1.01,0.03675520122051239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,1,8,power_law_1.2,0.3654016017913818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,128,power_law_1.01,0.03898240029811859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,1,8,power_law_1.2,0.403334379196167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,4,power_law_1.01,2.6129024505615233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,128,power_law_1.01,0.039468801021575926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,1,8,power_law_1.2,0.5751039981842041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,32,4,power_law_1.2,1.1943231582641602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,1,8,power_law_1.2,0.8148544311523438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,128,power_law_1.01,0.04023039937019348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,1,8,power_law_1.2,1.225055980682373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,128,power_law_1.01,0.042771199345588685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,128,power_law_1.01,0.044460800290107724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,1,8,power_law_1.2,1.378495979309082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,128,power_law_1.01,0.049472001194953916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,32,4,power_law_1.2,1.5254783630371094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,1,8,power_law_1.2,2.1469568252563476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,128,power_law_1.01,0.055199998617172244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,128,power_law_1.01,0.06273279786109924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,1,8,power_law_1.2,2.686604881286621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,4,power_law_1.01,5.1703041076660154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,128,power_law_1.01,0.07626240253448487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,1,8,power_law_1.2,3.729267120361328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,128,power_law_1.01,0.10658559799194336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,32,4,power_law_1.2,2.159257507324219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,128,power_law_1.01,0.13226239681243895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,1,8,power_law_1.2,5.966854476928711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,128,power_law_1.01,0.20572800636291505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,128,power_law_1.01,0.24995839595794678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,128,power_law_1.01,0.2873471975326538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,32,4,power_law_1.2,4.364479827880859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,1,8,power_law_1.2,11.93340835571289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,128,power_law_1.01,0.4658368110656738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,128,power_law_1.01,0.9582592010498047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,2,power_law_1.01,0.09695360064506531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,2,power_law_1.01,0.14462080001831054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,2,power_law_1.01,0.24126079082489013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,2,power_law_1.01,0.17294080257415773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,8,power_law_1.2,0.046911999583244324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,2,power_law_1.01,0.33740160465240476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,4,power_law_1.01,0.0801472008228302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,8,power_law_1.2,0.05208960175514221
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,2,power_law_1.01,0.48721919059753416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,8,power_law_1.2,0.058412802219390866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,2,power_law_1.01,0.5803008079528809
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,4,power_law_1.01,0.0890175998210907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,4,power_law_1.01,0.04970879852771759
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,8,power_law_1.2,0.05289599895477295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,2,power_law_1.01,0.7565567970275879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,4,power_law_1.01,0.08636800050735474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,8,power_law_1.2,0.05800319910049438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,4,power_law_1.01,0.06521599888801574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,2,power_law_1.01,0.805299186706543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,4,power_law_1.01,0.09898880124092102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,8,power_law_1.2,0.06490880250930786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,2,power_law_1.01,0.8615551948547363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,4,power_law_1.01,0.06590719819068909
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,4,power_law_1.01,0.08388479948043823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,8,power_law_1.2,0.0697983980178833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,4,power_law_1.01,0.09342079758644103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,2,power_law_1.01,0.8762111663818359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,8,power_law_1.2,0.08049920201301575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,4,power_law_1.01,0.1094208002090454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,4,power_law_1.01,0.09738240242004395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,2,power_law_1.01,0.9164480209350586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,8,power_law_1.2,0.08288639783859253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,4,power_law_1.01,0.127839994430542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,2,power_law_1.01,0.9553919792175293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,8,power_law_1.2,0.08602880239486695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,4,power_law_1.01,0.12885119915008544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,4,power_law_1.01,0.15088000297546386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,2,power_law_1.01,0.9786175727844239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,8,power_law_1.2,0.0870464026927948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,4,power_law_1.01,0.15571839809417726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,2,power_law_1.01,1.0217344284057617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,4,power_law_1.01,0.1364416003227234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,8,power_law_1.2,0.09041919708251953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,4,power_law_1.01,0.16060160398483275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,2,power_law_1.01,1.0442496299743653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,8,power_law_1.2,0.0972927987575531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,4,power_law_1.01,0.1651136040687561
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,4,power_law_1.01,0.1408128023147583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,8,power_law_1.2,0.10184320211410522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,2,power_law_1.01,1.104748821258545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,4,power_law_1.01,0.17538559436798096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,8,power_law_1.2,0.10926079750061035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,2,power_law_1.01,1.1542719841003417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,4,power_law_1.01,0.17749760150909424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,4,power_law_1.01,0.14702080488204955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,8,power_law_1.2,0.11380480527877808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,4,power_law_1.01,0.1870144009590149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,2,power_law_1.01,1.2451583862304687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,8,power_law_1.2,0.1245568037033081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,4,power_law_1.01,0.1511296033859253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,4,power_law_1.01,0.19251840114593505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,2,power_law_1.01,1.4052672386169434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,8,power_law_1.2,0.14101120233535766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,4,power_law_1.01,0.20149760246276854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,2,power_law_1.01,1.429638385772705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,8,power_law_1.2,0.1575103998184204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,4,power_law_1.01,0.21198079586029053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,2,power_law_1.01,1.6869247436523438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,8,power_law_1.2,0.1992319941520691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,4,power_law_1.01,0.1571071982383728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,4,power_law_1.01,0.22716159820556642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,8,power_law_1.2,0.21850240230560303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,2,power_law_1.01,1.8569536209106445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,4,power_law_1.01,0.24913280010223388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,8,power_law_1.2,0.284172797203064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,4,power_law_1.01,0.2915776014328003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,2,power_law_1.01,2.279999923706055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,4,power_law_1.01,0.16054400205612182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,4,power_law_1.01,0.320032000541687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,8,power_law_1.2,0.3628864049911499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,2,power_law_1.01,2.4726144790649416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,4,power_law_1.01,0.39806718826293946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,8,power_law_1.2,0.5027584075927735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,4,power_law_1.01,0.16328959465026854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,4,power_law_1.01,0.48497281074523924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,2,power_law_1.01,3.1009536743164063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,8,power_law_1.2,0.6400320053100585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,4,power_law_1.01,0.6098048210144043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,8,power_law_1.2,0.8753215789794921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,4,power_law_1.01,0.16933759450912475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,2,power_law_1.01,3.742732620239258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,4,power_law_1.01,0.7343167781829834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,8,power_law_1.2,1.1248831748962402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,4,power_law_1.01,1.0635007858276366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,4,power_law_1.01,0.17558399438858033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,2,power_law_1.01,4.4551551818847654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,8,power_law_1.2,1.414739227294922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,4,power_law_1.01,1.310649585723877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,4,power_law_1.01,0.18574719429016112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,4,power_law_1.01,1.5555264472961425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,2,power_law_1.01,6.515865325927734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,8,power_law_1.2,2.551136016845703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,4,power_law_1.01,0.1933632016181946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,4,power_law_1.01,2.413983917236328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,8,power_law_1.2,4.463711929321289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,4,power_law_1.01,0.2234879970550537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,2,power_law_1.01,12.418150329589844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,4,power_law_1.01,4.431724929809571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,4,power_law_1.01,0.24079999923706055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,4,power_law_1.01,0.2954495906829834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,4,power_law_1.01,0.37342081069946287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,4,power_law_1.01,0.4778304100036621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,4,power_law_1.01,0.5153728008270264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,8,power_law_1.01,0.03922559916973114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,8,power_law_1.01,0.03402880132198334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,4,power_law_1.01,0.701855993270874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,8,power_law_1.01,0.03387520015239716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,2,power_law_1.2,0.053286397457122804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,8,power_law_1.01,0.03429119884967804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,2,power_law_1.2,0.05371519923210144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,8,power_law_1.01,0.035718399286270144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,2,power_law_1.2,0.06565759778022766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,8,power_law_1.01,0.0382207989692688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,1,power_law_1.01,0.05315840244293213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,2,power_law_1.2,0.08831999897956848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,8,power_law_1.01,0.038176000118255615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,4,power_law_1.01,0.8488896369934082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,2,power_law_1.2,0.10268800258636475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,8,power_law_1.01,0.03874559998512268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,2,power_law_1.2,0.17140480279922485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,8,power_law_1.01,0.039212799072265624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,1,power_law_1.01,0.05767040252685547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,8,power_law_1.01,0.03929600119590759
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,2,power_law_1.2,0.18689919710159303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,8,power_law_1.01,0.040556800365447995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,2,power_law_1.2,0.18751360177993776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,8,power_law_1.01,0.041171199083328246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,2,power_law_1.2,0.19943679571151735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,1,power_law_1.01,0.08173440098762512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,8,power_law_1.01,0.04219520092010498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,2,power_law_1.2,0.20072319507598876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,4,power_law_1.01,1.1752384185791016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,8,power_law_1.01,0.044038400053977966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,2,power_law_1.2,0.20032639503479005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,1,power_law_1.01,0.0951744019985199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,8,power_law_1.01,0.04704639911651611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,2,power_law_1.2,0.21403520107269286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,8,power_law_1.01,0.048895999789237976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,2,power_law_1.2,0.2193984031677246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,8,power_law_1.01,0.05366399884223938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,2,power_law_1.2,0.2248960018157959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,1,power_law_1.01,0.14176000356674195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,8,power_law_1.01,0.05953279733657837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,2,power_law_1.2,0.24264318943023683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,8,power_law_1.01,0.06812800168991089
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,2,power_law_1.2,0.24903039932250975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,1,power_law_1.01,0.19359359741210938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,4,power_law_1.01,1.577907180786133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,8,power_law_1.01,0.07466880083084107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,2,power_law_1.2,0.2515775918960571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,8,power_law_1.01,0.0873856008052826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,2,power_law_1.2,0.2797631978988647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,1,power_law_1.01,0.21870079040527343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,8,power_law_1.01,0.1068608045578003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,2,power_law_1.2,0.3009984016418457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,2,power_law_1.2,0.3556544065475464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,8,power_law_1.01,0.12460160255432129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,1,power_law_1.01,0.26873600482940674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,2,power_law_1.2,0.38943359851837156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,8,power_law_1.01,0.1817728042602539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,2,power_law_1.2,0.4896128177642822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,8,power_law_1.01,0.21509120464324952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,1,power_law_1.01,0.2823935985565186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,2,power_law_1.2,0.6129536151885986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,4,4,power_law_1.01,1.8262016296386718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,8,power_law_1.01,0.3190464019775391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,2,power_law_1.2,0.7805759906768799
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,1,power_law_1.01,0.2869951963424683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,8,power_law_1.01,0.38034560680389407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,2,power_law_1.2,1.1202816009521483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,1,power_law_1.01,0.3080895900726318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,8,power_law_1.01,0.4645823955535889
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,2,power_law_1.2,1.3112768173217773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,1,power_law_1.01,0.3167360067367554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,8,power_law_1.01,0.7650623798370362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,2,power_law_1.2,1.8900159835815429
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,1,power_law_1.01,0.3290816068649292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,4,4,power_law_1.01,2.8241600036621093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,2,power_law_1.2,2.3262399673461913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,1,power_law_1.01,0.3550463914871216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,8,power_law_1.01,1.3358783721923828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,1,power_law_1.01,0.3704767942428589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,2,power_law_1.2,3.5816768646240233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,1,power_law_1.01,0.3963007926940918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,1,power_law_1.01,0.4347583770751953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,1,power_law_1.01,0.5187327861785889
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,2,power_law_1.2,7.165286254882813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,1,power_law_1.01,0.5250048160552978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,4,4,power_law_1.01,5.70513916015625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,1,power_law_1.01,0.6255743980407715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,1,power_law_1.01,0.7108863830566406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,1,power_law_1.01,0.9038399696350098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,8,8,power_law_1.01,0.04536960124969482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,8,8,power_law_1.01,0.055961602926254274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,1,power_law_1.01,1.104377555847168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,8,8,power_law_1.01,0.03319680094718933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,8,8,power_law_1.01,0.033471998572349546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,8,8,power_law_1.01,0.03356159925460815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,8,8,power_law_1.01,0.034272000193595886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,8,8,power_law_1.01,0.03697279989719391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,8,8,power_law_1.01,0.03798399865627289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,8,8,power_law_1.01,0.04071680009365082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,8,8,power_law_1.01,0.0434688001871109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,1,power_law_1.01,1.480185604095459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,8,8,power_law_1.01,0.04298880100250244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,8,8,power_law_1.01,0.04593279957771301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,8,8,power_law_1.01,0.04599680006504059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,8,8,power_law_1.01,0.04883840084075928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,8,8,power_law_1.01,0.06166399717330932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,8,8,power_law_1.01,0.06321280002593994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,8,8,power_law_1.01,0.07068799734115601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,8,8,power_law_1.01,0.08704000115394592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,8,8,power_law_1.01,0.09909759759902954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,1,power_law_1.01,1.8975103378295899
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,8,8,power_law_1.01,0.12680959701538086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,8,8,power_law_1.01,0.13512320518493653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,8,8,power_law_1.01,0.1845695972442627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,4,balanced,0.05645333230495453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,4,balanced,0.058005332946777344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,4,balanced,0.06428800026575725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,8,8,power_law_1.01,0.20609281063079835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,4,balanced,0.08971200386683147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,4,balanced,0.1239306628704071
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,4,balanced,0.1848213275273641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,4,balanced,0.24399999777475992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,8,8,power_law_1.01,0.3164671897888184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,4,balanced,0.24900799989700317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,4,balanced,0.2505600055058797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,4,balanced,0.2490773399670919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,4,balanced,0.2524799903233846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,8,8,power_law_1.01,0.35942399501800537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,4,balanced,0.25731199979782104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,4,balanced,0.26048000653584796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,4,balanced,0.2632906635602315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,4,balanced,0.2706720034281413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,8,8,power_law_1.01,0.5910336017608643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,4,balanced,0.2720693349838257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,4,balanced,0.28675733009974164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,1,power_law_1.01,2.5831104278564454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,4,balanced,0.29837334156036377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,4,balanced,0.317029337088267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,8,8,power_law_1.01,0.6839680194854736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,4,balanced,0.3508479992548625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,4,balanced,0.380730668703715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,4,balanced,0.4535253445307414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,4,balanced,0.5020373264948527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,8,8,power_law_1.01,0.8789312362670898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,4,balanced,0.6218080123265585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,4,balanced,0.8098719914754232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,4,balanced,1.026410659154256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,8,8,power_law_1.01,1.361996841430664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,4,balanced,1.370672067006429
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,4,balanced,1.7114453315734863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,2,power_law_1.2,0.0433023989200592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,1,power_law_1.01,3.2075454711914064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,4,balanced,2.5129226048787436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,8,8,power_law_1.01,2.5764543533325197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,2,power_law_1.2,0.04257920086383819
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,4,balanced,4.783754666646321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,2,power_law_1.2,0.047577598690986635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,1,power_law_1.01,3.906739044189453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,2,power_law_1.2,0.05397760272026062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,2,power_law_1.2,0.057868802547454835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,2,power_law_1.2,0.08382080197334289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,2,power_law_1.2,0.08476799726486206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,1,power_law_1.01,6.17402229309082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,2,power_law_1.2,0.08635519742965699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,2,power_law_1.2,0.08915839791297912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,2,power_law_1.2,0.09014400243759155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,2,power_law_1.2,0.09349120259284974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,2,power_law_1.2,0.09651839733123779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,2,power_law_1.2,0.09968000054359435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,2,power_law_1.2,0.10218240022659301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,2,power_law_1.2,0.10844800472259522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,1,power_law_1.01,11.708006286621094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,2,power_law_1.2,0.11258879899978638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,1,1,power_law_1.2,0.09767680168151856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,2,power_law_1.2,0.12021119594573974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,1,1,power_law_1.2,0.11634559631347656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,1,1,power_law_1.2,0.11143039464950562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,1,1,power_law_1.2,0.23843200206756593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,1,1,power_law_1.2,0.23941121101379395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,2,power_law_1.2,0.13498879671096803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,1,1,power_law_1.2,0.2410304069519043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,1,1,power_law_1.2,0.24391040802001954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,1,1,power_law_1.2,0.24862079620361327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,2,power_law_1.2,0.15753599405288696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,1,1,power_law_1.2,0.2506623983383179
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,8,power_law_1.01,0.10095360279083251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,1,1,power_law_1.2,0.25381760597229003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,1,1,power_law_1.2,0.25566720962524414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,1,1,power_law_1.2,0.26391680240631105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,2,power_law_1.2,0.1952831983566284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,1,1,power_law_1.2,0.2712255954742432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,1,1,power_law_1.2,0.2777535915374756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,1,1,power_law_1.2,0.30415360927581786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,2,power_law_1.2,0.22533121109008789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,1,1,power_law_1.2,0.3103872060775757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,8,power_law_1.01,0.12272000312805176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,1,1,power_law_1.2,0.3409535884857178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,1,1,power_law_1.2,0.3853951930999756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,1,1,power_law_1.2,0.44809598922729493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,2,power_law_1.2,0.31445119380950926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,1,1,power_law_1.2,0.5570432186126709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,1,1,power_law_1.2,0.6743040084838867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,2,power_law_1.2,0.41368961334228516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,1,1,power_law_1.2,0.8917632102966309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,8,power_law_1.01,0.06450560092926025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,1,1,power_law_1.2,1.1021568298339843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,2,power_law_1.2,0.5567232131958008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,1,1,power_law_1.2,1.5675200462341308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,1,1,power_law_1.2,2.0068735122680663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,8,power_law_1.01,0.07893760204315185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,2,power_law_1.2,0.693017578125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,1,1,power_law_1.2,2.966060829162598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,2,power_law_1.2,0.9659968376159668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,8,power_law_1.01,0.09419519901275634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,1,1,power_law_1.2,3.759417724609375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,2,1,balanced,0.06204266846179962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,2,power_law_1.2,1.4385087966918946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,2,1,balanced,0.10322133700052898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,2,1,balanced,0.13451199730237326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,1,1,power_law_1.2,4.657855987548828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,8,power_law_1.01,0.13000320196151732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,2,1,balanced,0.135754664738973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,2,1,balanced,0.13522133231163025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,2,1,balanced,0.13697600364685059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,2,1,balanced,0.13820266723632812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,2,1,balanced,0.14130666851997375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,2,1,balanced,0.1394719978173574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,2,1,balanced,0.1411146620909373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,2,1,balanced,0.1439520021279653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,2,1,balanced,0.14782933394114176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,2,1,balanced,0.15037866433461508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,2,1,balanced,0.15293866395950317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,2,1,balanced,0.1679626703262329
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,2,1,balanced,0.17091200749079385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,2,1,balanced,0.17667200167973837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,2,1,balanced,0.21546133359273276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,2,1,balanced,0.220960001150767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,2,1,balanced,0.28971733649571735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,2,1,balanced,0.32546667257944745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,2,1,balanced,0.43012801806132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,2,1,balanced,0.5368800163269043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,2,power_law_1.2,1.7971839904785156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,8,power_law_1.01,0.13794560432434083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,2,1,balanced,0.7755733331044515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,2,1,balanced,1.0017386277516682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,1,1,power_law_1.2,7.067129516601563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,2,1,balanced,1.4689547220865886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,2,1,balanced,1.9439627329508464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,2,1,balanced,2.40884796778361
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,8,power_law_1.01,0.13439359664916992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,2,1,balanced,3.780165354410807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,2,power_law_1.2,2.5624128341674806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,8,power_law_1.01,0.14369280338287355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,2,1,balanced,7.467669169108073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,1,1,power_law_1.2,13.580455017089843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,8,power_law_1.01,0.14175360202789306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,2,power_law_1.2,6.226476669311523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,8,power_law_1.01,0.147052800655365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,8,power_law_1.01,0.1522879958152771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,8,power_law_1.01,0.15442559719085694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,8,power_law_1.01,0.15509120225906373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,8,power_law_1.01,0.1670912027359009
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,8,power_law_1.01,0.17294080257415773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,8,power_law_1.01,0.18050559759140014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,64,power_law_1.2,0.07317759990692138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,16,power_law_1.01,0.09919999837875366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,16,balanced,0.05328000088532766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,16,balanced,0.055999999245007835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,16,balanced,0.07784533500671387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,16,balanced,0.10822932918866475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,8,power_law_1.01,0.20298240184783936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,16,balanced,0.13896532853444418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,16,balanced,0.1572480003039042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,16,balanced,0.1588053305943807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,16,balanced,0.15618133544921875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,16,balanced,0.15918399890263876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,16,power_law_1.01,0.14818559885025023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,16,balanced,0.1581493318080902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,16,balanced,0.15973333517710367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,16,balanced,0.1633386711279551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,64,power_law_1.2,0.06792960166931153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,16,balanced,0.1625173290570577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,16,balanced,0.16510933637619019
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,16,balanced,0.1718613306681315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,16,balanced,0.1713226636250814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,16,balanced,0.17822933197021484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,16,balanced,0.20089600483576456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,16,power_law_1.01,0.08835840225219727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,8,power_law_1.01,0.2255039930343628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,16,balanced,0.19831999142964682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,16,balanced,0.2488266626993815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,16,balanced,0.2480693260828654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,16,balanced,0.37166400750478107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,16,balanced,0.32572799921035767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,16,balanced,0.4601866801579793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,64,power_law_1.2,0.05846400260925293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,16,power_law_1.01,0.07816320061683654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,16,balanced,0.47307733694712323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,16,balanced,0.687509298324585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,16,balanced,0.7708426316579183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,8,power_law_1.01,0.276198410987854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,1,16,balanced,0.9521173636118571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,16,power_law_1.01,0.09683200120925903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,1,16,balanced,1.3931840260823567
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,64,power_law_1.2,0.05845119953155518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,16,power_law_1.01,0.09909759759902954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,1,16,balanced,2.806901295979818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,8,power_law_1.01,0.30113279819488525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,16,power_law_1.01,0.12216320037841796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,64,power_law_1.2,0.0657151997089386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,16,power_law_1.01,0.12311040163040161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,8,power_law_1.01,0.38172159194946287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,16,power_law_1.01,0.12606719732284546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,64,power_law_1.2,0.0667519986629486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,16,power_law_1.01,0.1327296018600464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,8,power_law_1.01,0.43176960945129395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,16,power_law_1.01,0.13720959424972534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,64,power_law_1.2,0.0693120002746582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,16,power_law_1.01,0.14158079624176026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,64,power_law_1.2,0.07528319954872131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,16,power_law_1.01,0.14529279470443726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,8,power_law_1.01,0.5867136001586915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,16,power_law_1.01,0.15511679649353027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,64,power_law_1.2,0.07744640111923218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,16,power_law_1.01,0.16327040195465087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,8,power_law_1.01,0.6873727798461914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,16,power_law_1.01,0.165555202960968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,64,power_law_1.2,0.07747200131416321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,16,power_law_1.01,0.1609344005584717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,64,power_law_1.2,0.0735360026359558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,16,power_law_1.01,0.1717695951461792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,8,power_law_1.01,1.0049152374267578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,8,power_law_1.2,0.03706879913806915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,16,power_law_1.01,0.18564480543136597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,64,power_law_1.2,0.07900800108909607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,8,power_law_1.2,0.037241598963737486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,16,power_law_1.01,0.20869119167327882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,8,power_law_1.2,0.036646398901939395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,64,power_law_1.2,0.07971199750900268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,16,power_law_1.01,0.21901440620422363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,8,power_law_1.2,0.03873279988765717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,8,power_law_1.01,1.2127872467041017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,16,power_law_1.01,0.2626559972763062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,8,power_law_1.2,0.039776000380516055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,64,power_law_1.2,0.07946239709854126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,8,power_law_1.2,0.0546176016330719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,16,power_law_1.01,0.3077568054199219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,8,power_law_1.2,0.05633919835090637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,16,power_law_1.01,0.4099264144897461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,64,power_law_1.2,0.08554880023002624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,8,power_law_1.2,0.05916799902915955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,8,power_law_1.2,0.05930240154266357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,16,power_law_1.01,0.46790399551391604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,2,8,power_law_1.01,1.4424960136413574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,64,power_law_1.2,0.08849279880523682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,8,power_law_1.2,0.05839359760284424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,16,power_law_1.01,0.6503039836883545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,8,power_law_1.2,0.059411197900772095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,64,power_law_1.2,0.09764479994773864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,8,power_law_1.2,0.06128000020980835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,16,power_law_1.01,0.7878975868225098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,8,power_law_1.2,0.06477439999580384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,64,power_law_1.2,0.11075199842453003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,8,power_law_1.2,0.07070720195770264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,16,power_law_1.01,0.9959808349609375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,2,8,power_law_1.01,2.1399808883666993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,8,power_law_1.2,0.07644799947738648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,64,power_law_1.2,0.1262719988822937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,16,power_law_1.01,1.4965888023376466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,8,power_law_1.2,0.0784063994884491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,64,power_law_1.2,0.15807360410690308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,8,power_law_1.2,0.08019199967384338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,8,power_law_1.2,0.0927295982837677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,64,power_law_1.2,0.18513920307159423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,8,power_law_1.2,0.10593279600143432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,16,power_law_1.01,2.8397247314453127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,64,power_law_1.2,0.2431999921798706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,8,power_law_1.2,0.13583359718322754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,8,power_law_1.2,0.13814400434494017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,2,8,power_law_1.01,4.526208114624024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,64,power_law_1.2,0.26876800060272216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,8,power_law_1.2,0.17380479574203492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,8,power_law_1.2,0.2274240016937256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,8,power_law_1.2,0.3369472026824951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,64,power_law_1.2,0.3592447996139526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,8,power_law_1.2,0.4393343925476074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,8,power_law_1.2,0.610649585723877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,64,power_law_1.2,0.4802559852600098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,8,power_law_1.2,0.8064767837524414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,64,power_law_1.2,0.669152021408081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,8,power_law_1.2,1.1064703941345215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,8,power_law_1.2,1.612224006652832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,4,8,power_law_1.2,0.026502400636672974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,4,8,power_law_1.2,0.023039999604225158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,64,power_law_1.2,0.9664640426635742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,4,8,power_law_1.2,0.02494720071554184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,4,8,power_law_1.2,0.023129600286483764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,1,balanced,0.054330666859944664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,1,balanced,0.10247466961542766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,1,balanced,0.1186346709728241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,1,balanced,0.20218133926391602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,4,8,power_law_1.2,0.026015999913215637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,1,balanced,0.37090134620666504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,1,balanced,0.37644267082214355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,1,balanced,0.3781760136286418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,1,balanced,0.3767626682917277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,1,balanced,0.38064531485239667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,4,8,power_law_1.2,0.02728320062160492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,1,balanced,0.38176000118255615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,1,balanced,0.38176532586415607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,1,balanced,0.3860960006713867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,1,balanced,0.38731201489766437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,8,power_law_1.2,3.6630783081054688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,1,balanced,0.3952159881591797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,1,balanced,0.4032906691233317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,4,8,power_law_1.2,0.027020800113677978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,1,balanced,0.40909866491953534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,1,balanced,0.4235626856486003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,1,balanced,0.44536534945170086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,1,balanced,0.48727468649546307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,1,balanced,0.5291093190511068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,4,8,power_law_1.2,0.02958720028400421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,1,balanced,0.5691306591033936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,1,balanced,0.6910773118336996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,1,balanced,0.7674132982889811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,2,64,power_law_1.2,1.0449407577514649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,4,8,power_law_1.2,0.031327998638153075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,1,balanced,1.044975996017456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,1,balanced,1.221461296081543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,4,8,power_law_1.2,0.03190400004386902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,1,balanced,1.7435626983642578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,4,8,power_law_1.2,0.031788799166679385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,1,balanced,2.262159983317057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,4,8,power_law_1.2,0.032518398761749265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,1,balanced,2.8065760930379233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,4,8,power_law_1.2,0.03406080007553101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,1,balanced,4.355850537618001
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,4,8,power_law_1.2,0.03456639945507049
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,4,8,power_law_1.2,0.03835520148277283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,4,8,power_law_1.2,0.04033919870853424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,1,balanced,8.303738911946615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,4,8,power_law_1.2,0.04356479942798615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,2,64,power_law_1.2,1.7913536071777343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,4,8,power_law_1.2,0.0496832013130188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,8,power_law_1.2,0.05429760217666626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,8,power_law_1.2,0.06179839968681335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,8,power_law_1.2,0.07041919827461243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,8,power_law_1.2,0.0907584011554718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,8,power_law_1.2,0.09946879744529724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,8,power_law_1.2,0.14401919841766359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,8,power_law_1.2,0.2111423969268799
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,4,balanced,0.047397335370381675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,4,balanced,0.04693333307902018
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,4,balanced,0.048997332652409874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,4,balanced,0.05654400090376536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,4,balanced,0.0825493335723877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,4,balanced,0.08916800220807393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,4,balanced,0.0904960036277771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,4,balanced,0.09307199716567993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,8,power_law_1.2,0.280947208404541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,4,balanced,0.09148266911506653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,4,balanced,0.09353599945704143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,4,balanced,0.09273599584897359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,4,balanced,0.09452799956003825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,4,balanced,0.09685867031415303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,4,balanced,0.09911466638247173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,4,balanced,0.10361599922180176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,4,balanced,0.10683199763298035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,4,balanced,0.1111893355846405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,4,balanced,0.11921599507331848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,4,balanced,0.12797333796819052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,8,power_law_1.2,0.2915519952774048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,4,balanced,0.16268799702326456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,4,balanced,0.1775253415107727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,4,balanced,0.23903467257817587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,4,balanced,0.2622293432553609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,4,balanced,0.38725332419077557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,2,64,power_law_1.2,4.220512008666992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,4,balanced,0.4538346529006958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,4,balanced,0.6169759829839071
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,8,power_law_1.2,0.43457918167114257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,4,balanced,0.7988320191701254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,4,balanced,0.9646826585133871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,4,balanced,1.4780853589375813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,4,balanced,2.8455947240193686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,8,power_law_1.2,0.7007232189178467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,2,power_law_1.01,0.0695743978023529
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,2,power_law_1.01,0.11282559633255004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,8,power_law_1.2,1.5734463691711427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,2,power_law_1.01,0.07925760149955749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,2,power_law_1.01,0.11323519945144653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,2,power_law_1.01,0.17537280321121215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,2,power_law_1.01,0.2286207914352417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,2,power_law_1.01,0.31178879737854004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,2,power_law_1.01,0.34653439521789553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,2,power_law_1.01,0.3514496088027954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,2,power_law_1.01,0.3563199996948242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,2,power_law_1.01,0.37957119941711426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,32,8,power_law_1.2,0.05312640070915222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,2,power_law_1.01,0.38499839305877687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,1,128,power_law_1.01,0.03468799889087677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,32,8,power_law_1.2,0.05304960012435913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,2,power_law_1.01,0.3932096004486084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,1,128,power_law_1.01,0.03439359962940216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,2,power_law_1.01,0.4138751983642578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,1,1,balanced,0.12825066844622293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,32,8,power_law_1.2,0.04960640072822571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,1,1,balanced,0.2738719979921977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,1,128,power_law_1.01,0.03377279937267304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,1,1,balanced,0.3808906475702922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,1,1,balanced,0.3847786585489909
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,1,1,balanced,0.38494400183359784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,1,1,balanced,0.3896853526433309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,2,power_law_1.01,0.44632320404052733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,1,1,balanced,0.39112532138824463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,1,1,balanced,0.3911466598510742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,1,1,balanced,0.39238401254018146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,1,1,balanced,0.3946666717529297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,1,1,balanced,0.40278931458791095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,1,1,balanced,0.4076213439305623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,32,8,power_law_1.2,0.04878079891204834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,1,128,power_law_1.01,0.03392640054225922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,1,1,balanced,0.4109813372294108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,1,1,balanced,0.43837865193684894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,1,1,balanced,0.4382400115331014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,2,power_law_1.01,0.4275775909423828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,1,1,balanced,0.4437813361485799
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,1,1,balanced,0.47715731461842853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,1,1,balanced,0.5786453485488892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,1,1,balanced,0.5811893145243326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,1,1,balanced,0.7605600357055664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,1,128,power_law_1.01,0.03470720052719116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,1,1,balanced,0.8235413233439127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,32,8,power_law_1.2,0.054611200094223024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,2,power_law_1.01,0.459443187713623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,1,1,balanced,1.1137279669443767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,1,1,balanced,1.44760529200236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,1,128,power_law_1.01,0.03455359935760498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,1,1,balanced,2.111413319905599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,2,power_law_1.01,0.5024640083312988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,32,8,power_law_1.2,0.05985280275344849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,1,1,balanced,2.7732693354288735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,1,128,power_law_1.01,0.035155200958251955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,2,power_law_1.01,0.5233407974243164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,1,1,balanced,4.10260804494222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,1,128,power_law_1.01,0.03543039858341217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,32,8,power_law_1.2,0.06302080154418946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,1,1,balanced,5.495168050130208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,2,power_law_1.01,0.6209152221679688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,1,128,power_law_1.01,0.035718399286270144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,32,8,power_law_1.2,0.06716160178184509
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,2,power_law_1.01,0.643782377243042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,1,1,balanced,6.846778869628906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,1,128,power_law_1.01,0.03566080033779144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,1,128,power_law_1.01,0.03592320084571839
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,32,8,power_law_1.2,0.06925439834594727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,2,power_law_1.01,0.7883903980255127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,1,128,power_law_1.01,0.0366784006357193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,1,1,balanced,10.929285685221354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,2,power_law_1.01,0.9428223609924317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,32,8,power_law_1.2,0.06928640007972717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,1,128,power_law_1.01,0.03609600067138672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,1,128,power_law_1.01,0.03676159977912903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,2,power_law_1.01,1.2234432220458984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,32,8,power_law_1.2,0.07382400035858154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,1,128,power_law_1.01,0.038796800374984744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,32,8,power_law_1.2,0.07700480222702026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,2,power_law_1.01,1.4457216262817383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,1,128,power_law_1.01,0.0387584000825882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,1,1,balanced,22.028602600097656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,1,128,power_law_1.01,0.03947519958019256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,32,8,power_law_1.2,0.07919999957084656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,2,power_law_1.01,1.9871807098388672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,1,128,power_law_1.01,0.04085119962692261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,32,8,power_law_1.2,0.08215680122375488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,128,power_law_1.01,0.042438399791717527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,128,power_law_1.01,0.04601599872112274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,2,power_law_1.01,2.5884544372558596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,32,8,power_law_1.2,0.0913536012172699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,128,power_law_1.01,0.05030400156974792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,32,8,power_law_1.2,0.09367039799690247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,128,power_law_1.01,0.05906559824943543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,2,power_law_1.01,3.0512895584106445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,128,power_law_1.01,0.06572800278663635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,32,8,power_law_1.2,0.10152959823608398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,128,power_law_1.01,0.08186240196228027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,32,8,power_law_1.2,0.11516799926757812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,128,power_law_1.01,0.10392960309982299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,2,power_law_1.01,4.650425720214844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,32,8,power_law_1.2,0.12505600452423096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,128,power_law_1.01,0.14795520305633544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,32,8,power_law_1.2,0.1562175989151001
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,128,power_law_1.01,0.18931200504302978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,32,8,power_law_1.2,0.1729856014251709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,128,power_law_1.01,0.22777600288391114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,4,8,power_law_1.01,0.03961600065231323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,4,8,power_law_1.01,0.035231998562812804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,32,8,power_law_1.2,0.2211519956588745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,2,power_law_1.01,8.891744232177734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,4,8,power_law_1.01,0.034067198634147644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,128,power_law_1.01,0.3733887910842896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,4,8,power_law_1.01,0.03495680093765259
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,32,8,power_law_1.2,0.2950272083282471
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,4,8,power_law_1.01,0.035148799419403076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,4,8,power_law_1.01,0.03903999924659729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,4,8,power_law_1.01,0.038924801349639895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,32,8,power_law_1.2,0.3929728031158447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,128,power_law_1.01,0.629798412322998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,4,8,power_law_1.01,0.03887999951839447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,4,8,power_law_1.01,0.03984639942646027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,32,8,power_law_1.2,0.5002304077148437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,4,8,power_law_1.01,0.039724799990653994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,4,8,power_law_1.01,0.040652799606323245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,4,8,power_law_1.01,0.04099839925765991
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,4,8,power_law_1.01,0.04268800020217896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,32,8,power_law_1.2,0.674348783493042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,4,8,power_law_1.01,0.044038400053977966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,4,8,power_law_1.01,0.04764800071716309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,4,8,power_law_1.01,0.04828799962997436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,32,8,power_law_1.2,0.9170944213867187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,4,8,power_law_1.01,0.05454080104827881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,4,8,power_law_1.01,0.05802879929542541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,4,8,power_law_1.01,0.06467840075492859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,4,8,power_law_1.01,0.07530239820480347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,32,8,power_law_1.2,1.1396672248840332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,4,8,power_law_1.01,0.0806656002998352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,4,8,power_law_1.01,0.0996288001537323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,64,power_law_1.01,0.042156800627708435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,4,8,power_law_1.01,0.12851840257644653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,64,power_law_1.01,0.030476799607276915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,32,8,power_law_1.2,1.9400192260742188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,64,power_law_1.01,0.027628800272941588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,4,8,power_law_1.01,0.17262719869613646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,64,power_law_1.01,0.027001601457595826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,4,8,power_law_1.01,0.2046976089477539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,64,power_law_1.01,0.0282943993806839
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,8,balanced,0.044778664906819664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,8,balanced,0.04535466432571411
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,64,power_law_1.01,0.028115200996398925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,8,balanced,0.04698666433493296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,4,8,power_law_1.01,0.3089600086212158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,8,balanced,0.0479360024134318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,8,balanced,0.05409066875775655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,8,balanced,0.061792001128196716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,64,power_law_1.01,0.02900480031967163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,8,balanced,0.06381866832574208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,8,balanced,0.06549866497516632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,8,balanced,0.0647626668214798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,8,balanced,0.06538666784763336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,8,balanced,0.06806933383146922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,64,power_law_1.01,0.02839039862155914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,8,balanced,0.06844800213972728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,4,8,power_law_1.01,0.40660481452941893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,8,balanced,0.06906666855017345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,8,balanced,0.07506666580835979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,8,balanced,0.07970133423805237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,8,balanced,0.08669867118199666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,64,power_law_1.01,0.028697600960731505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,8,balanced,0.0920960009098053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,8,balanced,0.10380267103513081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,8,balanced,0.11506666739781697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,8,balanced,0.1349066694577535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,64,power_law_1.01,0.029241600632667543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,32,8,power_law_1.2,3.35333137512207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,8,balanced,0.15569067001342773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,8,balanced,0.20722667376200357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,4,8,power_law_1.01,0.4300992012023926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,8,balanced,0.2483146588007609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,8,balanced,0.3437173366546631
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,64,power_law_1.01,0.02945919930934906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,8,balanced,0.43395201365152997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,8,balanced,0.6223200162251791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,64,power_law_1.01,0.03000960052013397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,8,balanced,0.7941439946492513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,8,balanced,0.981498638788859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,64,power_law_1.01,0.030579200387001036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,8,balanced,1.5152106285095215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,4,8,power_law_1.01,0.7136896133422852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,64,power_law_1.01,0.03258880078792572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,8,balanced,2.9749492009480796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,64,power_law_1.01,0.03495680093765259
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,64,power_law_1.01,0.03555839955806732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,64,power_law_1.01,0.03576320111751556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,64,power_law_1.01,0.037894400954246524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,4,8,power_law_1.01,1.5542400360107422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,64,power_law_1.01,0.03837440013885498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,64,power_law_1.01,0.041222399473190306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,64,power_law_1.01,0.04357120096683502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,64,power_law_1.01,0.04785920083522797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,64,power_law_1.01,0.05557760000228882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,64,power_law_1.01,0.07531520128250122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,64,power_law_1.01,0.09662079811096191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,64,power_law_1.01,0.12480640411376953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,64,power_law_1.01,0.138099205493927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,4,64,power_law_1.01,0.16044800281524657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,4,64,power_law_1.01,0.25827200412750245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,4,power_law_1.2,0.06854400038719177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,2,power_law_1.2,0.04859519898891449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,4,power_law_1.2,0.08920959830284118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,4,64,power_law_1.01,0.5255360126495361
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,2,power_law_1.2,0.056441599130630495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,4,power_law_1.2,0.08110079765319825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,2,power_law_1.2,0.05842559933662415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,2,balanced,0.07889600098133087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,2,balanced,0.10756267110506694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,2,balanced,0.12622400124867758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,2,balanced,0.1999733249346415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,2,balanced,0.33033599456151325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,2,balanced,0.45526933670043945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,2,balanced,0.4570986827214559
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,2,power_law_1.2,0.07671679854393006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,2,balanced,0.4603840112686157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,2,balanced,0.4617813428243001
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,4,power_law_1.2,0.08698880076408386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,2,balanced,0.46484267711639404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,2,balanced,0.46695466836293537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,2,balanced,0.47140268484751385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,2,balanced,0.4768746693929036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,2,balanced,0.48044268290201825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,2,power_law_1.2,0.10045440196990967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,2,balanced,0.4914453426996867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,2,balanced,0.49723200003306073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,2,balanced,0.5096213420232137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,2,balanced,0.5464160044987997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,2,balanced,0.5701599915822347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,4,power_law_1.2,0.1047551989555359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,2,power_law_1.2,0.16733440160751342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,2,balanced,0.7437067031860352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,2,balanced,0.6982293128967285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,2,balanced,1.0085333188374836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,2,balanced,0.9542826811472574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,2,power_law_1.2,0.18154879808425903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,2,balanced,1.403706709543864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,4,power_law_1.2,0.12585599422454835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,2,balanced,1.5476800600687664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,2,power_law_1.2,0.18351999521255494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,2,balanced,2.2576427459716797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,2,balanced,2.6846933364868164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,4,power_law_1.2,0.1434048056602478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,2,2,balanced,3.46451727549235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,2,power_law_1.2,0.191539204120636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,4,power_law_1.2,0.17813119888305665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,2,2,balanced,5.23093859354655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,2,power_law_1.2,0.19137279987335204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,4,power_law_1.2,0.1816704034805298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,2,power_law_1.2,0.20159358978271485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,2,2,balanced,10.111562728881836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,2,power_law_1.2,0.2048959970474243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,4,power_law_1.2,0.1923200011253357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,2,power_law_1.2,0.20639359951019287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,4,power_law_1.2,0.20176000595092775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,32,power_law_1.01,0.07722880244255066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,2,power_law_1.2,0.21575040817260743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,4,power_law_1.2,0.20090880393981933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,2,power_law_1.2,0.21777920722961425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,32,power_law_1.01,0.07046399712562561
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,2,power_law_1.2,0.22308480739593506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,4,power_law_1.2,0.2044991970062256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,32,power_law_1.01,0.06220160126686096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,2,power_law_1.2,0.23637120723724364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,4,power_law_1.2,0.21891839504241944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,32,power_law_1.01,0.047040000557899475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,2,power_law_1.2,0.26321280002593994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,4,power_law_1.2,0.23476479053497315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,2,power_law_1.2,0.27100160121917727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,32,power_law_1.01,0.05187839865684509
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,2,power_law_1.2,0.3107327938079834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,4,power_law_1.2,0.24332160949707032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,32,power_law_1.01,0.05366399884223938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,2,power_law_1.2,0.3519615888595581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,4,power_law_1.2,0.25829119682312013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,2,power_law_1.2,0.4049856185913086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,32,power_law_1.01,0.05569919943809509
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,2,power_law_1.2,0.4398784160614014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,4,power_law_1.2,0.2857215881347656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,32,power_law_1.01,0.06295679807662964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,2,power_law_1.2,0.5464831829071045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,32,2,power_law_1.01,0.047244799137115476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,4,power_law_1.2,0.3166719913482666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,32,power_law_1.01,0.06293119788169861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,2,power_law_1.2,0.6716095924377441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,4,power_law_1.2,0.36382720470428465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,32,power_law_1.01,0.061459201574325564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,32,2,power_law_1.01,0.04997119903564453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,2,power_law_1.2,0.9594623565673828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,4,power_law_1.2,0.41741437911987306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,32,power_law_1.01,0.06348800063133239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,2,power_law_1.2,0.9727168083190918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,32,2,power_law_1.01,0.057158398628234866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,32,power_law_1.01,0.06402559876441956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,4,power_law_1.2,0.5484416007995605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,1,2,power_law_1.2,1.2038975715637208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,32,2,power_law_1.01,0.05870720148086548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,32,power_law_1.01,0.06638079881668091
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,4,power_law_1.2,0.5980735778808594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,32,2,power_law_1.01,0.07504000067710877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,1,2,power_law_1.2,1.7397056579589845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,32,power_law_1.01,0.0676479995250702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,32,2,power_law_1.01,0.08926079869270324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,4,power_law_1.2,0.7339200019836426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,32,power_law_1.01,0.07572479844093323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,32,2,power_law_1.01,0.09926400184631348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,4,power_law_1.2,0.8951104164123536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,32,power_law_1.01,0.0766207993030548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,1,2,power_law_1.2,3.422003173828125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,32,2,power_law_1.01,0.10987520217895508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,32,power_law_1.01,0.08395519852638245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,4,power_law_1.2,1.3459136009216308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,32,2,power_law_1.01,0.11676160097122193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,32,power_law_1.01,0.09181439876556396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,32,2,power_law_1.01,0.12184319496154786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,32,power_law_1.01,0.09953280091285706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,4,power_law_1.2,1.731135940551758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,32,2,power_law_1.01,0.12338559627532959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,32,power_law_1.01,0.11665279865264892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,4,power_law_1.2,2.073747253417969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,32,2,power_law_1.01,0.1355072021484375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,32,power_law_1.01,0.13070080280303956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,32,2,power_law_1.01,0.13880959749221802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,32,power_law_1.01,0.16211199760437012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,32,2,power_law_1.01,0.14892159700393676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,32,power_law_1.01,0.20764799118041993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,4,power_law_1.2,2.8534080505371096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,32,2,power_law_1.01,0.16398719549179078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,32,power_law_1.01,0.26407680511474607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,32,2,power_law_1.01,0.17185280323028565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,32,power_law_1.01,0.3248255968093872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,16,power_law_1.2,0.03873920142650604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,32,2,power_law_1.01,0.19240959882736205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,4,power_law_1.2,6.080966567993164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,32,power_law_1.01,0.4679999828338623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,16,power_law_1.2,0.03415040075778961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,32,2,power_law_1.01,0.23661439418792723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,16,power_law_1.2,0.028966400027275085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,32,power_law_1.01,0.5699520111083984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,32,2,power_law_1.01,0.23311359882354737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,16,power_law_1.2,0.031020799279212953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,32,2,power_law_1.01,0.2936575889587402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,16,power_law_1.2,0.03162879943847656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,32,power_law_1.01,0.6779391765594482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,16,power_law_1.2,0.032671999931335446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,32,2,power_law_1.01,0.3480639934539795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,16,power_law_1.2,0.033478400111198424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,16,power_law_1.2,0.0328575998544693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,32,power_law_1.01,1.0575679779052733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,32,2,power_law_1.01,0.45742077827453614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,16,power_law_1.2,0.03251200020313263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,32,2,power_law_1.01,0.5874112129211426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,16,power_law_1.2,0.03378559947013855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,16,power_law_1.2,0.03516800105571747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,32,2,power_law_1.01,0.7882431983947754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,16,power_law_1.2,0.035596799850463864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,32,power_law_1.01,1.999001693725586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,16,power_law_1.2,0.03699840009212494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,32,2,power_law_1.01,1.0427712440490722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,16,power_law_1.2,0.03741439878940582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,32,4,power_law_1.2,0.030425599217414855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,32,4,power_law_1.2,0.03079040050506592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,32,4,power_law_1.2,0.03023360073566437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,32,4,power_law_1.2,0.030028799176216127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,16,power_law_1.2,0.042419201135635375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,32,4,power_law_1.2,0.030272001028060914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,32,4,power_law_1.2,0.03086079955101013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,32,4,power_law_1.2,0.031116798520088196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,32,4,power_law_1.2,0.03187839984893799
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,16,power_law_1.2,0.042745599150657655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,32,4,power_law_1.2,0.032742398977279666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,32,4,power_law_1.2,0.032716798782348636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,32,2,power_law_1.01,1.4650815963745116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,32,4,power_law_1.2,0.0338239997625351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,32,4,power_law_1.2,0.03402239978313446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,16,power_law_1.2,0.045510399341583255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,32,4,power_law_1.2,0.03561600148677826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,32,4,power_law_1.2,0.03681919872760773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,32,4,power_law_1.2,0.04133760035037994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,32,4,power_law_1.2,0.04280959963798523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,16,power_law_1.2,0.05011199712753296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,32,4,power_law_1.2,0.04835200011730194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,32,4,power_law_1.2,0.052147197723388675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,32,4,power_law_1.2,0.06427519917488098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,32,4,power_law_1.2,0.08082559704780579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,16,power_law_1.2,0.06115840077400207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,32,4,power_law_1.2,0.08634880185127258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,32,2,power_law_1.01,1.904198455810547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,32,4,power_law_1.2,0.12367360591888428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,32,4,power_law_1.2,0.16183040142059327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,16,power_law_1.2,0.06689280271530151
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,32,4,power_law_1.2,0.22469120025634765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,32,4,power_law_1.2,0.28025600910186765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,32,4,power_law_1.2,0.41114239692687987
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,16,power_law_1.2,0.08091520071029663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,32,4,power_law_1.2,0.5349760055541992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,32,4,power_law_1.2,0.6816512107849121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,32,2,power_law_1.01,2.217638397216797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,16,power_law_1.2,0.10001280307769775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,32,4,power_law_1.2,1.055296039581299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,16,power_law_1.2,0.11194239854812622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,32,4,power_law_1.2,2.027903938293457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,16,power_law_1.2,0.16636799573898314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,32,2,power_law_1.01,3.6171329498291014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,16,power_law_1.2,0.2595711946487427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,32,8,balanced,0.041850666205088295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,32,8,balanced,0.043621331453323364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,32,8,balanced,0.041722665230433144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,32,8,balanced,0.04574400186538696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,32,8,balanced,0.051776001850763954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,32,8,balanced,0.0543039987484614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,32,8,balanced,0.054133335749308266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,32,8,balanced,0.05635733405749003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,32,8,balanced,0.05593599875768026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,32,8,balanced,0.05619200070699056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,32,8,balanced,0.056186666091283165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,32,8,balanced,0.058378666639328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,32,8,balanced,0.05794133245944977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,16,power_law_1.2,0.27987840175628664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,32,8,balanced,0.05827199916044871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,32,8,balanced,0.06277333199977875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,32,8,balanced,0.0647626668214798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,32,8,balanced,0.06646933158238728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,32,8,balanced,0.07877866427103679
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,32,8,balanced,0.0804906686147054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,32,8,balanced,0.10549867153167725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,32,8,balanced,0.11374933520952861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,32,8,balanced,0.15210666259129843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,32,8,balanced,0.16897066434224448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,32,8,balanced,0.2332800030708313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,32,8,balanced,0.2799839973449707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,32,8,balanced,0.375274658203125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,32,8,balanced,0.47173333168029785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,16,power_law_1.2,0.4019711971282959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,32,8,balanced,0.5873013337453207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,32,8,balanced,0.8663040002187093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,32,8,balanced,1.6682507197062175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,16,power_law_1.2,0.5211711883544922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,32,2,power_law_1.01,7.004640197753906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,16,power_law_1.2,1.0618751525878907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,16,power_law_1.2,1.808639907836914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,64,power_law_1.01,0.08511360287666321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,64,power_law_1.01,0.0625216007232666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,64,power_law_1.01,0.05500800013542175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,64,power_law_1.01,0.052697598934173584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,64,power_law_1.01,0.05408639907836914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,1,balanced,0.036101333796978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,1,balanced,0.03887466589609782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,1,balanced,0.04470400015513102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,1,balanced,0.06600533425807953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,1,balanced,0.09700266520182292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,1,balanced,0.09826667110125224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,1,balanced,0.10022399822870891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,1,balanced,0.10150933265686035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,1,balanced,0.10143466790517171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,1,balanced,0.10361066460609436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,1,balanced,0.10625599821408589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,64,power_law_1.01,0.05611519813537598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,1,balanced,0.10965866843859355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,1,balanced,0.11185066898663838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,1,balanced,0.11512000362078349
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,1,balanced,0.12145599722862244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,1,balanced,0.12659200032552084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,1,balanced,0.13545599579811096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,1,balanced,0.15432533621788025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,1,balanced,0.1666719913482666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,1,balanced,0.23536533117294312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,128,balanced,0.03379733363787333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,128,balanced,0.033413333197434746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,1,balanced,0.2595573266347249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,128,balanced,0.03389866650104523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,128,balanced,0.033733333150545754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,1,balanced,0.37434665362040204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,128,balanced,0.03544000039498011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,128,balanced,0.03572266548871994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,64,power_law_1.01,0.058963197469711306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,1,balanced,0.4691893259684245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,128,balanced,0.03589333345492681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,128,balanced,0.03572800010442734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,128,balanced,0.03566399961709976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,128,balanced,0.03547733277082443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,1,balanced,0.6842079957326254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,128,balanced,0.035775999228159584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,128,balanced,0.03786666691303253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,128,balanced,0.03777066618204117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,1,balanced,0.7915519873301188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,128,balanced,0.03772799919048945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,128,balanced,0.04146666576464971
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,128,balanced,0.0412266676624616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,128,balanced,0.043696001172065735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,128,balanced,0.04289066791534424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,1,balanced,1.1641440391540527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,128,balanced,0.045663997530937195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,128,balanced,0.04584533472855886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,128,balanced,0.0499839981396993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,128,balanced,0.05388799806435903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,128,balanced,0.05804799993832906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,128,balanced,0.07038400073846181
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,1,balanced,1.5077439943949382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,64,power_law_1.01,0.060140800476074216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,128,balanced,0.09262933333714803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,128,balanced,0.13110933701197305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,128,balanced,0.15960533420244852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,128,balanced,0.1893120010693868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,1,balanced,1.877072016398112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,128,balanced,0.27930132548014325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,128,balanced,0.5167839924494425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,1,balanced,2.8955891927083335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,64,power_law_1.01,0.05872640013694763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,64,balanced,0.0386613334218661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,64,balanced,0.03510933369398117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,64,power_law_1.01,0.06046720147132874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,64,balanced,0.03542399903138479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,1,balanced,5.615119934082031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,64,balanced,0.036714665591716766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,64,balanced,0.03576533248027166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,64,balanced,0.03705599904060364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,64,balanced,0.03534399966398875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,64,balanced,0.03723733375469843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,64,balanced,0.03713600089152654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,64,balanced,0.03711999952793121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,64,balanced,0.03924266745646795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,64,balanced,0.03681066632270813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,64,balanced,0.03854399919509888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,64,power_law_1.01,0.06116480231285095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,64,balanced,0.036992001036802925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,64,balanced,0.04053333401679993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,64,balanced,0.042117332418759666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,64,balanced,0.04241600135962168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,64,balanced,0.04115733255942663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,64,balanced,0.04275733232498169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,64,balanced,0.04411733150482178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,64,balanced,0.04637866715590159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,64,balanced,0.04663999875386556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,64,balanced,0.05236800014972687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,64,balanced,0.0633653352657954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,64,power_law_1.01,0.06216959953308106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,64,balanced,0.07050666709740956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,64,balanced,0.09057600299517314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,64,balanced,0.10547733306884766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,64,balanced,0.12384000420570374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,64,balanced,0.17428799470265707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,64,balanced,0.29461334149042767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,64,power_law_1.01,0.06613759994506836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,64,power_law_1.01,0.06757760047912598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,64,power_law_1.01,0.07562239766120911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,64,power_law_1.01,0.07881600260734559
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,64,power_law_1.01,0.08225280046463013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,64,power_law_1.01,0.09326080083847046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,64,power_law_1.01,0.10339200496673584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,64,power_law_1.01,0.11874560117721558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,64,power_law_1.01,0.14028160572052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,4,balanced,0.04602666695912679
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,4,balanced,0.04900800188382467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,4,balanced,0.05000533163547516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,4,balanced,0.054058666030565895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,4,balanced,0.08051733175913493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,4,balanced,0.11110400160153706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,4,balanced,0.1132319966952006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,4,balanced,0.11487467090288798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,4,balanced,0.1156213382879893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,4,balanced,0.11653866370519002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,4,balanced,0.11748266220092773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,64,power_law_1.01,0.16531200408935548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,4,balanced,0.12175466616948445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,4,balanced,0.12180800239245097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,4,balanced,0.1263040006160736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,4,balanced,0.13447466492652893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,4,balanced,0.1377066671848297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,4,balanced,0.1471999982992808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,4,balanced,0.1628320018450419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,4,balanced,0.1752799948056539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,4,balanced,0.20503999789555868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,4,balanced,0.2315733234087626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,4,balanced,0.3109333316485087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,4,balanced,0.36686933040618896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,4,balanced,0.5132266680399576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,64,power_law_1.01,0.202239990234375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,1,power_law_1.01,0.04710400104522705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,4,balanced,0.6272106568018595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,4,balanced,0.8988853295644125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,4,balanced,1.1536800066630046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,1,power_law_1.01,0.04924800097942352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,4,balanced,1.4198400179545085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,64,power_law_1.01,0.25880320072174073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,1,power_law_1.01,0.05554559826850891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,4,balanced,2.196693261464437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,16,4,power_law_1.2,0.03624320030212402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,1,power_law_1.01,0.06444799900054932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,16,4,power_law_1.2,0.03553920090198517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,4,balanced,4.235616048177083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,1,power_law_1.01,0.07776640057563781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,64,power_law_1.01,0.3244352102279663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,16,4,power_law_1.2,0.03370879888534546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,1,power_law_1.01,0.10382080078125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,16,4,power_law_1.2,0.03403519988059998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,16,4,power_law_1.2,0.03493120074272156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,1,power_law_1.01,0.1102720022201538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,64,power_law_1.01,0.46065921783447267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,16,4,power_law_1.2,0.03482879996299744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,1,power_law_1.01,0.11244159936904907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,16,4,power_law_1.2,0.03550719916820526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,16,4,power_law_1.2,0.036051198840141296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,1,power_law_1.01,0.11514879465103149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,16,4,power_law_1.2,0.03678080141544342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,1,power_law_1.01,0.11667840480804444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,16,4,power_law_1.2,0.03705599904060364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,64,power_law_1.01,0.6161407947540283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,16,4,power_law_1.2,0.03940480053424835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,1,power_law_1.01,0.1218559980392456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,16,4,power_law_1.2,0.04036479890346527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,16,4,power_law_1.2,0.04192639887332916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,1,power_law_1.01,0.12624640464782716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,16,4,power_law_1.2,0.04505600035190582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,1,power_law_1.01,0.12778240442276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,16,4,power_law_1.2,0.048416000604629514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,64,power_law_1.01,0.7040768146514893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,16,4,power_law_1.2,0.055366402864456175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,1,power_law_1.01,0.13306879997253418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,16,4,power_law_1.2,0.056704002618789676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,16,4,power_law_1.2,0.07162879705429077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,1,power_law_1.01,0.1418303966522217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,16,4,power_law_1.2,0.07729920148849487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,16,4,power_law_1.2,0.1093440055847168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,1,power_law_1.01,0.1457856059074402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,16,4,power_law_1.2,0.14009599685668944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,64,power_law_1.01,1.0987903594970703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,1,power_law_1.01,0.1575935959815979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,16,4,power_law_1.2,0.1831231951713562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,16,4,power_law_1.2,0.23244800567626953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,1,power_law_1.01,0.18216960430145263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,16,4,power_law_1.2,0.35011839866638184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,1,power_law_1.01,0.20341119766235352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,16,4,power_law_1.2,0.4037312030792236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,16,4,power_law_1.2,0.5315904140472412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,1,power_law_1.01,0.25174400806427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,16,4,power_law_1.2,0.8620223999023438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,1,power_law_1.01,0.2986687898635864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,16,4,power_law_1.2,1.0165184020996094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,1,power_law_1.01,0.4047232151031494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,64,power_law_1.01,2.230284881591797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,16,4,power_law_1.2,1.422860813140869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,1,power_law_1.01,0.5080319881439209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,1,power_law_1.01,0.7089536190032959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,16,4,power_law_1.2,2.8496383666992187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,1,power_law_1.01,0.9020352363586426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,64,power_law_1.01,0.06817920207977295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,1,power_law_1.01,1.3034560203552246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,64,power_law_1.01,0.058348798751831056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,1,power_law_1.01,1.6952384948730468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,1,power_law_1.01,2.0896703720092775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,64,power_law_1.01,0.049830400943756105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,64,power_law_1.01,0.05294079780578613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,1,power_law_1.01,3.2603073120117188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,64,power_law_1.01,0.05420799851417542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,64,power_law_1.01,0.057062399387359616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,4,power_law_1.01,0.05490559935569763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,1,power_law_1.01,6.339398574829102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,4,power_law_1.01,0.040089601278305055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,4,power_law_1.01,0.04982399940490723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,64,power_law_1.01,0.055846399068832396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,4,power_law_1.01,0.06839039921760559
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,4,power_law_1.01,0.08183680176734924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,4,power_law_1.01,0.10906239748001098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,64,power_law_1.01,0.05627520084381103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,1,power_law_1.2,0.14498560428619384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,4,power_law_1.01,0.11125760078430176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,4,power_law_1.01,0.11059199571609497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,4,power_law_1.01,0.11509759426116943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,1,power_law_1.2,0.24243199825286865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,64,power_law_1.01,0.057094401121139525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,4,power_law_1.01,0.11439360380172729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,4,power_law_1.01,0.1195520043373108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,4,power_law_1.01,0.12240639925003052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,64,power_law_1.01,0.05725439786911011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,4,power_law_1.01,0.125600004196167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,1,power_law_1.2,0.26813440322875975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,4,power_law_1.01,0.13133440017700196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,4,power_law_1.01,0.13937920331954956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,64,power_law_1.01,0.05774719715118408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,4,power_law_1.01,0.14436479806900024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,1,power_law_1.2,0.5484928131103516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,4,power_law_1.01,0.14544639587402344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,4,power_law_1.01,0.16311039924621581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,64,power_law_1.01,0.06223359704017639
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,4,power_law_1.01,0.18968960046768188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,4,power_law_1.01,0.21548159122467042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,1,power_law_1.2,0.8989567756652832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,4,power_law_1.01,0.2356544017791748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,64,power_law_1.01,0.0639743983745575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,4,power_law_1.01,0.3082880020141602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,1,power_law_1.2,1.2835328102111816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,4,power_law_1.01,0.3773247957229614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,64,power_law_1.01,0.06609280109405517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,2,power_law_1.01,0.07805439829826355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,4,power_law_1.01,0.485536003112793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,4,power_law_1.01,0.5972608089447021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,64,power_law_1.01,0.07131519913673401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,1,power_law_1.2,2.052364730834961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,4,power_law_1.01,0.8674495697021485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,2,power_law_1.01,0.10764800310134888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,64,power_law_1.01,0.07641599774360656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,1,power_law_1.2,2.1946624755859374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,4,power_law_1.01,1.1312319755554199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,2,power_law_1.01,0.07774080038070678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,64,power_law_1.01,0.08216320276260376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,4,power_law_1.01,1.403264045715332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,1,power_law_1.2,2.309254455566406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,2,power_law_1.01,0.10734080076217652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,64,power_law_1.01,0.09674879908561707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,4,power_law_1.01,1.9653440475463868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,1,power_law_1.2,2.395782470703125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,2,power_law_1.01,0.13308800458908082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,64,power_law_1.01,0.10084480047225952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,2,power_law_1.01,0.19667199850082398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,1,power_law_1.2,2.5398143768310546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,64,power_law_1.01,0.12035200595855713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,4,power_law_1.01,4.214368057250977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,2,power_law_1.01,0.21466240882873536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,1,power_law_1.2,2.6549440383911134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,64,power_law_1.01,0.14558720588684082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,2,power_law_1.01,0.22104959487915038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,1,power_law_1.2,2.6579072952270506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,64,power_law_1.01,0.1933184027671814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,2,power_law_1.01,0.22943999767303466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,1,power_law_1.2,2.7070335388183593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,2,power_law_1.01,0.23470079898834229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,64,power_law_1.01,0.1902783989906311
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,1,power_law_1.2,2.8856767654418944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,2,power_law_1.01,0.24395520687103273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,64,power_law_1.01,0.24354560375213624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,1,power_law_1.2,2.9959104537963865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,2,power_law_1.01,0.25001599788665774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,1,power_law_1.2,3.022559928894043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,64,power_law_1.01,0.28631041049957273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,2,power_law_1.01,0.2544447898864746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,16,power_law_1.2,0.05110399723052979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,1,power_law_1.2,3.3304447174072265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,2,power_law_1.01,0.26600959300994875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,16,power_law_1.2,0.03927040100097656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,64,power_law_1.01,0.3995840072631836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,16,power_law_1.2,0.03477759957313538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,2,power_law_1.01,0.2757440090179443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,16,power_law_1.2,0.037913599610328676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,1,power_law_1.2,3.5094913482666015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,16,power_law_1.2,0.03932799994945526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,16,power_law_1.2,0.05187199711799621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,2,power_law_1.01,0.286841607093811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,16,power_law_1.2,0.052262401580810545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,16,power_law_1.2,0.0531328022480011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,64,power_law_1.01,0.4733695983886719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,1,power_law_1.2,3.9450817108154297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,16,power_law_1.2,0.053600001335144046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,2,power_law_1.01,0.3027008056640625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,16,power_law_1.2,0.0548799991607666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,16,power_law_1.2,0.05597440004348755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,16,power_law_1.2,0.05502079725265503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,1,power_law_1.2,3.8244224548339845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,2,power_law_1.01,0.3499392032623291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,16,power_law_1.2,0.05688959956169128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,16,power_law_1.2,0.06238080263137817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,4,64,power_law_1.01,0.5548543930053711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,16,power_law_1.2,0.06738560199737549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,16,power_law_1.2,0.06760960221290588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,2,power_law_1.01,0.3818432092666626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,1,power_law_1.2,4.25032958984375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,16,power_law_1.2,0.07159039974212647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,16,power_law_1.2,0.0782144010066986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,16,power_law_1.2,0.08305919766426087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,2,power_law_1.01,0.4725056171417236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,16,power_law_1.2,0.10582400560379028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,1,power_law_1.2,4.5310718536376955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,16,power_law_1.2,0.10334080457687378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,16,power_law_1.2,0.13945599794387817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,4,64,power_law_1.01,0.851750373840332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,2,power_law_1.01,0.578604793548584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,16,power_law_1.2,0.1652799963951111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,16,power_law_1.2,0.22012801170349122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,1,power_law_1.2,5.447763061523437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,16,power_law_1.2,0.32778880596160886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,2,power_law_1.01,0.7344895839691162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,16,power_law_1.2,0.4738175868988037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,2,power_law_1.01,0.8179967880249024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,16,power_law_1.2,0.7137087821960449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,1,power_law_1.2,6.252735900878906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,16,power_law_1.2,0.7803008079528808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,2,power_law_1.01,1.1419712066650392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,16,power_law_1.2,1.3501503944396973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,4,64,power_law_1.01,1.7812288284301758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,1,power_law_1.2,7.9419395446777346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,2,power_law_1.01,1.3586496353149413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,16,power_law_1.2,2.612575912475586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,1,power_law_1.2,9.21154556274414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,2,power_law_1.01,1.9460351943969727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,1,power_law_1.2,10.879519653320312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,2,power_law_1.01,2.4827648162841798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,4,2,power_law_1.01,2.9368831634521486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,1,power_law_1.2,15.604197692871093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,4,power_law_1.2,0.095551997423172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,2,1,power_law_1.01,0.03333120048046112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,4,power_law_1.2,0.17918720245361328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,2,1,power_law_1.01,0.03654400110244751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,2,1,power_law_1.01,0.04379520118236542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,4,power_law_1.2,0.15219199657440186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,2,1,power_law_1.01,0.06156799793243408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,2,1,power_law_1.01,0.07604479789733887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,4,2,power_law_1.01,4.684384155273437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,2,1,power_law_1.01,0.10459519624710083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,4,power_law_1.2,0.13111679553985595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,2,1,power_law_1.01,0.10958080291748047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,2,1,power_law_1.01,0.11471999883651733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,4,power_law_1.2,0.1918720006942749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,2,1,power_law_1.01,0.11526399850845337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,2,1,power_law_1.01,0.11955840587615967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,2,1,power_law_1.01,0.12315520048141479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,4,power_law_1.2,0.26103041172027586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,2,1,power_law_1.01,0.12814079523086547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,2,1,power_law_1.01,0.12762240171432496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,1,power_law_1.2,27.844244384765624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,4,power_law_1.2,0.2901439905166626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,2,1,power_law_1.01,0.13249919414520264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,2,1,power_law_1.01,0.14092799425125122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,2,1,power_law_1.01,0.14485119581222533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,4,power_law_1.2,0.4193920135498047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,2,1,power_law_1.01,0.15048960447311402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,2,1,power_law_1.01,0.16758400201797485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,4,power_law_1.2,0.42905597686767577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,1,power_law_1.01,0.18542720079421998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,1,power_law_1.01,0.222707200050354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,1,power_law_1.01,0.24459519386291503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,4,2,power_law_1.01,9.383980560302735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,4,power_law_1.2,0.43926401138305665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,1,power_law_1.01,0.31179521083831785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,1,power_law_1.01,0.3775552034378052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,4,power_law_1.2,0.4716159820556641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,1,power_law_1.01,0.5136000156402588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,4,power_law_1.2,0.4709184169769287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,1,power_law_1.01,0.635968017578125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,1,power_law_1.01,0.8879039764404297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,4,power_law_1.2,0.5092991828918457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,1,power_law_1.01,1.1637375831604004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,4,power_law_1.2,0.5319104194641113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,1,power_law_1.01,1.4242815971374512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,4,power_law_1.2,0.5241663932800293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,4,power_law_1.2,0.5601856231689453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,1,power_law_1.01,2.1451839447021483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,32,power_law_1.2,0.05976960062980652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,4,power_law_1.2,0.5733119964599609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,32,power_law_1.2,0.04568960070610047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,64,balanced,0.0629120022058487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,64,balanced,0.0642986645301183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,4,power_law_1.2,0.6551487922668457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,64,balanced,0.06446933249632518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,64,balanced,0.06438399851322174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,64,balanced,0.06840000053246816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,32,power_law_1.2,0.045542401075363156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,64,balanced,0.09287466605504353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,64,balanced,0.11055466532707214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,64,balanced,0.10918399691581726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,64,balanced,0.1132533351580302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,64,balanced,0.11154133081436157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,64,balanced,0.11148800452550252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,64,balanced,0.12147733569145203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,1,power_law_1.01,4.194112014770508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,64,balanced,0.11341333389282227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,64,balanced,0.11691199739774068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,64,balanced,0.12688000003496805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,64,balanced,0.1276800036430359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,64,balanced,0.13783466815948486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,64,balanced,0.14197867115338644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,64,balanced,0.14723733067512512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,64,balanced,0.16755733887354532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,64,balanced,0.18922134240468344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,4,power_law_1.2,0.7095039844512939
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,64,balanced,0.2286346753438314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,32,power_law_1.2,0.0450111985206604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,64,balanced,0.26729599634806317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,64,balanced,0.34515198071797687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,64,balanced,0.4458666642506917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,64,balanced,0.5932480096817017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,64,balanced,0.7637920379638672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,64,balanced,0.9737386703491211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,32,power_law_1.2,0.045414400100708005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,64,balanced,1.4794774055480957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,4,power_law_1.2,0.8390080451965332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,64,balanced,2.83406925201416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,32,power_law_1.2,0.04725759923458099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,4,power_law_1.2,0.8422464370727539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,32,power_law_1.2,0.047654399275779726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,32,power_law_1.2,0.04840959906578064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,4,power_law_1.2,1.0302080154418944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,32,power_law_1.2,0.04839679896831513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,32,power_law_1.2,0.0515392005443573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,4,power_law_1.2,1.1491840362548829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,32,power_law_1.2,0.04935039877891541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,32,power_law_1.2,0.04983679950237274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,4,power_law_1.2,1.4643712043762207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,32,power_law_1.2,0.05090559720993042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,32,power_law_1.2,0.05311359763145447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,4,power_law_1.2,1.548844814300537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,32,power_law_1.2,0.05997440218925476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,8,balanced,0.04576000074545542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,8,balanced,0.048122664292653404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,8,balanced,0.045738667249679565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,8,balanced,0.05202666421731313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,8,balanced,0.07694399853547414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,8,balanced,0.08398933211962382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,8,balanced,0.08428800106048584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,8,balanced,0.08542399605115254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,8,balanced,0.08539199829101562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,8,balanced,0.0846506655216217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,8,balanced,0.08578667044639587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,8,balanced,0.0862559974193573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,8,balanced,0.08872532844543457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,8,balanced,0.08726933598518372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,32,power_law_1.2,0.0624064028263092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,8,balanced,0.09483200311660767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,8,balanced,0.09508267045021057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,8,balanced,0.09904000163078308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,8,balanced,0.10710400342941284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,8,balanced,0.11170666416486104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,8,balanced,0.13190399607022604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,8,balanced,0.1406613290309906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,8,balanced,0.18003199497858682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,4,power_law_1.2,2.0200063705444338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,8,balanced,0.19715199867884317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,8,balanced,0.2650986711184184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,8,balanced,0.3150186737378438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,8,balanced,0.4288586775461833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,32,power_law_1.2,0.06581760048866273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,8,balanced,0.5426400105158488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,8,balanced,0.6449973185857137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,8,balanced,0.9854666392008463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,32,power_law_1.2,0.07546240091323853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,8,balanced,1.8524959882100422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,2,2,power_law_1.01,0.06949120163917541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,2,2,power_law_1.01,0.10739200115203858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,2,2,power_law_1.01,0.09108480215072631
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,2,2,power_law_1.01,0.11279360055923462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,32,power_law_1.2,0.08721920251846313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,4,power_law_1.2,2.4467584609985353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,2,2,power_law_1.01,0.12097280025482178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,2,2,power_law_1.01,0.1220479965209961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,2,2,power_law_1.01,0.12501120567321777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,2,2,power_law_1.01,0.1289023995399475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,32,power_law_1.2,0.10760320425033569
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,2,2,power_law_1.01,0.1368191957473755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,2,2,power_law_1.01,0.1316799998283386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,2,2,power_law_1.01,0.13479039669036866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,2,2,power_law_1.01,0.1368064045906067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,2,2,power_law_1.01,0.14679039716720582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,32,power_law_1.2,0.12012799978256225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,2,2,power_law_1.01,0.15010559558868408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,2,2,power_law_1.01,0.1649216055870056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,2,2,power_law_1.01,0.17469439506530762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,2,2,power_law_1.01,0.189900803565979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,4,power_law_1.2,2.8944320678710938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,32,power_law_1.2,0.15448960065841674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,2,2,power_law_1.01,0.21675519943237304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,2,2,power_law_1.01,0.24277760982513427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,2,2,power_law_1.01,0.3023296117782593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,2,2,power_law_1.01,0.3738431930541992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,32,power_law_1.2,0.18566399812698364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,2,2,power_law_1.01,0.5305280208587646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,2,2,power_law_1.01,0.6894015789031982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,2,2,power_law_1.01,0.9952383995056152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,32,power_law_1.2,0.25708160400390623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,2,2,power_law_1.01,1.0827199935913085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,2,2,power_law_1.01,1.917203140258789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,4,power_law_1.2,4.665350341796875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,32,power_law_1.2,0.31677439212799074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,2,2,power_law_1.01,2.5677183151245115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,32,power_law_1.2,0.4744895935058594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,2,2,power_law_1.01,2.7956928253173827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,2,2,power_law_1.01,3.767238235473633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,32,power_law_1.2,0.5635519981384277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,32,power_law_1.2,0.6084671974182129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,16,power_law_1.01,0.038124799728393555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,4,power_law_1.2,8.236978912353516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,2,2,power_law_1.01,7.874150085449219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,16,power_law_1.01,0.03304960131645203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,16,power_law_1.01,0.02919679880142212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,16,power_law_1.01,0.028703999519348145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,32,power_law_1.2,1.1520575523376464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,16,power_law_1.01,0.029683199524879456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,16,power_law_1.01,0.03251200020313263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,16,power_law_1.01,0.0323199987411499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,16,power_law_1.01,0.03188480138778686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,16,power_law_1.01,0.0321727991104126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,16,power_law_1.01,0.032364800572395325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,32,power_law_1.2,2.4996416091918947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,16,power_law_1.01,0.03269760012626648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,16,power_law_1.01,0.03338879942893982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,16,power_law_1.01,0.033939200639724734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,16,power_law_1.01,0.03462400138378143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,16,power_law_1.01,0.03834879994392395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,16,power_law_1.01,0.038649600744247434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,16,power_law_1.01,0.04131200015544891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,16,power_law_1.01,0.04593920111656189
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,16,power_law_1.01,0.046598398685455324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,16,power_law_1.01,0.05601919889450073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,16,power_law_1.01,0.057792001962661745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,16,power_law_1.01,0.06743680238723755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,16,power_law_1.01,0.08095999956130981
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,2,power_law_1.2,0.051526397466659546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,16,power_law_1.01,0.09950720071792603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,64,power_law_1.2,0.040294399857521056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,2,power_law_1.2,0.05979520082473755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,16,power_law_1.01,0.11441919803619385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,64,power_law_1.2,0.03426559865474701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,2,power_law_1.2,0.06606720089912414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,16,power_law_1.01,0.14985599517822265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,64,power_law_1.2,0.035078400373458864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,2,power_law_1.2,0.08614400029182434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,64,power_law_1.2,0.03509120047092438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,16,power_law_1.01,0.21175680160522461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,2,power_law_1.2,0.1044543981552124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,64,power_law_1.2,0.03543039858341217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,2,power_law_1.2,0.15092480182647705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,4,16,power_law_1.01,0.2327552080154419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,64,power_law_1.2,0.03450239896774292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,2,power_law_1.2,0.1649791955947876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,64,power_law_1.2,0.03485440015792847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,32,4,power_law_1.2,0.045561599731445315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,64,power_law_1.2,0.03516159951686859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,2,power_law_1.2,0.16548479795455934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,4,16,power_law_1.01,0.37336320877075196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,32,4,power_law_1.2,0.03568640053272247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,64,power_law_1.2,0.035155200958251955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,32,4,power_law_1.2,0.033958399295806886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,2,power_law_1.2,0.1717568039894104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,32,4,power_law_1.2,0.033241599798202515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,64,power_law_1.2,0.03603839874267578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,2,power_law_1.2,0.18443520069122316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,32,4,power_law_1.2,0.03405439853668213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,64,power_law_1.2,0.036447998881340024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,32,4,power_law_1.2,0.03357439935207367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,2,power_law_1.2,0.18812799453735352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,32,4,power_law_1.2,0.03452160060405731
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,4,16,power_law_1.01,0.7024831771850586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,64,power_law_1.2,0.03784320056438446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,32,4,power_law_1.2,0.03456639945507049
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,2,power_law_1.2,0.18821120262145996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,64,power_law_1.2,0.0383103996515274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,32,4,power_law_1.2,0.035743999481201175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,32,4,power_law_1.2,0.035718399286270144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,64,power_law_1.2,0.039904001355171206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,2,power_law_1.2,0.197817599773407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,32,4,power_law_1.2,0.035872000455856326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,32,4,power_law_1.2,0.037529599666595456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,64,power_law_1.2,0.04366720020771027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,32,4,power_law_1.2,0.03888640105724335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,2,power_law_1.2,0.20035200119018554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,32,4,power_law_1.2,0.03944959938526153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,64,power_law_1.2,0.04598399996757507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,32,4,power_law_1.2,0.04459519982337952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,2,power_law_1.2,0.21589760780334472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,64,power_law_1.2,0.04967679977416992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,32,4,power_law_1.2,0.04708479940891266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,32,4,power_law_1.2,0.05136640071868896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,2,power_law_1.2,0.22655999660491943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,64,power_law_1.2,0.053939199447631835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,32,4,power_law_1.2,0.06976640224456787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,32,4,power_law_1.2,0.07488639950752259
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,64,power_law_1.2,0.06409599781036376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,32,4,power_law_1.2,0.0896511971950531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,2,power_law_1.2,0.22230401039123535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,32,4,power_law_1.2,0.1186560034751892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,64,power_law_1.2,0.08007680177688599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,32,4,power_law_1.2,0.15322240591049194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,2,power_law_1.2,0.2634111881256104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,32,4,power_law_1.2,0.17278720140457154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,64,power_law_1.2,0.09152640104293823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,32,4,power_law_1.2,0.1996224045753479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,2,power_law_1.2,0.2924799919128418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,64,power_law_1.2,0.11131520271301269
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,32,4,power_law_1.2,0.31305599212646484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,2,power_law_1.2,0.35757439136505126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,32,4,power_law_1.2,0.4455615997314453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,64,power_law_1.2,0.1463104009628296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,32,4,power_law_1.2,0.5985472202301025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,2,power_law_1.2,0.4315968036651611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,64,power_law_1.2,0.2591423988342285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,32,4,power_law_1.2,0.6759424209594727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,2,power_law_1.2,0.5464704036712646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,64,power_law_1.2,0.2571327924728394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,32,4,power_law_1.2,0.9962944030761719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,2,power_law_1.2,0.6593344211578369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,64,power_law_1.2,0.37039999961853026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,32,4,power_law_1.2,1.9899776458740235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,2,power_law_1.2,0.8711359977722168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,64,power_law_1.2,0.48448638916015624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,2,power_law_1.2,1.1763392448425294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,1,32,balanced,0.027386667827765148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,1,32,balanced,0.025797332326571148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,1,32,balanced,0.029109333952267964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,1,32,balanced,0.027610667049884796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,1,32,balanced,0.030026666820049286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,1,32,balanced,0.027877333263556164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,1,32,balanced,0.02762666592995326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,64,power_law_1.2,0.6318848133087158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,1,32,balanced,0.029552000264326733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,1,32,balanced,0.029850666721661884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,1,32,balanced,0.029834667841593426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,1,32,balanced,0.029493334392706554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,1,32,balanced,0.030080000559488933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,1,32,balanced,0.03159466634194056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,1,32,balanced,0.03363733241955439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,1,32,balanced,0.0372533326347669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,1,32,balanced,0.035930665830771126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,1,32,balanced,0.03793599953254064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,1,32,balanced,0.03817066550254822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,32,balanced,0.039733332892258964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,32,balanced,0.041722665230433144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,32,balanced,0.04197333256403605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,32,balanced,0.04655466477076212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,32,balanced,0.049882665276527405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,2,power_law_1.2,1.5719167709350585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,32,balanced,0.0642986645301183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,32,balanced,0.06855999926726024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,32,balanced,0.08523199955622356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,32,balanced,0.1074720025062561
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,32,balanced,0.12531200051307678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,32,balanced,0.15998400251070657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,32,balanced,0.2728959918022156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,64,power_law_1.2,1.0688960075378418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,2,power_law_1.2,1.9517248153686524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,2,power_law_1.2,2.60247688293457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,64,power_law_1.2,2.1655296325683593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,2,power_law_1.2,3.4677505493164062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,2,32,power_law_1.2,0.025574401021003723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,2,32,power_law_1.2,0.02483839988708496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,2,power_law_1.2,8.112281799316406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,2,32,power_law_1.2,0.024377599358558655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,2,32,power_law_1.2,0.024281600117683412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,2,32,power_law_1.2,0.023020799458026885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,8,4,power_law_1.01,0.040454399585723874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,2,32,power_law_1.2,0.02640640139579773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,8,4,power_law_1.01,0.03800320029258728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,2,32,power_law_1.2,0.026521599292755126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,8,4,power_law_1.01,0.03927040100097656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,2,32,power_law_1.2,0.026726400852203368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,8,4,power_law_1.01,0.04301440119743347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,8,4,power_law_1.01,0.04483200013637543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,2,32,power_law_1.2,0.029516801238059998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,8,4,power_law_1.01,0.05181440114974976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,2,32,power_law_1.2,0.02964479923248291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,8,4,power_law_1.01,0.05261440277099609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,8,4,power_law_1.01,0.05251839756965637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,2,32,power_law_1.2,0.0332863986492157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,8,4,power_law_1.01,0.05267840027809143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,256,power_law_1.2,0.06245120167732239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,8,4,power_law_1.01,0.0536575973033905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,2,32,power_law_1.2,0.03249920010566711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,8,4,power_law_1.01,0.0554751992225647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,8,4,power_law_1.01,0.05717759728431702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,256,power_law_1.2,0.06424959897994995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,2,32,power_law_1.2,0.03255679905414581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,8,4,power_law_1.01,0.058796799182891844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,8,4,power_law_1.01,0.06161919832229614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,2,32,power_law_1.2,0.03245440125465393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,8,4,power_law_1.01,0.06863999962806702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,256,power_law_1.2,0.06314240097999572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,8,4,power_law_1.01,0.07057920098304749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,2,32,power_law_1.2,0.035411199927330016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,8,4,power_law_1.01,0.07978240251541138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,1,power_law_1.2,0.03299199938774109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,8,4,power_law_1.01,0.09482880234718323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,2,32,power_law_1.2,0.03604480028152466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,256,power_law_1.2,0.062009602785110474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,1,power_law_1.2,0.038227200508117676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,8,4,power_law_1.01,0.09626880288124084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,1,power_law_1.2,0.04260480105876922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,2,32,power_law_1.2,0.03767040073871612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,8,4,power_law_1.01,0.11990400552749633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,1,power_law_1.2,0.0496832013130188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,8,4,power_law_1.01,0.1382464051246643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,256,power_law_1.2,0.06480000019073487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,2,32,power_law_1.2,0.042131200432777405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,1,power_law_1.2,0.0642624020576477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,8,4,power_law_1.01,0.1754047989845276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,1,power_law_1.2,0.08925439715385437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,32,power_law_1.2,0.04547840058803558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,8,4,power_law_1.01,0.20959360599517823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,1,power_law_1.2,0.09228799939155578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,256,power_law_1.2,0.06217600107192993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,32,power_law_1.2,0.05373439788818359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,1,power_law_1.2,0.09637119770050048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,8,4,power_law_1.01,0.29466240406036376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,1,power_law_1.2,0.10046720504760742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,32,power_law_1.2,0.05495679974555969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,8,4,power_law_1.01,0.38750081062316893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,1,power_law_1.2,0.1039423942565918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,256,power_law_1.2,0.06410880088806152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,32,power_law_1.2,0.07061120271682739
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,1,power_law_1.2,0.11067520380020142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,8,4,power_law_1.01,0.5809599876403808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,1,power_law_1.2,0.11395839452743531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,32,power_law_1.2,0.08686720132827759
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,1,power_law_1.2,0.12108800411224366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,256,power_law_1.2,0.06413440108299255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,8,4,power_law_1.01,0.7470399856567382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,32,power_law_1.2,0.12368639707565307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,1,power_law_1.2,0.1208448052406311
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,1,power_law_1.2,0.1313472032546997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,8,4,power_law_1.01,1.0041472434997558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,32,power_law_1.2,0.14642560482025146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,1,power_law_1.2,0.1386623978614807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,256,power_law_1.2,0.06407679915428162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,1,power_law_1.2,0.15075839757919313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,32,power_law_1.2,0.17011200189590453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,1,power_law_1.2,0.18318719863891603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,256,power_law_1.2,0.06468480229377746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,8,4,power_law_1.01,1.4506943702697754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,1,power_law_1.2,0.20893440246582032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,32,power_law_1.2,0.27337601184844973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,1,power_law_1.2,0.2686912059783936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,1,power_law_1.2,0.3238784074783325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,256,power_law_1.2,0.06437119841575623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,32,power_law_1.2,0.3324287891387939
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,1,power_law_1.2,0.4401408195495605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,1,power_law_1.2,0.5544256210327149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,8,4,power_law_1.01,2.8310592651367186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,256,power_law_1.2,0.06848639845848084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,1,power_law_1.2,0.7729087829589844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,32,power_law_1.2,0.6443327903747559
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,1,power_law_1.2,0.9970944404602051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,256,power_law_1.2,0.06779519915580749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,1,power_law_1.2,1.4275712013244628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,256,power_law_1.2,0.06881279945373535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,32,power_law_1.2,1.2110591888427735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,1,power_law_1.2,1.8568256378173829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,256,power_law_1.2,0.07344639897346497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,1,power_law_1.2,2.2867456436157227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,256,power_law_1.2,0.07448959946632386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,1,power_law_1.2,3.6070785522460938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,256,power_law_1.2,0.08072959780693054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,256,power_law_1.2,0.09752960205078125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,1,power_law_1.2,6.965817260742187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,256,power_law_1.2,0.1078719973564148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,16,power_law_1.2,0.0384768009185791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,16,power_law_1.2,0.031513598561286923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,16,power_law_1.2,0.02922239899635315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,256,power_law_1.2,0.13628799915313722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,16,power_law_1.2,0.02852480113506317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,16,power_law_1.2,0.03018240034580231
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,4,2,balanced,0.04240000247955322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,4,2,balanced,0.044154668847719826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,4,2,balanced,0.06890666484832764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,4,2,balanced,0.06856533388296764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,4,2,balanced,0.06770666440327962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,16,power_law_1.2,0.0318336009979248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,4,2,balanced,0.06899733344713847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,4,2,balanced,0.06829333305358887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,4,2,balanced,0.07085333267847697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,4,2,balanced,0.06861333549022675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,4,2,balanced,0.06891733407974243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,256,power_law_1.2,0.16398719549179078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,16,power_law_1.2,0.03281280100345611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,4,2,balanced,0.07046933472156525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,4,2,balanced,0.07129066685835521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,4,2,balanced,0.07216533521811168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,4,2,balanced,0.07433600227038066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,4,2,balanced,0.08113066852092743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,4,2,balanced,0.0853653351465861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,16,power_law_1.2,0.03230080008506775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,4,2,balanced,0.08410666386286418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,4,2,balanced,0.0983893374602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,4,2,balanced,0.12013866504033406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,4,2,balanced,0.1363040010134379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,4,2,balanced,0.1458613375822703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,4,2,balanced,0.1827253301938375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,16,power_law_1.2,0.0322816014289856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,4,2,balanced,0.22273600101470947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,4,2,balanced,0.2953760027885437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,4,2,balanced,0.37105600039164227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,4,2,balanced,0.5267733335494995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,16,power_law_1.2,0.03192319869995117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,256,power_law_1.2,0.2196671962738037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,4,2,balanced,0.6712266604105631
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,4,2,balanced,0.8147466977437338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,16,power_law_1.2,0.03265919983386993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,4,2,balanced,1.2720320224761963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,16,power_law_1.2,0.03386879861354828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,4,2,balanced,2.477359930674235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,16,power_law_1.2,0.03407999873161316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,16,power_law_1.2,0.03664000034332275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,256,power_law_1.2,0.24405760765075685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,16,power_law_1.2,0.03828479945659637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,16,power_law_1.2,0.03941119909286499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,16,power_law_1.2,0.04334079921245575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,256,power_law_1.2,0.31686398983001707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,16,power_law_1.2,0.046374401450157164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,16,power_law_1.2,0.0486272007226944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,16,power_law_1.2,0.05740799903869629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,16,balanced,0.03226666649182638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,16,balanced,0.03355200091997782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,16,balanced,0.03579733272393545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,16,balanced,0.05492799977461497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,16,balanced,0.07720000048478444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,16,power_law_1.2,0.06257280111312866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,16,balanced,0.07851199805736542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,16,balanced,0.07638933261235555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,16,balanced,0.07745600243409474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,16,balanced,0.0763679991165797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,16,balanced,0.07704000174999237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,16,balanced,0.07834666470686595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,16,balanced,0.07878399888674419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,16,balanced,0.0798826664686203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,16,balanced,0.0813973347345988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,16,balanced,0.08326933284600575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,256,power_law_1.2,0.43438081741333007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,16,balanced,0.08504000306129456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,16,power_law_1.2,0.07264000177383423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,16,balanced,0.08635733524958293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,16,balanced,0.0869760016600291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,16,balanced,0.08966400225957234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,16,balanced,0.0997226635615031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,16,balanced,0.10401599605878194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,16,balanced,0.11754133303960164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,16,balanced,0.13081066807111105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,16,balanced,0.16153599818547568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,16,balanced,0.19198399782180786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,16,power_law_1.2,0.08460800051689148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,16,balanced,0.23857067028681436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,16,balanced,0.2842026750246684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,16,balanced,0.33473066488901776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,16,balanced,0.4824639956156413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,16,power_law_1.2,0.10808320045471191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,16,balanced,0.8865280151367188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,16,power_law_1.2,0.14998400211334229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,256,power_law_1.2,0.6395648002624512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,16,power_law_1.2,0.2120192050933838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,16,power_law_1.2,0.23208959102630616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,4,16,power_law_1.2,0.26109440326690675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,1,balanced,0.06196799874305725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,256,power_law_1.2,0.9385791778564453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,1,balanced,0.07867733140786488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,1,balanced,0.10935466488202412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,4,16,power_law_1.2,0.5975423812866211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,1,balanced,0.15763200322786966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,1,balanced,0.2598453362782796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,1,balanced,0.4591999848683675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,1,balanced,0.6619360049565634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,1,balanced,0.6650559902191162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,1,balanced,0.6751999855041504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,1,balanced,0.6750293572743734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,1,balanced,0.6771840254465739
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,1,balanced,0.6825013160705566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,1,balanced,0.6913653214772543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,1,balanced,0.6992692947387695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,4,16,power_law_1.2,1.0202495574951171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,1,balanced,0.7110773722330729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,1,balanced,0.7187893390655518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,1,balanced,0.7454079786936442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,1,balanced,0.7861173152923584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,1,balanced,0.8256693681081136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,1,balanced,0.9000266393025717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,1,256,power_law_1.2,1.0213824272155763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,1,balanced,0.9815786679585775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,1,balanced,1.1546773115793865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,1,balanced,1.3602933883666992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,1,balanced,1.6792960166931152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,1,balanced,2.112069288889567
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,1,balanced,2.746709187825521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,1,balanced,3.743626594543457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,1,balanced,0.16884799798329672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,1,balanced,4.715162595113118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,1,balanced,0.19350934028625488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,1,balanced,0.2759946584701538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,1,balanced,0.4548106590906779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,1,balanced,0.7970560391743978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,1,balanced,1.14246400197347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,1,256,power_law_1.2,1.6875007629394532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,1,balanced,1.1524266401926677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,1,balanced,1.155951976776123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,1,balanced,6.728410720825195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,1,balanced,1.160042683283488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,1,balanced,1.1669920285542805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,1,balanced,1.1791893641153972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,1,balanced,1.1910879611968994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,1,balanced,1.2006133397420247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,1,balanced,1.219546635945638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,1,balanced,1.243557294209798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,1,balanced,1.2627253532409668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,1,balanced,1.3073279857635498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,1,balanced,12.815291086832682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,1,balanced,1.402250607808431
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,1,balanced,1.4817867279052734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,1,balanced,2.113797346750895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,1,balanced,1.8646507263183594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,1,balanced,2.6050186157226562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,1,balanced,2.7866026560465493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,1,balanced,4.119951883951823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,2,power_law_1.01,0.03175039887428284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,1,balanced,4.706362724304199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,2,power_law_1.01,0.03482879996299744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,2,power_law_1.01,0.03681280016899109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,2,power_law_1.01,0.04620159864425659
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,1,balanced,7.348688125610352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,1,256,power_law_1.2,4.0223041534423825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,2,power_law_1.01,0.05252479910850525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,2,power_law_1.01,0.06947839856147767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,2,power_law_1.01,0.07507839798927307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,1,balanced,8.499386469523111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,2,power_law_1.01,0.07495679855346679
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,2,power_law_1.01,0.07555840015411378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,2,power_law_1.01,0.0766207993030548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,1,1,balanced,10.57862408955892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,2,power_law_1.01,0.0784063994884491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,2,power_law_1.01,0.08160640001296997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,2,power_law_1.01,0.08072320222854615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,2,power_law_1.01,0.08380159735679626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,2,power_law_1.01,0.08947200179100037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,4,balanced,0.030746666093667347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,4,balanced,0.033530667424201965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,4,balanced,0.03495999922355016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,4,balanced,0.037578667203585304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,1,1,balanced,16.470272064208984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,4,balanced,0.04505600035190582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,2,power_law_1.01,0.09052159786224365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,4,balanced,0.04275733232498169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,4,balanced,0.04156800111134847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,4,balanced,0.04450666904449463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,4,balanced,0.04359466830889384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,4,balanced,0.042912001411120095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,4,balanced,0.044591998060544334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,2,power_law_1.01,0.09879680275917054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,4,balanced,0.04461866617202759
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,4,balanced,0.04523199796676636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,4,balanced,0.04691733419895172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,4,balanced,0.04938666522502899
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,4,balanced,0.05301866432030996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,4,balanced,0.05796800057093302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,4,balanced,0.06518933176994324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,2,power_law_1.01,0.10695040225982666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,4,balanced,0.07154133419195811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,4,balanced,0.08781333764394124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,4,balanced,0.10256000359853108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,4,balanced,0.12773332993189493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,4,balanced,0.15440533558527628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,2,power_law_1.01,0.11456639766693115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,4,balanced,0.20834134022394815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,4,balanced,0.254528005917867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,4,balanced,0.368010679880778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,4,balanced,0.4676053524017334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,2,power_law_1.01,0.136953604221344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,4,balanced,0.5774293343226115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,4,balanced,0.8710613250732422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,2,power_law_1.01,0.151692795753479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,4,balanced,1.6479199727376301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,2,power_law_1.01,0.19162240028381347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,2,power_law_1.01,0.21235840320587157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,2,power_law_1.01,0.2831104040145874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,1,balanced,0.0400693342089653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,1,balanced,0.058431997895240784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,1,balanced,0.08119999865690868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,1,balanced,0.1256160040696462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,1,balanced,0.20989867051442465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,1,balanced,0.21211733420689902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,1,balanced,0.21543467044830322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,1,balanced,0.2169333299001058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,1,balanced,0.21957866350809732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,1,1,balanced,32.243184407552086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,1,balanced,0.22558399041493735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,2,power_law_1.01,0.35053439140319825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,1,balanced,0.22097599506378174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,1,balanced,0.22988800207773843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,1,balanced,0.2346880038579305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,1,balanced,0.23635733127593994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,1,balanced,0.24438933531443277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,1,balanced,0.2537013292312622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,1,balanced,0.26657599210739136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,1,balanced,0.29496532678604126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,1,balanced,0.31833066542943317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,2,power_law_1.01,0.495084810256958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,1,balanced,0.396565318107605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,1,balanced,0.43852798144022626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,1,balanced,0.5882720152537028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,1,balanced,0.6843732992808024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,1,balanced,0.9709493319193522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,1,balanced,1.1937066713968914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,2,power_law_1.01,0.5948544025421143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,1,balanced,1.7220373153686523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,1,balanced,2.235328038533529
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,2,2,power_law_1.01,0.7607744216918946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,1,balanced,2.758853276570638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,1,balanced,4.2899627685546875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,2,2,power_law_1.01,1.1459967613220214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,1,balanced,8.384938557942709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,2,2,power_law_1.01,2.2636224746704103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,16,power_law_1.2,0.05431680083274841
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,16,power_law_1.2,0.07251840233802795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,16,power_law_1.2,0.057708799839019775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,256,balanced,0.0617386649052302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,256,balanced,0.0583840012550354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,256,balanced,0.05919999877611796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,256,balanced,0.059989333152770996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,256,balanced,0.06003733476003011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,256,balanced,0.06195733447869619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,256,balanced,0.0625439981619517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,256,balanced,0.06244266529877981
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,256,balanced,0.06032533446947733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,256,balanced,0.06440000236034393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,256,balanced,0.06232533355553945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,256,balanced,0.06438399851322174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,256,balanced,0.06446399788061778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,256,balanced,0.06621866424878438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,256,balanced,0.07029866675535838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,256,balanced,0.07056533296902974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,256,balanced,0.07452266911665599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,256,balanced,0.08081066608428955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,256,balanced,0.08500799536705017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,256,balanced,0.0978559950987498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,16,power_law_1.2,0.04937599897384644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,256,balanced,0.11086400349934895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,256,balanced,0.13954666256904602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,256,balanced,0.16269333163897196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,256,balanced,0.2077066699663798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,256,balanced,0.25245867172876996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,256,balanced,0.3476159969965617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,256,balanced,0.4346239964167277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,256,balanced,0.5306453307469686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,256,balanced,0.8031413555145264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,256,balanced,1.5713280042012532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,32,2,balanced,0.04831466575463613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,32,2,balanced,0.04910400013128916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,32,2,balanced,0.05639466643333435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,16,power_law_1.2,0.05185920000076294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,32,2,balanced,0.06066666543483734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,32,2,balanced,0.07812266548474629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,32,2,balanced,0.11247467001279195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,32,2,balanced,0.14196800192197165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,32,2,balanced,0.14271466930707297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,32,2,balanced,0.14538133144378662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,32,2,balanced,0.14680533607800803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,32,2,balanced,0.14714133739471436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,4,8,balanced,0.039120001097520195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,32,2,balanced,0.1529706617196401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,4,8,balanced,0.03950933367013931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,32,2,balanced,0.1539306640625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,4,8,balanced,0.03762666632731756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,32,2,balanced,0.15876266360282898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,4,8,balanced,0.037434667348861694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,32,2,balanced,0.16752000649770102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,4,8,balanced,0.03746666759252548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,4,8,balanced,0.03730666637420654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,32,2,balanced,0.1716746687889099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,4,8,balanced,0.037776000797748566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,32,2,balanced,0.1836586594581604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,4,8,balanced,0.03953066716591517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,4,8,balanced,0.03754133234421412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,32,2,balanced,0.2116853396097819
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,4,8,balanced,0.03749866783618927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,4,8,balanced,0.039077334105968475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,32,2,balanced,0.23256532351175943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,4,8,balanced,0.03967999915281931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,4,8,balanced,0.0397119993964831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,32,2,balanced,0.28039999802907306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,4,8,balanced,0.039674667020638786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,4,8,balanced,0.04577066500981649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,32,2,balanced,0.32068800926208496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,4,8,balanced,0.04570133487383524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,4,8,balanced,0.045834665497144066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,32,2,balanced,0.41118399302164715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,4,8,balanced,0.04780266682306925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,4,8,balanced,0.05189866820971171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,4,8,balanced,0.05905066430568695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,32,2,balanced,0.5267626841862997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,4,8,balanced,0.06417599817117055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,4,8,balanced,0.08120533327261607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,32,2,balanced,0.7098293304443359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,4,8,balanced,0.09125866492589314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,16,power_law_1.2,0.056704002618789676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,4,8,balanced,0.11616533001263936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,4,8,balanced,0.136245330174764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,32,2,balanced,0.9722399711608887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,4,8,balanced,0.1795733372370402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,4,8,balanced,0.21846934159596762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,4,8,balanced,0.2594933311144511
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,32,2,balanced,1.3413492838541667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,4,8,balanced,0.38925333817799884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,4,8,balanced,0.7153013547261556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,32,2,balanced,1.7822507222493489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,32,2,balanced,2.2272213300069175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,16,power_law_1.2,0.05750399827957153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,32,2,balanced,3.4306348164876304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,32,2,balanced,6.695557276407878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,16,power_law_1.2,0.06410880088806152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,16,power_law_1.2,0.0667136013507843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,16,power_law_1.2,0.06409599781036376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,16,power_law_1.2,0.0702015995979309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,64,power_law_1.01,0.03157120048999786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,16,power_law_1.2,0.07037439942359924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,64,power_law_1.01,0.030406400561332703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,64,power_law_1.01,0.02995840013027191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,64,power_law_1.01,0.030028799176216127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,16,power_law_1.2,0.074508798122406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,64,power_law_1.01,0.030611199140548707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,64,power_law_1.01,0.03116160035133362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,32,balanced,0.04980266590913137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,32,balanced,0.05007466673851013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,32,balanced,0.050848002235094704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,32,balanced,0.056074668963750206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,32,balanced,0.05974400043487549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,32,balanced,0.0620959997177124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,32,balanced,0.06213866670926412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,32,balanced,0.06198933223883311
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,64,power_law_1.01,0.03177599906921387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,32,balanced,0.06628266473611195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,32,balanced,0.06421866516272227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,32,balanced,0.061109334230422974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,32,balanced,0.0683786670366923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,32,balanced,0.06603733201821645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,16,power_law_1.2,0.07871360182762147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,32,balanced,0.06835199892520905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,32,balanced,0.07449066638946533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,32,balanced,0.08403199911117554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,32,balanced,0.08045866588751475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,64,power_law_1.01,0.03247359991073608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,32,balanced,0.09911466638247173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,32,balanced,0.10270933310190837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,32,balanced,0.13357866803805032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,32,balanced,0.1516266663869222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,32,balanced,0.2038080096244812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,32,balanced,0.19669334093729654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,64,power_law_1.01,0.03250559866428375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,32,balanced,0.25967466831207275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,32,balanced,0.310805340607961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,32,balanced,0.4195786714553833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,32,balanced,0.5114719867706299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,64,power_law_1.01,0.032716798782348636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,4,32,balanced,0.6275680065155029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,16,balanced,0.05194133520126343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,16,balanced,0.05193600058555603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,16,balanced,0.052005335688591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,16,power_law_1.2,0.08723199963569642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,16,balanced,0.05401599903901418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,16,balanced,0.05570666491985321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,4,32,balanced,0.9234453042348226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,16,balanced,0.06609599788983662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,64,power_law_1.01,0.033792001008987424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,16,balanced,0.08461866776148479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,16,balanced,0.0872320036093394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,16,balanced,0.08711999654769897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,16,balanced,0.08876799543698628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,16,balanced,0.08915199836095174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,16,balanced,0.09218666950861613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,16,balanced,0.09194133679072063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,4,32,balanced,1.9434399604797363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,16,balanced,0.0913866659005483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,16,balanced,0.09896000226338704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,64,power_law_1.01,0.03456639945507049
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,16,balanced,0.09810133775075276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,16,balanced,0.10578667124112447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,16,balanced,0.11424000064531963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,16,balanced,0.12147733569145203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,16,balanced,0.14059733351071677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,16,balanced,0.1548960010210673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,16,balanced,0.19118932882944742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,64,power_law_1.01,0.03550719916820526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,16,balanced,0.21990400552749634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,16,balanced,0.2884586652119954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,16,balanced,0.36287466684977215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,16,balanced,0.48978133996327716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,16,power_law_1.2,0.09223039746284485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,16,balanced,0.6292853355407715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,64,power_law_1.01,0.036620798707008365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,16,balanced,0.778335968653361
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,16,balanced,1.1753919919331868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,64,power_law_1.01,0.039417600631713866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,16,balanced,2.2894879976908364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,64,power_law_1.01,0.04113920032978058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,16,power_law_1.2,0.1019327998161316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,64,power_law_1.01,0.04216960072517395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,64,power_law_1.01,0.046623998880386354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,64,power_law_1.01,0.05011199712753296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,16,power_law_1.2,0.11809920072555542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,64,power_law_1.01,0.05804160237312317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,64,power_law_1.01,0.06661760210990905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,64,power_law_1.01,0.08394240140914917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,16,power_law_1.2,0.12319999933242798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,64,power_law_1.01,0.11717120409011841
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,64,power_law_1.01,0.1585536003112793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,16,power_law_1.2,0.1481727957725525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,64,power_law_1.01,0.20117759704589844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,64,power_law_1.01,0.29489281177520754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,16,power_law_1.2,0.17623679637908934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,64,power_law_1.01,0.36020479202270506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,16,power_law_1.2,0.22728960514068602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,64,power_law_1.01,0.4346816062927246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,16,balanced,0.05256533126036326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,16,balanced,0.05629866818586985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,16,balanced,0.0521066685517629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,16,balanced,0.054560000697771706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,16,balanced,0.06253866851329803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,16,balanced,0.08505066235860188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,16,balanced,0.08721066514650981
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,16,balanced,0.08901333808898926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,16,balanced,0.08704533179601033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,16,balanced,0.08813866972923279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,16,balanced,0.08905067046483357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,16,balanced,0.09086933732032776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,16,balanced,0.0911253293355306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,16,balanced,0.0930560032526652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,16,balanced,0.09726933638254802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,16,balanced,0.09876799583435059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,16,balanced,0.1035093367099762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,16,balanced,0.10980799794197083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,16,balanced,0.11869866649309795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,16,balanced,0.1309599975744883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,64,power_law_1.01,0.6651008129119873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,16,balanced,0.1530080040295919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,16,balanced,0.17889066537221274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,16,power_law_1.2,0.30094079971313475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,16,balanced,0.19925866524378458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,16,balanced,0.26523733139038086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,16,balanced,0.3113226691881816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,16,balanced,0.4367786645889282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,16,balanced,0.5407093365987142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,16,balanced,0.6633919874827067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,32,8,power_law_1.2,0.030099201202392577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,32,8,power_law_1.2,0.029465600848197937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,16,balanced,0.9947466850280762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,32,8,power_law_1.2,0.029440000653266907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,32,8,power_law_1.2,0.02874239981174469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,16,balanced,1.9360586802164714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,32,8,power_law_1.2,0.030131199955940248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,32,8,power_law_1.2,0.030348798632621764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,16,power_law_1.2,0.38725121021270753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,32,8,power_law_1.2,0.03115519881248474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,32,8,power_law_1.2,0.031455999612808226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,32,8,power_law_1.2,0.03245440125465393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,64,power_law_1.01,1.2934080123901368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,32,8,power_law_1.2,0.03251200020313263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,32,8,power_law_1.2,0.03423359990119934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,32,8,power_law_1.2,0.03553920090198517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,32,8,power_law_1.2,0.03562879860401154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,32,8,power_law_1.2,0.03685120046138764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,32,8,power_law_1.2,0.04354560077190399
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,32,8,power_law_1.2,0.044659200310707095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,32,8,power_law_1.2,0.046828800439834596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,16,power_law_1.2,0.5181759834289551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,32,8,power_law_1.2,0.05390080213546753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,32,8,power_law_1.2,0.061286401748657224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,32,8,power_law_1.2,0.0763584017753601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,32,8,power_law_1.2,0.0945088028907776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,32,8,power_law_1.2,0.14158719778060913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,32,8,power_law_1.2,0.17619839906692505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,32,8,power_law_1.2,0.2471679925918579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,32,8,power_law_1.2,0.30062079429626465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,16,power_law_1.2,0.7782527923583984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,32,8,power_law_1.2,0.44252800941467285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,32,8,power_law_1.2,0.5619711875915527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,32,8,power_law_1.2,0.7088064193725586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,32,8,power_law_1.2,1.0567487716674804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,16,power_law_1.2,0.9735039710998535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,32,8,power_law_1.2,2.0919551849365234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,16,power_law_1.2,1.1279423713684082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,1,1,balanced,0.046165332198143005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,1,1,balanced,0.06178666651248932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,1,1,balanced,0.09001599748929341
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,16,power_law_1.2,1.9557567596435548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,1,1,balanced,0.13665067156155905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,1,1,balanced,0.22217599550882974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,1,1,balanced,0.2172586719195048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,1,1,balanced,0.2191093365351359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,1,1,balanced,0.21755200624465942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,1,1,balanced,0.2197386622428894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,1,1,balanced,0.22019733985265097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,1,1,balanced,0.22462934255599976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,1,1,balanced,0.22652800877888998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,1,1,balanced,0.22721600532531738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,1,1,balanced,0.24709866444269815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,1,1,balanced,0.2504533330599467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,1,1,balanced,0.251909335454305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,1,1,balanced,0.2572373350461324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,1,1,balanced,0.2755519946416219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,1,balanced,0.28643733263015747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,1,balanced,0.31650133927663165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,1,balanced,0.34675200780232746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,1,balanced,0.4594240188598633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,1,balanced,0.5046079953511556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,1,balanced,0.6828906536102295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,1,balanced,0.870901346206665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,1,balanced,1.2616480191548665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,32,power_law_1.01,0.04928640127182007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,1,balanced,1.683578650156657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,1,balanced,2.0882719357808432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,1,balanced,3.280634562174479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,32,power_law_1.01,0.04989440143108368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,16,power_law_1.2,3.9346687316894533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,1,balanced,6.451194763183594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,32,power_law_1.01,0.0527616024017334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,32,balanced,0.035690667728583016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,32,balanced,0.029781334102153778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,32,balanced,0.029418667157491047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,32,balanced,0.029418667157491047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,32,balanced,0.03107733279466629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,32,power_law_1.01,0.046060800552368164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,32,balanced,0.03292799989382426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,32,balanced,0.0315733328461647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,32,balanced,0.033861334125200905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,32,balanced,0.03153600047032038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,32,balanced,0.031504000226656594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,32,balanced,0.03332799921433131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,32,balanced,0.03339733431736628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,32,balanced,0.03372266640265783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,32,balanced,0.033439998825391136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,32,balanced,0.0372533326347669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,32,balanced,0.03920000046491623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,32,balanced,0.03781333317359289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,32,balanced,0.03955200066169103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,32,balanced,0.039647998909155525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,32,balanced,0.04159999887148539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,32,balanced,0.043663998444875084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,32,balanced,0.04781866570313772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,32,balanced,0.054431999723116554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,32,balanced,0.06840000053246816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,32,balanced,0.08494933446248372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,32,balanced,0.10260799527168274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,32,balanced,0.12522133191426596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,32,balanced,0.15019200245539346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,32,power_law_1.01,0.047014400362968445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,32,balanced,0.19590399662653604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,32,balanced,0.3448479970296224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,32,power_law_1.01,0.04896639883518219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,32,power_law_1.01,0.05211520195007324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,32,power_law_1.01,0.05392640233039856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,32,power_law_1.01,0.05260159969329834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,32,power_law_1.01,0.05423359870910645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,8,2,power_law_1.01,0.032864001393318173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,8,2,power_law_1.01,0.033670398592948916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,8,2,power_law_1.01,0.032678401470184325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,32,power_law_1.01,0.05364480018615723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,8,2,power_law_1.01,0.034227201342582704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,8,2,power_law_1.01,0.03518719971179962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,4,power_law_1.2,0.03314560055732727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,8,2,power_law_1.01,0.038380798697471616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,8,2,power_law_1.01,0.04257920086383819
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,32,power_law_1.01,0.05683839917182922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,8,2,power_law_1.01,0.03914879858493805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,4,power_law_1.2,0.031481599807739256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,8,2,power_law_1.01,0.04386560022830963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,8,2,power_law_1.01,0.04485760033130646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,4,power_law_1.2,0.0318015992641449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,8,2,power_law_1.01,0.0449535995721817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,8,2,power_law_1.01,0.048076799511909483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,32,power_law_1.01,0.05914239883422852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,4,power_law_1.2,0.03394559919834137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,8,2,power_law_1.01,0.05428479909896851
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,4,power_law_1.2,0.0378495991230011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,16,balanced,0.0440586656332016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,16,balanced,0.04612799982229868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,8,2,power_law_1.01,0.0554751992225647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,16,balanced,0.0461706668138504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,16,balanced,0.049770668148994446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,4,power_law_1.2,0.04377599954605103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,16,balanced,0.07206933200359344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,16,balanced,0.10201600193977356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,16,balanced,0.12571199735005698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,32,power_law_1.01,0.061324799060821535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,16,balanced,0.12678933143615723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,8,2,power_law_1.01,0.0652288019657135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,16,balanced,0.12809066971143088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,16,balanced,0.12755733728408813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,16,balanced,0.12965866923332214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,4,power_law_1.2,0.04362240135669708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,16,balanced,0.1322826643784841
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,16,balanced,0.13184533516565958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,8,2,power_law_1.01,0.06872320175170898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,16,balanced,0.13409066200256348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,16,balanced,0.14105066657066345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,16,balanced,0.14257599910100302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,16,balanced,0.14921599626541138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,16,balanced,0.16293866435686746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,8,2,power_law_1.01,0.07594879865646362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,4,power_law_1.2,0.046623998880386354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,16,balanced,0.17108267545700073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,16,balanced,0.19326933224995932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,16,balanced,0.2132426699002584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,16,balanced,0.25914667050043744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,8,2,power_law_1.01,0.08106240034103393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,16,balanced,0.30403733253479004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,32,power_law_1.01,0.07041280269622803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,16,balanced,0.3946880102157593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,4,power_law_1.2,0.0469760000705719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,16,balanced,0.4803893168767293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,8,2,power_law_1.01,0.09344000220298768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,16,balanced,0.6539520025253296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,16,balanced,0.842415968577067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,4,power_law_1.2,0.04843519926071167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,8,2,power_law_1.01,0.11681280136108399
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,16,balanced,1.0364906787872314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,8,2,power_law_1.01,0.14415359497070312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,16,balanced,1.5514720280965169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,4,power_law_1.2,0.05303040146827698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,32,power_law_1.01,0.06989439725875854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,8,2,power_law_1.01,0.17138559818267823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,4,power_law_1.2,0.05478399991989136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,16,balanced,3.02622922261556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,8,2,power_law_1.01,0.22627840042114258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,8,2,power_law_1.01,0.31492478847503663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,4,power_law_1.2,0.059724801778793336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,32,power_law_1.01,0.07881600260734559
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,8,2,power_law_1.01,0.3542207956314087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,4,power_law_1.2,0.058790397644042966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,8,2,power_law_1.01,0.5837183952331543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,4,power_law_1.2,0.06369280219078063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,8,2,power_law_1.01,0.6919551849365234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,32,power_law_1.01,0.09454079866409301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,4,power_law_1.2,0.06715520024299622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,8,2,power_law_1.01,0.9144895553588868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,4,power_law_1.2,0.0718720018863678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,8,2,power_law_1.01,1.2586751937866212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,4,power_law_1.2,0.08565120100975036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,32,power_law_1.01,0.10180480480194092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,4,power_law_1.2,0.10296319723129273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,4,power_law_1.2,0.12025599479675293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,8,2,power_law_1.01,2.924947166442871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,32,power_law_1.01,0.12845439910888673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,4,power_law_1.2,0.14800000190734863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,4,power_law_1.2,0.18120959997177125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,32,power_law_1.01,0.15023360252380372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,4,power_law_1.2,0.26495358943939207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,4,power_law_1.2,0.3660736083984375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,32,power_law_1.01,0.1962496042251587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,4,power_law_1.2,0.48648319244384763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,32,power_law_1.01,0.24504320621490477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,4,power_law_1.2,0.7010176181793213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,4,power_law_1.2,0.9796223640441895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,32,power_law_1.01,0.33251841068267823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,16,2,balanced,0.031504000226656594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,16,2,balanced,0.031471999982992806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,16,2,balanced,0.03575466573238373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,16,2,balanced,0.03555733213822047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,16,2,balanced,0.03555733213822047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,16,2,balanced,0.035546667873859406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,16,2,balanced,0.03578133384386698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,16,2,balanced,0.03756266583998998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,16,2,balanced,0.03762666632731756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,16,2,balanced,0.03989866624275843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,16,2,balanced,0.0415786678592364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,16,2,balanced,0.04180799921353658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,16,2,balanced,0.0397173340121905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,16,2,balanced,0.0432640016078949
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,4,power_law_1.2,1.0617919921875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,16,2,balanced,0.055914665261904396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,16,2,balanced,0.055999999245007835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,16,2,balanced,0.058037335673967995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,16,2,balanced,0.07429333527882893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,16,2,balanced,0.08430400490760803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,16,2,balanced,0.10372266173362732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,32,power_law_1.01,0.421619176864624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,16,2,balanced,0.12798399726549783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,16,2,balanced,0.16487999757130942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,16,2,balanced,0.1891253391901652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,16,2,balanced,0.26316267251968384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,16,2,balanced,0.3435680071512858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,16,2,balanced,0.4885813395182292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,16,2,balanced,0.6243093411127726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,16,2,balanced,0.7724053064982096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,16,2,balanced,1.1967306931813557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,4,power_law_1.2,1.8089536666870116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,16,2,balanced,2.306842645009359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,32,power_law_1.01,0.6089983940124511
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,32,power_law_1.01,0.7604735851287842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,2,balanced,0.10714667042096455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,2,balanced,0.12667199969291687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,2,balanced,0.19371734062830606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,2,balanced,0.3383893171946208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,2,balanced,0.5991359949111938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,4,power_law_1.2,3.8551105499267577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,2,balanced,0.8457813262939453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,2,balanced,0.8477866649627686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,2,balanced,0.8506613572438558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,2,balanced,0.8540266354878744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,2,balanced,0.8552853266398112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,2,balanced,0.8632319768269857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,32,power_law_1.01,0.9651583671569824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,2,balanced,0.8701439698537191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,2,balanced,0.873680035273234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,2,balanced,0.8865013122558594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,2,balanced,0.8968053658803304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,2,balanced,0.900159994761149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,2,balanced,0.9119040171305338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,2,balanced,1.1200053691864014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,2,balanced,0.9811253547668457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,2,balanced,1.2366507053375244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,2,balanced,1.121840000152588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,32,power_law_1.01,1.470751953125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,2,balanced,1.8436959584554036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,2,balanced,1.3705919583638508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,2,balanced,2.0819360415140786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,2,balanced,2.07479461034139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,2,balanced,3.140618642171224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,2,balanced,3.369429270426432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,1,2,balanced,4.283178647359212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,32,power_law_1.01,2.8971839904785157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,1,balanced,0.035631999373435974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,1,2,balanced,6.2417653401692705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,1,balanced,0.0395413339138031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,1,balanced,0.04154666761557261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,1,balanced,0.054010664423306785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,1,balanced,0.08059733112653096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,1,balanced,0.08119999865690868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,1,balanced,0.08541333675384521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,1,balanced,0.08483200271924336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,1,balanced,0.08648000160853068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,1,balanced,0.08875200152397156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,1,balanced,0.09026133020718892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,1,balanced,0.09556266665458679
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,1,balanced,0.09716266393661499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,1,balanced,0.10373333096504211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,1,balanced,0.11339733004570007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,1,balanced,0.11569066842397054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,1,balanced,0.12345066666603088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,1,balanced,0.14680533607800803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,1,balanced,0.17667200167973837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,1,balanced,0.21985065937042236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,1,balanced,0.2640320062637329
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,1,balanced,0.38592533270517987
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,1,balanced,0.4785653352737427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,1,balanced,0.7010986804962158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,1,balanced,0.912992000579834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,1,balanced,1.3510826428731282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,1,2,balanced,12.35476303100586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,1,balanced,1.7596906026204426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,1,balanced,2.19651206334432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,16,4,power_law_1.01,0.03749760091304779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,16,4,power_law_1.01,0.035417601466178894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,1,balanced,3.4272588094075522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,16,4,power_law_1.01,0.03427839875221252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,16,4,power_law_1.01,0.034431999921798705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,16,4,power_law_1.01,0.03473919928073883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,16,4,power_law_1.01,0.0350847989320755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,16,4,power_law_1.01,0.0359360009431839
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,16,4,power_law_1.01,0.03573119938373566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,1,balanced,6.685754776000977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,16,4,power_law_1.01,0.03573119938373566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,16,4,power_law_1.01,0.03607679903507233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,16,4,power_law_1.01,0.036595198512077334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,16,4,power_law_1.01,0.03772799968719483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,16,4,power_law_1.01,0.03814400136470795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,16,4,power_law_1.01,0.038771200180053714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,16,4,power_law_1.01,0.04387199878692627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,16,4,power_law_1.01,0.045075199007987975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,16,4,power_law_1.01,0.04878079891204834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,16,4,power_law_1.01,0.05680000185966492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,16,4,power_law_1.01,0.0629696011543274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,16,4,power_law_1.01,0.06931840181350708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,16,4,power_law_1.01,0.07582079768180847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,16,4,power_law_1.01,0.12097920179367065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,16,4,power_law_1.01,0.1508288025856018
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,16,4,power_law_1.01,0.2226815938949585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,16,4,power_law_1.01,0.2783423900604248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,16,4,power_law_1.01,0.3196415901184082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,16,4,power_law_1.01,0.4367231845855713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,16,4,power_law_1.01,0.599398422241211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,16,4,power_law_1.01,0.7961279869079589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,16,4,power_law_1.01,1.7745983123779296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,128,power_law_1.01,0.06488320231437683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,128,power_law_1.01,0.06529920101165772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,128,power_law_1.01,0.06297600269317627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,1,16,balanced,0.025487999121348064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,1,16,balanced,0.02754666656255722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,128,power_law_1.01,0.061484801769256595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,1,16,balanced,0.02515200028816859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,1,16,balanced,0.025744001070658367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,1,16,balanced,0.029477333029111225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,1,16,balanced,0.0312266672650973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,1,16,balanced,0.03166933357715607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,1,16,balanced,0.03549866626660029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,1,16,balanced,0.03346133232116699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,1,16,balanced,0.033733333150545754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,1,16,balanced,0.03350399931271871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,1,16,balanced,0.03336533407370249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,1,16,balanced,0.034186666210492454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,1,16,balanced,0.035487999518712364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,1,16,balanced,0.037471999724706016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,1,16,balanced,0.039701332648595176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,1,16,balanced,0.041802664597829185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,1,16,balanced,0.04197333256403605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,16,balanced,0.045653333266576133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,16,balanced,0.04766400158405304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,16,balanced,0.052629331747690834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,16,balanced,0.06612800061702728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,16,balanced,0.07167999943097432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,16,balanced,0.08371200164159139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,16,balanced,0.10316800077756245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,16,balanced,0.12611200412114462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,128,power_law_1.01,0.06082559823989868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,16,balanced,0.14672000209490457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,16,balanced,0.17693867286046347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,16,balanced,0.2427306572596232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,16,balanced,0.43610668182373047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,128,power_law_1.01,0.06285439729690552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,128,power_law_1.01,0.06291199922561645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,128,power_law_1.01,0.06429439783096313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,32,power_law_1.01,0.039155200123786926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,128,power_law_1.01,0.06537600159645081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,32,power_law_1.01,0.03463680148124695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,32,power_law_1.01,0.03327359855175018
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,128,power_law_1.01,0.06414719820022582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,32,power_law_1.01,0.03275519907474518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,32,power_law_1.01,0.03367680013179779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,128,power_law_1.01,0.06447359919548035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,32,power_law_1.01,0.033888000249862674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,32,power_law_1.01,0.033843201398849485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,32,power_law_1.01,0.03470079898834229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,128,power_law_1.01,0.06606720089912414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,32,power_law_1.01,0.034092798829078674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,32,power_law_1.01,0.03468799889087677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,128,power_law_1.01,0.06706560254096985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,32,power_law_1.01,0.03561600148677826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,32,power_law_1.01,0.036313599348068236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,64,balanced,0.0425546665986379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,64,balanced,0.031258667508761086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,64,balanced,0.031354665756225586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,64,balanced,0.03139200061559677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,64,balanced,0.031386665999889374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,64,balanced,0.03159466634194056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,64,balanced,0.03133866687615713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,64,balanced,0.033200000723203026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,64,balanced,0.03142400085926056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,64,balanced,0.03310399999221166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,32,power_law_1.01,0.03699840009212494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,64,balanced,0.03344533344109853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,64,balanced,0.033413333197434746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,64,balanced,0.03364266703526179
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,128,power_law_1.01,0.06884480118751526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,64,balanced,0.03340800106525421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,64,balanced,0.03737599899371465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,64,balanced,0.03749866783618927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,64,balanced,0.03961600114901861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,64,balanced,0.03957866628964742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,64,balanced,0.04144533226887385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,64,balanced,0.04562133550643921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,32,power_law_1.01,0.03781760036945343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,64,balanced,0.04771199822425842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,64,balanced,0.05385066568851471
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,64,balanced,0.06015466650327047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,64,balanced,0.08970133463541667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,64,balanced,0.10422933101654053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,64,balanced,0.13577600320180258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,64,balanced,0.16750933726628622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,1,64,balanced,0.20014933745066324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,32,power_law_1.01,0.040780800580978396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,1,64,balanced,0.2864533265431722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,1,64,balanced,0.5229386488596598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,128,power_law_1.01,0.07372159957885742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,32,power_law_1.01,0.04140160083770752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,32,power_law_1.01,0.04264320135116577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,128,power_law_1.01,0.07448319792747497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,32,power_law_1.01,0.047539201378822324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,32,power_law_1.01,0.04888960123062134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,32,power_law_1.01,0.05451520085334778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,128,power_law_1.01,0.0787775993347168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,32,power_law_1.01,0.061452800035476686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,32,power_law_1.01,0.07243520021438599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,128,power_law_1.01,0.08620799779891967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,32,power_law_1.01,0.07970560193061829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,128,power_law_1.01,0.09526399970054626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,32,power_law_1.01,0.11534719467163086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,32,power_law_1.01,0.1367616057395935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,128,power_law_1.01,0.11498880386352539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,32,power_law_1.01,0.18698240518569947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,128,power_law_1.01,0.13642239570617676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,32,power_law_1.01,0.22873599529266359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,32,power_law_1.01,0.32020480632781984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,128,power_law_1.01,0.17310080528259278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,128,power_law_1.01,0.20839040279388427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,32,power_law_1.01,0.4596415996551514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,32,power_law_1.2,0.04476799964904785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,32,power_law_1.2,0.04202240109443665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,128,power_law_1.01,0.27669761180877683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,32,power_law_1.2,0.04170880019664765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,32,power_law_1.2,0.041171199083328246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,32,power_law_1.01,0.9086400032043457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,32,power_law_1.2,0.04291839897632599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,128,power_law_1.01,0.35025920867919924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,32,power_law_1.2,0.04429439902305603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,32,power_law_1.2,0.04527359902858734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,128,power_law_1.01,0.4866367816925049
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,32,power_law_1.2,0.0466623991727829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,32,power_law_1.2,0.04699519872665405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,32,power_law_1.2,0.04826880097389221
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,128,power_law_1.01,0.6634880065917969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,32,power_law_1.2,0.04963839948177338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,32,power_law_1.2,0.05085440278053284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,32,power_law_1.2,0.0513152003288269
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,128,power_law_1.01,0.7876927852630615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,32,power_law_1.2,0.05364480018615723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,32,power_law_1.2,0.05913599729537964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,32,power_law_1.2,0.06199679970741272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,32,power_law_1.2,0.06342399716377259
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,128,power_law_1.01,1.2591232299804687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,32,power_law_1.2,0.07138559818267823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,32,power_law_1.2,0.08526080250740051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,32,power_law_1.2,0.10497920513153076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,32,power_law_1.2,0.12766079902648925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,4,1,power_law_1.2,0.056992000341415404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,32,power_law_1.2,0.18913919925689698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,4,1,power_law_1.2,0.0667967975139618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,4,1,power_law_1.2,0.09063680171966552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,4,1,power_law_1.2,0.13036799430847168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,128,power_law_1.01,2.490559959411621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,4,1,power_law_1.2,0.1290176033973694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,32,power_law_1.2,0.21440000534057618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,4,1,power_law_1.2,0.1344704031944275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,4,1,power_law_1.2,0.13326719999313355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,4,1,power_law_1.2,0.13574399948120117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,4,1,power_law_1.2,0.14062080383300782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,32,power_law_1.2,0.31139841079711916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,4,1,power_law_1.2,0.14042240381240845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,4,1,power_law_1.2,0.1361407995223999
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,4,1,power_law_1.2,0.14232319593429565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,4,1,power_law_1.2,0.148307204246521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,32,power_law_1.2,0.4337791919708252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,4,1,power_law_1.2,0.1553663969039917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,4,1,power_law_1.2,0.17397119998931884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,4,1,power_law_1.2,0.18103679418563842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,4,1,power_law_1.2,0.17847039699554443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,4,1,power_law_1.2,0.21196799278259276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,32,power_law_1.2,0.6373695850372314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,4,1,power_law_1.2,0.2284224033355713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,4,1,power_law_1.2,0.26629760265350344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,4,1,power_law_1.2,0.31500160694122314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,4,1,power_law_1.2,0.41013121604919434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,4,1,power_law_1.2,0.497651195526123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,32,power_law_1.2,0.7998528003692627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,4,1,power_law_1.2,0.6894527912139893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,4,1,power_law_1.2,0.8823488235473633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,4,1,power_law_1.2,1.2648127555847168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,32,power_law_1.2,1.153331184387207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,4,1,power_law_1.2,1.6327104568481445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,4,1,power_law_1.2,2.035225677490234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,32,power_law_1.2,1.5879936218261719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,4,1,power_law_1.2,3.102751922607422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,4,1,power_law_1.2,6.072415924072265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,32,power_law_1.2,3.5738624572753905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,8,8,power_law_1.01,0.038867199420928956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,8,8,power_law_1.01,0.0364544004201889
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,8,8,power_law_1.01,0.037459200620651244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,8,8,power_law_1.01,0.03785600066184998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,8,8,power_law_1.01,0.03954559862613678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,8,8,power_law_1.01,0.04213759899139404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,8,8,power_law_1.01,0.04314239919185638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,8,8,power_law_1.01,0.04283519983291626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,8,8,power_law_1.01,0.04387840032577515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,8,8,power_law_1.01,0.04399999976158142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,8,8,power_law_1.01,0.04514560103416443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,8,8,power_law_1.01,0.04679040014743805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,8,8,power_law_1.01,0.047839999198913574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,8,8,power_law_1.01,0.04996480047702789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,8,8,power_law_1.01,0.055327999591827395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,8,8,power_law_1.01,0.056652802228927615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,8,8,power_law_1.01,0.06269440054893494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,8,8,power_law_1.01,0.07292159795761108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,8,8,power_law_1.01,0.07698559761047363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,32,balanced,0.05022400120894114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,32,balanced,0.03370666752258936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,32,balanced,0.03425599883000056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,32,balanced,0.03392533212900162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,32,balanced,0.03376533339420954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,32,balanced,0.03602133442958196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,32,balanced,0.03570666660865148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,32,balanced,0.035818666219711304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,32,balanced,0.03621333340803782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,32,balanced,0.03597866743803024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,8,8,power_law_1.01,0.09193599820137024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,32,balanced,0.03563733398914337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,32,balanced,0.03698666642109553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,32,balanced,0.03792533278465271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,32,balanced,0.03760000069936117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,32,2,power_law_1.2,0.02863999903202057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,32,balanced,0.039887999494870506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,32,balanced,0.04005866746107737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,32,balanced,0.042463997999827065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,32,balanced,0.050800000627835594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,32,2,power_law_1.2,0.02794240117073059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,32,balanced,0.04402133325735728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,32,balanced,0.05519466598828634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,32,balanced,0.04972266654173533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,32,2,power_law_1.2,0.02842240035533905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,8,8,power_law_1.01,0.10556800365447998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,32,balanced,0.06637333333492279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,32,balanced,0.06620799998442332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,32,2,power_law_1.2,0.02996479868888855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,32,balanced,0.0828959991534551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,32,balanced,0.08364267150561015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,32,balanced,0.10944533348083496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,32,2,power_law_1.2,0.03047040104866028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,32,balanced,0.15222932895024618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,1,32,balanced,0.17918932437896729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,32,2,power_law_1.2,0.030720001459121703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,1,32,balanced,0.2453440030415853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,8,8,power_law_1.01,0.13859200477600098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,32,2,power_law_1.2,0.03210879862308502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,1,32,balanced,0.4479946692784627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,32,2,power_law_1.2,0.0324864000082016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,32,2,power_law_1.2,0.033580800890922545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,8,8,power_law_1.01,0.17989759445190429
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,32,2,power_law_1.2,0.03383040130138397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,32,2,power_law_1.2,0.03489919900894165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,32,2,power_law_1.2,0.035769599676132205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,32,2,power_law_1.2,0.03771519958972931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,8,8,power_law_1.01,0.21279358863830566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,32,2,power_law_1.2,0.037088000774383546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,32,2,power_law_1.2,0.044268798828125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,32,2,power_law_1.2,0.04794879853725433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,8,8,power_law_1.01,0.29342079162597656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,32,2,power_law_1.2,0.05067520141601563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,32,2,power_law_1.2,0.058713597059249875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,32,2,power_law_1.2,0.06711040139198303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,32,2,power_law_1.2,0.07200639843940734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,32,2,power_law_1.2,0.09775360226631165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,8,8,power_law_1.01,0.4107967853546143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,32,2,power_law_1.2,0.13035520315170288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,32,2,power_law_1.2,0.13808640241622924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,32,2,power_law_1.2,0.2019968032836914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,32,2,power_law_1.2,0.2531775951385498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,8,8,power_law_1.01,0.5356671810150146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,32,2,power_law_1.2,0.30453760623931886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,32,2,power_law_1.2,0.4136000156402588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,32,2,power_law_1.2,0.5230847835540772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,8,8,power_law_1.01,0.6378880023956299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,32,2,power_law_1.2,0.9039744377136231
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,32,2,power_law_1.2,1.4728575706481934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,8,8,power_law_1.01,1.091808032989502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,8,8,power_law_1.01,2.042848014831543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,1,4,power_law_1.2,0.1504256010055542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,1,4,power_law_1.2,0.1446720004081726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,1,4,power_law_1.2,0.13201919794082642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,1,4,power_law_1.2,0.1335935950279236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,1,4,power_law_1.2,0.13493759632110597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,1,4,power_law_1.2,0.13841919898986815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,1,4,power_law_1.2,0.16713600158691405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,1,4,power_law_1.2,0.18105599880218506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,1,4,power_law_1.2,0.19285759925842286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,1,4,power_law_1.2,0.18492799997329712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,1,4,power_law_1.2,0.15763200521469117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,1,4,power_law_1.2,0.15930240154266356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,1,4,power_law_1.2,0.16293120384216309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,1,4,power_law_1.2,0.19544960260391236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,1,4,power_law_1.2,0.216211199760437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,1,4,power_law_1.2,0.24520959854125976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,1,4,power_law_1.2,0.24430079460144044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,1,4,power_law_1.2,0.3067199945449829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,1,4,power_law_1.2,0.37635838985443115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,1,4,power_law_1.2,0.5996287822723388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,1,4,power_law_1.2,0.6259712219238281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,1,4,power_law_1.2,0.8696576118469238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,1,4,power_law_1.2,1.0063551902770995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,1,4,power_law_1.2,1.2711359977722168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,1,4,power_law_1.2,2.104198455810547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,1,4,power_law_1.2,3.306732940673828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,4,power_law_1.2,0.035411199927330016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,4,power_law_1.2,0.030406400561332703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,4,power_law_1.2,0.03139199912548065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,1,4,power_law_1.2,4.392313766479492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,4,power_law_1.2,0.03464959859848023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,4,power_law_1.2,0.036051198840141296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,1,4,power_law_1.2,5.170105743408203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,4,power_law_1.2,0.05149440169334411
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,64,power_law_1.2,0.03447040021419525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,4,power_law_1.2,0.05399680137634277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,64,power_law_1.2,0.030214399099349976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,4,power_law_1.2,0.05428479909896851
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,64,power_law_1.2,0.029407998919486998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,4,power_law_1.2,0.055641597509384154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,1,4,power_law_1.2,7.160256195068359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,64,power_law_1.2,0.029356798529624938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,4,power_law_1.2,0.05616639852523804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,64,power_law_1.2,0.030406400561332703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,4,power_law_1.2,0.0588096022605896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,64,power_law_1.2,0.03027839958667755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,4,power_law_1.2,0.06111360192298889
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,64,power_law_1.2,0.03059839904308319
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,4,power_law_1.2,0.061894398927688596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,64,power_law_1.2,0.030272001028060914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,4,power_law_1.2,0.06542720198631287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,64,power_law_1.2,0.031020799279212953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,4,power_law_1.2,0.06999040246009827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,64,power_law_1.2,0.030982398986816408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,4,power_law_1.2,0.07534719705581665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,64,power_law_1.2,0.03154560029506683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,4,power_law_1.2,0.07847679853439331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,64,power_law_1.2,0.032307198643684386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,4,power_law_1.2,0.08299520015716552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,1,4,power_law_1.2,22.31675567626953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,64,power_law_1.2,0.03271040022373199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,4,power_law_1.2,0.10157439708709717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,64,power_law_1.2,0.03403519988059998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,4,power_law_1.2,0.12442879676818848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,64,power_law_1.2,0.03821440041065216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,4,power_law_1.2,0.1343808054924011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,64,power_law_1.2,0.03865599930286408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,4,power_law_1.2,0.1700991988182068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,64,power_law_1.2,0.039129599928855896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,4,power_law_1.2,0.22298879623413087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,64,power_law_1.2,0.042393600940704344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,4,power_law_1.2,0.3105600118637085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,64,power_law_1.2,0.044889599084854126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,64,power_law_1.2,0.050963199138641356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,4,power_law_1.2,0.38035199642181394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,64,power_law_1.2,0.055430400371551516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,4,power_law_1.2,0.5583360195159912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,64,power_law_1.2,0.07274879813194275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,64,power_law_1.2,0.0880128026008606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,4,power_law_1.2,0.797382402420044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,64,power_law_1.2,0.12329599857330323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,64,power_law_1.2,0.15909759998321532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,4,power_law_1.2,0.9137791633605957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,64,power_law_1.2,0.2097088098526001
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,64,power_law_1.2,0.3010240077972412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,4,power_law_1.2,1.569375991821289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,2,64,power_law_1.2,0.3866944074630737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,8,2,balanced,0.03759466608365377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,8,2,balanced,0.0669653316338857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,8,2,balanced,0.07781333227952321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,8,2,balanced,0.07690133154392242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,8,2,balanced,0.07671999931335449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,8,2,balanced,0.07701333363850911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,8,2,balanced,0.07806933422883351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,2,64,power_law_1.2,0.5104512214660645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,8,2,balanced,0.07643199960390727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,8,2,balanced,0.0768746683994929
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,8,2,balanced,0.07679466903209686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,8,2,balanced,0.07985066870848338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,8,2,balanced,0.08016533156236012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,8,2,balanced,0.07878933350245158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,8,2,balanced,0.08583999673525493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,8,2,balanced,0.09082133571306865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,8,2,balanced,0.09283199906349182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,8,2,balanced,0.09893866380055745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,8,2,balanced,0.10346666971842448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,8,2,balanced,0.11565333604812622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,8,2,balanced,0.13621333241462708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,8,2,balanced,0.16345066825548807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,8,2,balanced,0.20699199040730795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,4,power_law_1.2,3.799379348754883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,8,2,balanced,0.24834666649500528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,8,2,balanced,0.33347733815511066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,8,2,balanced,0.39905067284901935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,8,2,balanced,0.580677350362142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,8,2,balanced,0.7408373355865479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,8,2,balanced,0.895909309387207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,8,2,balanced,1.3871092796325684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,2,64,power_law_1.2,1.130470371246338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,8,2,balanced,2.6915200551350913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,32,4,balanced,0.029477333029111225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,32,4,balanced,0.02939733366171519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,32,4,balanced,0.0314026673634847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,32,4,balanced,0.030229332546393078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,32,4,balanced,0.031290667752424874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,32,4,balanced,0.031285333136717476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,32,4,balanced,0.03141866624355316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,32,4,balanced,0.03335466732581457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,32,4,balanced,0.03156800071398417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,32,4,balanced,0.03147733211517334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,32,4,balanced,0.03372266640265783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,32,4,balanced,0.03352533280849457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,32,4,balanced,0.035461333890755974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,32,4,balanced,0.03369600077470144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,32,4,balanced,0.03940266619126002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,32,4,balanced,0.0395413339138031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,32,4,balanced,0.041562666495641075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,32,4,balanced,0.0436106671889623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,32,4,balanced,0.047653332352638245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,32,4,balanced,0.054666668176651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,32,4,balanced,0.059903999169667564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,32,4,balanced,0.0812906672557195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,32,4,balanced,0.10318932930628459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,32,4,balanced,0.149317334095637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,32,4,balanced,0.1811306675275167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,32,4,balanced,0.25171732902526855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,32,4,balanced,0.31222933530807495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,32,4,balanced,0.383407990137736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,4,power_law_1.01,0.07660800218582153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,32,4,balanced,0.5752640167872111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,32,4,balanced,1.1039893627166748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,8,power_law_1.2,0.052095997333526614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,4,power_law_1.01,0.08913279771804809
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,4,power_law_1.01,0.0673471987247467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,8,power_law_1.2,0.05130239725112915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,4,power_law_1.01,0.08887680172920227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,8,power_law_1.2,0.04848639965057373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,4,power_law_1.01,0.1112064003944397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,8,power_law_1.2,0.05203840136528015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,4,power_law_1.01,0.1515455961227417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,4,power_law_1.01,0.15500160455703735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,8,power_law_1.2,0.05363199710845947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,4,power_law_1.01,0.15831040143966674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,8,power_law_1.2,0.058719998598098753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,4,power_law_1.01,0.16357760429382323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,4,power_law_1.01,0.16591999530792237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,8,power_law_1.2,0.06938880085945129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,4,power_law_1.01,0.1704256057739258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,8,power_law_1.2,0.07178879976272583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,4,power_law_1.01,0.16960639953613282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,8,power_law_1.2,0.07178239822387696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,4,power_law_1.01,0.17598719596862794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,4,power_law_1.01,0.18442239761352539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,8,power_law_1.2,0.07605119943618774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,4,power_law_1.01,0.19690879583358764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,8,power_law_1.2,0.07580800056457519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,4,power_law_1.01,0.1990399956703186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,8,power_law_1.2,0.07740799784660339
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,4,power_law_1.01,0.21001598834991456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,4,power_law_1.01,0.23580799102783204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,8,power_law_1.2,0.07939839959144593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,4,power_law_1.01,0.23418240547180175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,8,power_law_1.2,0.08327680230140685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,4,balanced,0.06543466448783875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,4,balanced,0.07338133454322815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,4,power_law_1.01,0.278054404258728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,4,balanced,0.08989866574605306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,4,balanced,0.12126933534940083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,4,balanced,0.1699626644452413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,4,balanced,0.2098346749941508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,4,balanced,0.21242666244506836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,4,balanced,0.21263466278711954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,4,balanced,0.21469332774480185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,8,power_law_1.2,0.08937600255012512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,4,balanced,0.2165279984474182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,4,balanced,0.22190399964650473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,4,balanced,0.22509866952896118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,4,power_law_1.01,0.2998528003692627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,4,balanced,0.2283733288447062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,4,balanced,0.23224000136057535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,4,balanced,0.23946134249369302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,4,balanced,0.2432159980138143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,4,balanced,0.2553760011990865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,4,balanced,0.28523733218510944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,4,balanced,0.3118879993756612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,8,power_law_1.2,0.0982591986656189
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,4,balanced,0.46194132169087726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,4,balanced,0.42608535289764404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,4,power_law_1.01,0.3788480043411255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,4,balanced,0.6111679871877035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,4,balanced,0.6439359982808431
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,4,balanced,0.9287359714508057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,8,power_law_1.2,0.10227199792861938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,4,balanced,1.0894400278727214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,4,power_law_1.01,0.44594559669494627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,4,balanced,1.6248052914937336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,4,balanced,1.9858986536661785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,8,power_law_1.2,0.11121920347213746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,4,power_law_1.01,0.5746880054473877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,2,4,balanced,2.5044053395589194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,8,power_law_1.2,0.12785919904708862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,4,power_law_1.01,0.7142208099365235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,2,4,balanced,3.835439999898275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,8,power_law_1.2,0.14539519548416138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,4,power_law_1.01,1.0290559768676757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,2,4,balanced,7.563568115234375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,8,power_law_1.2,0.18243839740753173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,4,power_law_1.01,1.196070384979248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,8,power_law_1.2,0.22382719516754152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,8,power_law_1.2,0.266592001914978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,4,power_law_1.01,1.4584575653076173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,8,power_law_1.2,0.3953471899032593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,4,power_law_1.01,2.294361686706543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,8,power_law_1.2,0.48456320762634275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,8,power_law_1.2,0.6572800159454346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,8,power_law_1.2,0.8128704071044922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,4,power_law_1.01,4.281484985351563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,8,power_law_1.2,1.1006848335266113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,32,4,power_law_1.01,0.03534719944000244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,8,power_law_1.2,1.4542911529541016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,32,4,power_law_1.01,0.03246079981327057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,32,4,power_law_1.01,0.03296639919281006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,32,4,power_law_1.01,0.032467201352119446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,32,4,power_law_1.01,0.033190399408340454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,32,4,power_law_1.01,0.03326080143451691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,32,4,power_law_1.01,0.033888000249862674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,32,4,power_law_1.01,0.03438720107078552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,32,4,power_law_1.01,0.03490560054779053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,32,4,power_law_1.01,0.03596799969673157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,32,4,power_law_1.01,0.035148799419403076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,32,4,power_law_1.01,0.037478399276733396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,32,4,power_law_1.01,0.03663359880447388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,32,4,power_law_1.01,0.03814400136470795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,32,4,power_law_1.01,0.042854401469230655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,32,4,power_law_1.01,0.043270400166511534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,32,4,power_law_1.01,0.047116801142692566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,32,4,power_law_1.01,0.05400959849357605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,8,power_law_1.2,3.251424026489258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,32,4,power_law_1.01,0.05849599838256836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,32,4,power_law_1.01,0.07230719923973083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,32,4,power_law_1.01,0.07663999795913697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,32,4,power_law_1.01,0.0997376024723053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,32,4,power_law_1.01,0.12407040596008301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,32,4,power_law_1.01,0.17523200511932374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,32,4,power_law_1.01,0.2030400037765503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,32,4,power_law_1.01,0.2739840030670166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,32,4,power_law_1.01,0.4646463871002197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,32,4,power_law_1.01,0.4575551986694336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,32,4,power_law_1.01,0.7845695972442627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,32,4,power_law_1.01,1.4300479888916016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,32,8,power_law_1.2,0.037248000502586365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,32,8,power_law_1.2,0.036473599076271054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,32,8,power_law_1.2,0.03505280017852783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,32,8,power_law_1.2,0.03426559865474701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,32,8,power_law_1.2,0.035334399342536925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,32,8,power_law_1.2,0.035955199599266054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,32,8,power_law_1.2,0.03642880022525787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,32,8,power_law_1.2,0.03585920035839081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,32,8,power_law_1.2,0.0364544004201889
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,32,8,power_law_1.2,0.03682560026645661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,32,8,power_law_1.2,0.036959999799728395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,32,8,power_law_1.2,0.03845120072364807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,32,8,power_law_1.2,0.03824639916419983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,32,8,power_law_1.2,0.03922559916973114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,32,8,power_law_1.2,0.043968001008033754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,32,8,power_law_1.2,0.044787201285362246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,32,8,power_law_1.2,0.04803839921951294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,32,8,power_law_1.2,0.05295360088348389
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,32,8,power_law_1.2,0.05891839861869812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,32,8,power_law_1.2,0.06821759939193725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,32,8,power_law_1.2,0.08342400193214417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,32,8,power_law_1.2,0.11052800416946411
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,32,8,power_law_1.2,0.13038719892501832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,32,8,power_law_1.2,0.1877951979637146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,32,8,power_law_1.2,0.2295232057571411
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,32,8,power_law_1.2,0.3148159980773926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,32,8,power_law_1.2,0.4061823844909668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,32,8,power_law_1.2,0.464243221282959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,32,8,power_law_1.2,0.7672192096710205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,2,2,power_law_1.2,0.04715520143508911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,2,2,power_law_1.2,0.05625600218772888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,2,2,power_law_1.2,0.06160640120506287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,32,8,power_law_1.2,1.461081600189209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,2,2,power_law_1.2,0.06618239879608154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,2,2,power_law_1.2,0.06783360242843628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,2,2,power_law_1.2,0.06944000124931335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,2,2,power_law_1.2,0.070278400182724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,2,2,power_law_1.2,0.07087360024452209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,2,2,power_law_1.2,0.07276800274848938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,2,2,power_law_1.2,0.07661439776420594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,2,2,power_law_1.2,0.07322880029678344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,2,2,power_law_1.2,0.07801600098609925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,2,2,power_law_1.2,0.0810368001461029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,2,2,power_law_1.2,0.08654080033302307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,2,2,power_law_1.2,0.09456639885902404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,2,2,power_law_1.2,0.0988864004611969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,2,2,power_law_1.2,0.11002240180969239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,2,2,power_law_1.2,0.12836480140686035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,2,2,power_law_1.2,0.14071680307388307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,2,2,power_law_1.2,0.18439040184020997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,2,2,power_law_1.2,0.2209088087081909
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,2,2,power_law_1.2,0.3072448015213013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,2,2,power_law_1.2,0.3725696086883545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,2,2,power_law_1.2,0.49509758949279786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,2,power_law_1.2,0.1442304015159607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,2,2,power_law_1.2,0.6390399932861328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,2,2,power_law_1.2,1.0004608154296875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,2,power_law_1.2,0.23797760009765626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,2,2,power_law_1.2,1.1302911758422851
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,2,2,power_law_1.2,1.5099264144897462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,2,power_law_1.2,0.23905279636383056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,2,2,power_law_1.2,2.173356819152832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,2,power_law_1.2,0.3173311948776245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,2,2,power_law_1.2,5.154143905639648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,2,power_law_1.2,0.5199168205261231
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,2,power_law_1.2,0.8369600296020507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,2,4,balanced,0.05114666620890299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,2,4,balanced,0.047728002071380615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,2,4,balanced,0.067221333583196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,2,4,balanced,0.06881066660086314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,2,4,balanced,0.06816533207893372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,2,4,balanced,0.06885333359241486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,2,4,balanced,0.06896000107129414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,2,4,balanced,0.06843733290831248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,2,4,balanced,0.07046933472156525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,2,4,balanced,0.06855466465155284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,2,4,balanced,0.07014399766921997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,2,4,balanced,0.07107733190059662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,2,4,balanced,0.07133866846561432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,2,4,balanced,0.07507733503977458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,2,4,balanced,0.08263466755549113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,2,4,balanced,0.08281066517035167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,2,4,balanced,0.08805333574612935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,2,4,balanced,0.10254933436711629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,2,4,balanced,0.1032533347606659
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,2,4,balanced,0.11771733562151591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,2,4,balanced,0.14274666706720987
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,2,power_law_1.2,0.9652095794677734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,2,4,balanced,0.17617066701253256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,2,4,balanced,0.1998080015182495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,2,4,balanced,0.26501333713531494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,2,4,balanced,0.34851733843485516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,2,4,balanced,0.4744746685028076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,2,4,balanced,0.6090293327967325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,2,4,balanced,0.7444693247477213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,2,4,balanced,1.1428639888763428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,2,power_law_1.2,1.4145407676696777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,2,4,balanced,2.208255926767985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,2,power_law_1.2,1.500710391998291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,2,power_law_1.2,1.5844927787780763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,2,power_law_1.2,1.6396543502807617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,8,balanced,0.03389333436886469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,2,power_law_1.2,1.6746496200561523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,8,balanced,0.03269333392381668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,8,balanced,0.03268266717592875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,8,balanced,0.03398400048414866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,8,balanced,0.04678399860858917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,8,balanced,0.058229332168896995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,8,balanced,0.05871466795603434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,8,balanced,0.05810666580994924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,8,balanced,0.05624533196290334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,8,balanced,0.05650666852792104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,8,balanced,0.057999998331069946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,8,balanced,0.05855466425418854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,8,balanced,0.05852800110975901
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,8,balanced,0.05905066430568695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,8,balanced,0.06251200040181477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,8,balanced,0.0639573335647583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,8,balanced,0.0631520003080368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,8,balanced,0.06859200199445088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,8,balanced,0.06622933348019917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,8,balanced,0.07473066449165344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,8,balanced,0.07629866898059845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,8,balanced,0.09070932865142822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,8,balanced,0.09345600008964539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,8,balanced,0.12849066654841104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,8,balanced,0.14176533619562784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,8,balanced,0.1913493275642395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,2,power_law_1.2,1.7740928649902343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,8,balanced,0.1978986660639445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,2,8,balanced,0.24345600605010986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,2,8,balanced,0.3372960090637207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,2,8,balanced,0.6045973300933838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,2,power_law_1.2,1.8857088088989258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,2,power_law_1.2,1.982796859741211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,16,power_law_1.2,0.04204800128936768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,16,power_law_1.2,0.03669120073318481
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,2,power_law_1.2,1.9864639282226562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,16,power_law_1.2,0.03570559918880463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,16,power_law_1.2,0.03543039858341217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,2,power_law_1.2,2.098227119445801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,16,power_law_1.2,0.03650560081005096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,16,power_law_1.2,0.038841599225997926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,16,power_law_1.2,0.03871360123157501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,2,power_law_1.2,2.2346879959106447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,16,power_law_1.2,0.03866240084171295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,16,power_law_1.2,0.03901439905166626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,16,power_law_1.2,0.03951359987258911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,2,power_law_1.2,2.4523008346557615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,16,power_law_1.2,0.040582400560379026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,16,power_law_1.2,0.04184960126876831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,16,power_law_1.2,0.04215039908885956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,2,4,power_law_1.01,0.08342400193214417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,2,4,power_law_1.01,0.07123839855194092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,2,power_law_1.2,2.737740707397461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,2,4,power_law_1.01,0.07475839853286743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,16,power_law_1.2,0.044563201069831845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,2,4,power_law_1.01,0.0794368028640747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,2,4,power_law_1.01,0.08062080144882203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,2,4,power_law_1.01,0.08161280155181885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,16,power_law_1.2,0.0486272007226944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,2,4,power_law_1.01,0.08391680121421814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,2,4,power_law_1.01,0.09964159727096558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,16,power_law_1.2,0.049267199635505673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,2,4,power_law_1.01,0.10652799606323242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,2,4,power_law_1.01,0.11257599592208863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,2,4,power_law_1.01,0.09598079919815064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,16,power_law_1.2,0.05280640125274658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,2,power_law_1.2,2.7691455841064454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,2,4,power_law_1.01,0.10421119928359986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,2,4,power_law_1.01,0.09833599925041199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,16,power_law_1.2,0.05985280275344849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,2,4,power_law_1.01,0.11300480365753174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,2,4,power_law_1.01,0.13542399406433106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,2,4,power_law_1.01,0.1493056058883667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,16,power_law_1.2,0.06611199975013733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,2,4,power_law_1.01,0.144486403465271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,2,4,power_law_1.01,0.2051392078399658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,2,4,power_law_1.01,0.1903807997703552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,16,power_law_1.2,0.07431679964065552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,2,4,power_law_1.01,0.303436803817749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,2,power_law_1.2,3.352627182006836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,2,4,power_law_1.01,0.3306495904922485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,2,4,power_law_1.01,0.39123198986053465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,16,power_law_1.2,0.09525120258331299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,2,4,power_law_1.01,0.5114496231079102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,16,power_law_1.2,0.12641279697418212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,2,4,power_law_1.01,0.7721471786499023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,2,4,power_law_1.01,0.9988608360290527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,16,power_law_1.2,0.13752959966659545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,2,power_law_1.2,3.2283775329589846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,2,4,power_law_1.01,1.5696319580078124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,16,power_law_1.2,0.21911680698394775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,2,4,power_law_1.01,2.309324836730957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,16,power_law_1.2,0.277728009223938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,2,4,power_law_1.01,2.2693695068359374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,2,power_law_1.2,3.7164417266845704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,16,power_law_1.2,0.443449592590332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,2,4,power_law_1.01,4.045971298217774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,16,power_law_1.2,0.5338431835174561
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,2,power_law_1.2,4.315795135498047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,16,power_law_1.2,0.8043904304504395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,2,4,power_law_1.01,7.532032012939453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,16,power_law_1.2,1.3229503631591797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,2,power_law_1.2,5.385862350463867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,16,power_law_1.2,2.3352256774902345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,2,power_law_1.2,6.7654975891113285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,2,power_law_1.2,7.960288238525391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,2,balanced,0.07610133290290833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,2,balanced,0.08912533521652222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,2,balanced,0.12664000193277994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,2,balanced,0.19614400466283163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,2,balanced,0.32848532994588214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,2,balanced,0.45769067605336505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,2,balanced,0.46110932032267254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,2,balanced,0.4575893481572469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,2,balanced,0.460373322168986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,2,balanced,0.46023468176523846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,2,balanced,0.4662719964981079
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,2,balanced,0.4708799918492635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,2,balanced,0.4757973353068034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,2,power_law_1.2,10.022866821289062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,2,balanced,0.4819306532541911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,2,balanced,0.49460268020629883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,2,balanced,0.5046559969584147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,2,balanced,0.5149386723836263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,2,balanced,0.5921546618143717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,2,balanced,0.5807573397954305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,2,balanced,0.7401706377665201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,2,balanced,0.7116320133209229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,2,balanced,1.0351786613464355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,2,balanced,0.9806453386942545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,2,balanced,1.4036906560262044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,2,balanced,1.5562933286031086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,1,8,balanced,0.12351999680201213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,1,8,balanced,0.19712533553441366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,1,8,balanced,0.12504532933235168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,1,8,balanced,0.12272000312805176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,1,8,balanced,0.12433600425720215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,1,8,balanced,0.12266666690508525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,2,balanced,2.3982240358988443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,1,8,balanced,0.12409599622090657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,1,8,balanced,0.1256053348382314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,1,8,balanced,0.13091733058293661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,1,8,balanced,0.132832000652949
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,1,8,balanced,0.13615999619166055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,1,8,balanced,0.13429333766301474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,1,8,balanced,0.13385599851608276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,1,8,balanced,0.14009599884351095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,1,8,balanced,0.15312000115712485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,2,balanced,2.7655038833618164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,1,8,balanced,0.15611732999483743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,1,8,balanced,0.15507200360298157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,1,8,balanced,0.1742666761080424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,1,8,balanced,0.18132267395655313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,1,8,balanced,0.20664000511169434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,1,8,balanced,0.2559573252995809
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,1,8,balanced,0.31590400139490765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,1,8,balanced,0.38838398456573486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,4,2,balanced,3.502544085184733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,1,8,balanced,0.5424319903055826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,1,8,balanced,0.6718773047129313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,1,8,balanced,0.9508852958679199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,1,8,balanced,1.2862666447957356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,4,2,balanced,5.372698465983073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,1,8,balanced,1.572389284769694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,1,8,balanced,2.5143839518229165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,1,8,balanced,4.938725471496582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,4,2,balanced,10.149386723836264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,2,power_law_1.2,19.642579650878908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,64,balanced,0.03566933423280716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,64,balanced,0.035418666899204254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,64,balanced,0.031317333380381264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,64,balanced,0.03166399896144867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,64,balanced,0.033573334415753685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,64,balanced,0.03345066557327906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,64,balanced,0.03362133353948593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,64,balanced,0.033802665770053864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,64,balanced,0.033674667278925575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,64,balanced,0.03543466577927271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,64,balanced,0.03364266703526179
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,64,balanced,0.03367999941110611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,64,balanced,0.03558400024970373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,64,balanced,0.036789332826932274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,64,balanced,0.03805333375930786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,64,balanced,0.037808001041412354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,8,power_law_1.01,0.07277439832687378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,64,balanced,0.03931200007597605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,64,balanced,0.03969600051641464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,64,balanced,0.039488000174363456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,64,balanced,0.04381333291530609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,64,balanced,0.04377600053946177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,64,balanced,0.04996266464392344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,64,balanced,0.054330666859944664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,64,balanced,0.06826133529345195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,64,balanced,0.07464533547560374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,8,power_law_1.01,0.045952001214027406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,64,balanced,0.08917333682378133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,64,balanced,0.10819733142852783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,64,balanced,0.1272160013516744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,64,balanced,0.16612799962361655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,2,power_law_1.2,0.045388799905776975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,64,balanced,0.28301332394282025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,8,power_law_1.01,0.05249919891357422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,8,power_law_1.01,0.061401599645614625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,2,power_law_1.2,0.053273600339889524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,8,power_law_1.01,0.06323840022087097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,8,power_law_1.01,0.0787775993347168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,8,power_law_1.01,0.08094720244407654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,2,power_law_1.2,0.0587007999420166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,8,power_law_1.01,0.08004480004310607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,8,power_law_1.01,0.0843392014503479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,2,power_law_1.2,0.06936320066452026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,8,power_law_1.01,0.08569599986076355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,8,power_law_1.01,0.08727040290832519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,8,power_law_1.01,0.0852288007736206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,2,power_law_1.2,0.09033600091934205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,8,power_law_1.01,0.0870848000049591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,8,power_law_1.01,0.09104639887809754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,2,power_law_1.2,0.11107200384140015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,8,power_law_1.01,0.09255679845809936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,8,power_law_1.01,0.09450240135192871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,2,power_law_1.2,0.12776319980621337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,8,power_law_1.01,0.0986624002456665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,8,power_law_1.01,0.10068479776382447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,8,power_law_1.01,0.12316800355911255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,2,power_law_1.2,0.15475840568542482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,8,power_law_1.01,0.14011520147323608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,8,power_law_1.01,0.16374399662017822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,2,power_law_1.2,0.16713600158691405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,8,power_law_1.01,0.19696639776229857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,16,balanced,0.05057600140571594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,16,balanced,0.03573333223660787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,16,balanced,0.03156800071398417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,16,balanced,0.03624533365170161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,16,balanced,0.04585599899291992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,16,balanced,0.05704533557097117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,16,balanced,0.05890666445096334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,8,power_law_1.01,0.24833281040191652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,16,balanced,0.05719466507434845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,16,balanced,0.05680533250172933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,16,balanced,0.05776533484458923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,2,power_law_1.2,0.1757696032524109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,16,balanced,0.05731200178464254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,16,balanced,0.058277333776156105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,16,balanced,0.058917333682378135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,16,balanced,0.05819199979305267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,16,balanced,0.06316799918810527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,16,balanced,0.06263466676076253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,16,balanced,0.0642080008983612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,16,balanced,0.0684799998998642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,16,balanced,0.06481066842873891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,16,balanced,0.07088000078996022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,8,power_law_1.01,0.31417601108551024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,16,balanced,0.07263466715812683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,16,balanced,0.08460799853006999
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,16,balanced,0.08574933807055156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,16,balanced,0.107424000898997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,16,balanced,0.10969600081443787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,16,balanced,0.14573333660761514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,16,balanced,0.16288533806800842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,1,16,balanced,0.19121599197387695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,1,16,balanced,0.25891733169555664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,2,power_law_1.2,0.17989120483398438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,8,power_law_1.01,0.38412160873413087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,1,16,balanced,0.4562346537907918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,8,power_law_1.01,0.5629248142242431
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,2,power_law_1.2,0.19227520227432252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,8,power_law_1.01,0.7280447959899903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,2,power_law_1.2,0.1992319941520691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,8,power_law_1.01,0.9182656288146973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,2,power_law_1.2,0.20766079425811768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,2,power_law_1.2,0.2277695894241333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,8,power_law_1.01,1.303872013092041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,2,power_law_1.2,0.2396672010421753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,2,power_law_1.2,0.26060800552368163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,8,power_law_1.01,2.7314432144165037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,2,power_law_1.2,0.30193281173706055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,2,power_law_1.2,0.32737278938293457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,4,balanced,0.047775998711586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,4,balanced,0.05356266597906748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,4,balanced,0.07653333246707916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,4,balanced,0.10408000151316325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,4,balanced,0.1534933348496755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,4,balanced,0.18296533823013306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,4,balanced,0.18267732858657837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,2,power_law_1.2,0.4007872104644775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,4,balanced,0.18513067563374838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,4,balanced,0.18703999121983847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,4,balanced,0.18665599822998047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,4,balanced,0.18749332427978516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,4,balanced,0.19087467590967813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,4,balanced,0.19156799713770548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,4,balanced,0.19407999515533447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,4,balanced,0.20288532972335815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,4,balanced,0.20376000801722208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,4,balanced,0.21236799160639444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,4,balanced,0.22693334023157755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,4,balanced,0.2360960046450297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,4,balanced,0.2890239953994751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,4,balanced,0.29499200979868573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,4,balanced,0.4127306540807088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,2,power_law_1.2,0.4634687900543213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,4,balanced,0.42476801077524823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,4,balanced,0.607040007909139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,4,balanced,0.7023839950561523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,4,balanced,1.0329333146413167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,4,balanced,1.2611253261566162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,4,balanced,1.4730079968770344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,2,power_law_1.2,0.5939648151397705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,4,balanced,2.323568026224772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,4,balanced,4.422719955444336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,2,power_law_1.2,0.706169605255127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,2,power_law_1.2,0.9498368263244629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,2,power_law_1.2,1.1941311836242676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,1,power_law_1.2,0.06395519971847534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,1,power_law_1.2,0.07095040082931518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,2,power_law_1.2,1.668191909790039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,1,power_law_1.2,0.07824000120162963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,1,power_law_1.2,0.10149760246276855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,2,power_law_1.2,2.02926082611084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,1,power_law_1.2,0.1283903956413269
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,1,power_law_1.2,0.15552639961242676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,2,power_law_1.2,2.4708415985107424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,1,power_law_1.2,0.21057920455932616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,1,power_law_1.2,0.2202303886413574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,32,power_law_1.2,0.04377599954605103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,1,power_law_1.2,0.22899200916290283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,32,power_law_1.2,0.03325439989566803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,1,power_law_1.2,0.2370368003845215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,2,power_law_1.2,3.8665409088134766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,32,power_law_1.2,0.030579200387001036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,32,power_law_1.2,0.029504001140594482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,1,power_law_1.2,0.2489919900894165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,32,power_law_1.2,0.030073601007461547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,1,power_law_1.2,0.2617088079452515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,32,power_law_1.2,0.03206399977207184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,32,power_law_1.2,0.03141759932041168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,1,power_law_1.2,0.26986238956451414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,32,power_law_1.2,0.03128960132598877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,1,power_law_1.2,0.27669761180877683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,32,power_law_1.2,0.03207040131092072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,32,power_law_1.2,0.03132160007953644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,1,power_law_1.2,0.29423999786376953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,2,power_law_1.2,7.765113830566406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,32,power_law_1.2,0.03253119885921478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,32,power_law_1.2,0.03298560082912445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,1,power_law_1.2,0.3002239942550659
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,32,power_law_1.2,0.03454720079898834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,1,power_law_1.2,0.3231359958648682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,32,power_law_1.2,0.0353983998298645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,32,power_law_1.2,0.04002560079097748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,1,power_law_1.2,0.3766527891159058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,32,power_law_1.2,0.04223999977111816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,32,power_law_1.2,0.04294399917125702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,1,power_law_1.2,0.41052799224853515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,32,power_law_1.2,0.04655359983444214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,32,power_law_1.2,0.05097600221633911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,1,power_law_1.2,0.4973120212554932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,32,power_law_1.2,0.05707520246505737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,1,power_law_1.2,0.577561616897583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,32,power_law_1.2,0.0629696011543274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,32,power_law_1.2,0.07149440050125122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,1,power_law_1.2,0.7398784160614014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,32,power_law_1.2,0.09354239702224731
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,32,power_law_1.2,0.1319808006286621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,1,power_law_1.2,0.8954496383666992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,32,power_law_1.2,0.14972800016403198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,32,power_law_1.2,0.18222719430923462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,1,power_law_1.2,1.2073408126831056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,32,power_law_1.2,0.21969280242919922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,1,power_law_1.2,1.5302463531494142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,2,32,power_law_1.2,0.2659264087677002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,32,2,power_law_1.2,0.04510720074176788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,1,power_law_1.2,2.1518144607543945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,2,32,power_law_1.2,0.4383103847503662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,32,2,power_law_1.2,0.049881601333618165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,1,power_law_1.2,2.798182487487793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,2,32,power_law_1.2,0.9665599822998047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,32,2,power_law_1.2,0.05103359818458557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,32,2,power_law_1.2,0.0578495979309082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,1,power_law_1.2,3.444166564941406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,32,2,power_law_1.2,0.0697920024394989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,32,2,power_law_1.2,0.08484479784965515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,1,power_law_1.2,5.4216960906982425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,32,2,power_law_1.2,0.09515519738197327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,32,2,power_law_1.2,0.10995199680328369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,32,2,power_law_1.2,0.11502079963684082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,32,2,power_law_1.2,0.11931519508361817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,1,power_law_1.2,10.631161499023438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,32,2,power_law_1.2,0.1221119999885559
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,8,balanced,0.046298667788505554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,8,balanced,0.04730133215586344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,8,balanced,0.04764799773693085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,8,balanced,0.04985600213209788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,8,balanced,0.05579199890295664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,8,balanced,0.07970133423805237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,8,balanced,0.08010133107503255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,8,balanced,0.0811466674009959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,8,balanced,0.0813920001188914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,8,balanced,0.0825973351796468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,8,balanced,0.08161599934101105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,8,balanced,0.08434133728345235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,8,balanced,0.08477866649627686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,32,2,power_law_1.2,0.12830079793930055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,8,balanced,0.08556800087292989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,8,balanced,0.09147733449935913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,8,balanced,0.09267200032869975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,8,balanced,0.09584533174832661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,8,balanced,0.10542933146158855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,8,balanced,0.1132480005423228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,8,balanced,0.12390399972597758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,8,balanced,0.1525173286596934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,8,balanced,0.1780853271484375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,8,balanced,0.2080693244934082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,8,balanced,0.2743733326594035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,8,balanced,0.3344800074895223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,8,balanced,0.4612693389256795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,32,2,power_law_1.2,0.13871359825134277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,8,balanced,0.584885319073995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,8,balanced,0.7371573448181152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,8,balanced,1.0892266432444255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,8,balanced,2.116373380025228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,32,2,power_law_1.2,0.14776959419250488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,32,2,power_law_1.2,0.1625983953475952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,32,2,power_law_1.2,0.17210240364074708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,32,2,power_law_1.2,0.19406720399856567
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,32,2,power_law_1.2,0.23318400382995605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,32,2,power_law_1.2,0.23438079357147218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,32,2,power_law_1.2,0.29459199905395506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,32,2,power_law_1.2,0.35527040958404543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,2,power_law_1.01,0.14417279958724977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,32,2,power_law_1.2,0.47365121841430663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,2,power_law_1.01,0.16605440378189087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,2,power_law_1.01,0.09582719802856446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,32,2,power_law_1.2,0.5955584049224854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,2,power_law_1.01,0.16186879873275756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,2,power_law_1.01,0.1379263997077942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,32,2,power_law_1.2,0.8109888076782227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,2,power_law_1.01,0.1275904059410095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,2,power_law_1.01,0.19245439767837524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,2,power_law_1.01,0.2093951940536499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,32,2,power_law_1.2,1.096940803527832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,2,power_law_1.01,0.2834111928939819
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,2,power_law_1.01,0.2762687921524048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,2,power_law_1.01,0.42048001289367676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,32,2,power_law_1.2,1.5225791931152344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,2,power_law_1.01,0.5865791797637939
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,2,power_law_1.01,0.4402815818786621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,2,power_law_1.01,0.6095295906066894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,32,2,power_law_1.2,1.868889617919922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,2,power_law_1.01,0.632204818725586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,2,power_law_1.01,0.4875648021697998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,2,power_law_1.01,0.6708159923553467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,2,power_law_1.01,0.4969151973724365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,32,2,power_law_1.2,2.282828712463379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,2,power_law_1.01,0.6820032119750976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,2,power_law_1.01,0.7074687957763672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,2,power_law_1.01,0.5075647830963135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,2,power_law_1.01,0.71626877784729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,32,2,power_law_1.2,3.714303970336914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,2,power_law_1.01,0.7497856140136718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,2,power_law_1.01,0.5176383972167968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,2,power_law_1.01,0.7559360027313232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,2,power_law_1.01,0.5455743789672851
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,2,power_law_1.01,0.7823040008544921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,2,power_law_1.01,0.8207296371459961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,2,power_law_1.01,0.556217622756958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,2,power_law_1.01,0.9114815711975097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,32,2,power_law_1.2,7.310169219970703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,2,power_law_1.01,0.5644351959228515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,2,power_law_1.01,0.9093695640563965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,2,power_law_1.01,1.0446975708007813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,2,power_law_1.01,0.5848063945770263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,2,power_law_1.01,1.0271871566772461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,2,power_law_1.01,0.6054656028747558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,2,power_law_1.01,1.257209587097168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,2,power_law_1.01,0.6229311943054199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,2,power_law_1.01,1.3612223625183106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,2,power_law_1.01,1.7838848114013672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,2,power_law_1.01,0.6581567764282227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,2,power_law_1.01,2.0680896759033205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,2,power_law_1.01,0.7258687973022461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,2,power_law_1.01,2.720294380187988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,2,power_law_1.01,0.8131263732910157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,4,power_law_1.2,0.033792001008987424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,4,power_law_1.2,0.03361920118331909
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,2,power_law_1.01,3.2185855865478517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,2,power_law_1.01,0.9766271591186524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,4,power_law_1.2,0.03338879942893982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,4,power_law_1.2,0.03549439907073974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,4,power_law_1.2,0.04080640077590943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,2,power_law_1.01,1.0698880195617675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,4,power_law_1.2,0.05626239776611328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,2,power_law_1.01,4.01069450378418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,4,power_law_1.2,0.05865600109100342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,4,power_law_1.2,0.060070401430130003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,4,power_law_1.2,0.05974400043487549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,2,power_law_1.01,1.3642496109008788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,4,power_law_1.2,0.06059520244598389
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,4,power_law_1.2,0.0637440025806427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,2,power_law_1.01,5.9302207946777346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,4,power_law_1.2,0.06369919776916504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,4,power_law_1.2,0.0682752013206482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,2,power_law_1.01,1.6001983642578126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,4,power_law_1.2,0.07099519968032837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,4,power_law_1.2,0.07896320223808288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,4,power_law_1.2,0.08604800105094909
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,4,power_law_1.2,0.08518400192260742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,4,power_law_1.2,0.10016000270843506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,2,power_law_1.01,2.1472320556640625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,4,power_law_1.2,0.11137919425964356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,4,power_law_1.2,0.14042880535125732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,2,power_law_1.01,11.327718353271484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,4,power_law_1.2,0.16663039922714235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,4,power_law_1.2,0.2027008056640625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,2,power_law_1.01,2.6639551162719726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,4,power_law_1.2,0.2590656042098999
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,4,power_law_1.2,0.36313600540161134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,4,power_law_1.2,0.4706687927246094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,4,power_law_1.2,0.6957695960998536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,2,power_law_1.01,3.6911361694335936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,4,power_law_1.2,0.9870143890380859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,4,power_law_1.2,1.1191935539245605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,4,power_law_1.2,1.9799423217773438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,2,power_law_1.01,4.7976318359375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,32,balanced,0.0480320006608963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,32,balanced,0.051856001218159996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,32,balanced,0.05305600166320801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,32,balanced,0.07074133555094402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,32,balanced,0.08294933537642162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,32,balanced,0.0824533353249232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,32,balanced,0.0841493308544159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,4,power_law_1.2,4.0231681823730465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,32,balanced,0.08552533388137817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,32,balanced,0.08303999900817871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,32,balanced,0.08515733480453491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,32,balanced,0.08570133646329244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,32,balanced,0.08795733253161113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,32,balanced,0.08892800410588582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,32,balanced,0.08921066919962566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,32,balanced,0.0939626693725586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,32,balanced,0.09713600079218547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,32,balanced,0.09960533181826274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,32,balanced,0.11569600303967793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,32,balanced,0.12213333447774251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,32,balanced,0.14633599917093912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,32,balanced,0.16751466194788614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,32,balanced,0.22209600607554117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,32,balanced,0.21200533707936606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,32,balanced,0.2811200022697449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,1,2,power_law_1.01,5.725971221923828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,32,balanced,0.3255893389383952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,32,balanced,0.44686933358510333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,32,balanced,0.5372426509857178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,2,32,balanced,0.6613386472066244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,2,32,balanced,0.9714986483256022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,2,32,balanced,2.0283519426981607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,1,2,power_law_1.01,9.079827117919923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,32,power_law_1.01,0.06241919994354248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,32,power_law_1.01,0.056908798217773435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,32,power_law_1.01,0.043731200695037845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,32,power_law_1.01,0.04549759924411774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,4,power_law_1.01,0.04952319860458374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,32,power_law_1.01,0.04622719883918762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,32,power_law_1.01,0.05014399886131286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,4,power_law_1.01,0.08673279881477355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,32,power_law_1.01,0.05240960121154785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,1,2,power_law_1.01,17.423770141601562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,4,power_law_1.01,0.10248960256576538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,32,power_law_1.01,0.052108800411224364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,32,power_law_1.01,0.05074560046195984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,4,power_law_1.01,0.06849279999732971
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,32,power_law_1.01,0.05265920162200928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,4,power_law_1.01,0.09191679954528809
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,32,power_law_1.01,0.053337597846984865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,32,power_law_1.01,0.053881597518920896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,4,power_law_1.01,0.11755520105361938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,32,power_law_1.01,0.05411199927330017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,4,power_law_1.01,0.12568960189819336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,32,power_law_1.01,0.06020479798316956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,32,power_law_1.01,0.06471679806709289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,4,power_law_1.01,0.14967039823532105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,32,power_law_1.01,0.06673920154571533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,4,power_law_1.01,0.15887999534606934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,32,power_law_1.01,0.070278400182724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,32,power_law_1.01,0.08208640217781067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,4,power_law_1.01,0.16671359539031982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,32,power_law_1.01,0.08517760038375854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,4,power_law_1.01,0.16851199865341188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,32,power_law_1.01,0.1034559965133667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,8,8,power_law_1.2,0.031744000315666196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,8,8,power_law_1.2,0.03218559920787811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,4,power_law_1.01,0.17632639408111572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,32,power_law_1.01,0.11852799654006958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,8,8,power_law_1.2,0.03142400085926056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,8,8,power_law_1.2,0.03177599906921387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,8,8,power_law_1.2,0.03187839984893799
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,8,8,power_law_1.2,0.03328000009059906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,8,8,power_law_1.2,0.03488639891147614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,32,power_law_1.01,0.146943998336792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,8,8,power_law_1.2,0.035622400045394895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,8,8,power_law_1.2,0.038380798697471616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,4,power_law_1.01,0.17811199426651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,8,8,power_law_1.2,0.038489601016044615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,8,8,power_law_1.2,0.03909119963645935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,8,8,power_law_1.2,0.04302079975605011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,32,power_law_1.01,0.16634880304336547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,8,8,power_law_1.2,0.041171199083328246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,8,8,power_law_1.2,0.04266240000724793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,8,8,power_law_1.2,0.054655998945236206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,8,8,power_law_1.2,0.05604479908943176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,4,power_law_1.01,0.19132159948348998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,8,8,power_law_1.2,0.057766401767730714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,8,8,power_law_1.2,0.07349119782447815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,32,power_law_1.01,0.23899519443511963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,8,8,power_law_1.2,0.0821183979511261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,8,8,power_law_1.2,0.10473599433898925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,8,8,power_law_1.2,0.1322816014289856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,8,8,power_law_1.2,0.17320319414138793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,4,power_law_1.01,0.2002432107925415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,8,8,power_law_1.2,0.2100287914276123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,32,power_law_1.01,0.2875391960144043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,8,8,power_law_1.2,0.2933120012283325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,8,8,power_law_1.2,0.3779583930969238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,8,8,power_law_1.2,0.5619904041290283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,4,power_law_1.01,0.20790400505065917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,32,power_law_1.01,0.39818239212036133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,8,8,power_law_1.2,0.7151296138763428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,8,8,power_law_1.2,0.882528018951416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,4,power_law_1.01,0.22026240825653076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,8,8,power_law_1.2,1.2852288246154786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,32,power_law_1.01,0.48220157623291016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,4,power_law_1.01,0.2440959930419922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,8,8,power_law_1.2,2.6741695404052734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,32,power_law_1.01,0.5902912139892578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,4,power_law_1.01,0.27130880355834963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,32,power_law_1.01,0.9257792472839356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,4,power_law_1.01,0.3172287940979004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,4,power_law_1.01,0.3516607999801636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,4,power_law_1.01,0.4617599964141846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,32,power_law_1.01,1.7581695556640624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,4,power_law_1.01,0.5414527893066406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,4,power_law_1.01,0.7383296012878418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,4,power_law_1.01,0.8656064033508301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,32,2,balanced,0.031354665756225586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,32,2,balanced,0.03187733391920725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,32,2,balanced,0.03204799940188726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,32,2,balanced,0.03169599920511246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,32,2,balanced,0.033439998825391136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,32,2,balanced,0.03158933420976003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,32,2,balanced,0.033359999457995095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,32,2,balanced,0.03555200000603994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,32,2,balanced,0.03352533280849457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,32,2,balanced,0.03374933451414108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,32,2,balanced,0.03358400116364161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,32,2,balanced,0.035445332527160645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,32,2,balanced,0.03612799942493439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,32,2,balanced,0.03554133325815201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,32,2,balanced,0.04370133578777313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,32,2,balanced,0.045498669147491455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,32,2,balanced,0.045696000258127846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,32,2,balanced,0.051781331499417625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,32,2,balanced,0.056202664971351624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,32,2,balanced,0.07032000025113423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,4,power_law_1.01,1.270041561126709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,32,2,balanced,0.0819946676492691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,32,2,balanced,0.11342400312423706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,32,2,balanced,0.13462932904561362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,32,2,balanced,0.18878400325775146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,32,2,balanced,0.22622400522232056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,32,2,balanced,0.3243253429730733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,32,2,balanced,0.4310293197631836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,32,2,balanced,0.5268533229827881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,32,2,balanced,0.8014826774597168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,32,2,balanced,1.556320031483968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,4,power_law_1.01,1.5809727668762208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,4,power_law_1.01,1.8463232040405273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,4,power_law_1.01,0.0645695984363556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,4,power_law_1.01,0.06901760101318359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,4,power_law_1.01,0.06348800063133239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,4,power_law_1.01,2.8497087478637697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,4,power_law_1.01,0.07899519801139832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,4,power_law_1.01,0.09562879800796509
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,4,power_law_1.01,0.11505919694900513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,4,power_law_1.01,0.12691199779510498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,4,power_law_1.01,0.12726399898529053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,4,power_law_1.01,0.1286463975906372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,4,power_law_1.01,0.12743040323257446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,4,power_law_1.01,5.502783966064453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,4,power_law_1.01,0.13408639430999755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,4,power_law_1.01,0.14082560539245606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,4,power_law_1.2,0.07633919715881347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,4,power_law_1.01,0.14028160572052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,4,power_law_1.2,0.05917440056800842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,4,power_law_1.01,0.14661120176315307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,4,power_law_1.2,0.07251840233802795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,4,power_law_1.01,0.1546623945236206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,4,power_law_1.2,0.0818880021572113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,4,power_law_1.2,0.10660480260848999
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,4,power_law_1.01,0.1584831953048706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,4,power_law_1.2,0.17363200187683106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,4,power_law_1.01,0.16846079826354982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,4,power_law_1.2,0.18272639513015748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,4,power_law_1.01,0.19411200284957886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,4,power_law_1.2,0.1893056035041809
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,4,power_law_1.01,0.20967679023742675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,4,power_law_1.2,0.18465280532836914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,4,power_law_1.2,0.18665599822998047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,4,power_law_1.01,0.2622335910797119
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,4,power_law_1.2,0.1929471969604492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,4,power_law_1.01,0.2771008014678955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,4,power_law_1.2,0.20551679134368897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,4,power_law_1.01,0.36414079666137694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,4,power_law_1.2,0.2048896074295044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,4,power_law_1.2,0.21552639007568358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,4,power_law_1.01,0.44106879234313967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,4,power_law_1.2,0.23061120510101318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,4,power_law_1.01,0.5537727832794189
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,4,power_law_1.2,0.23161599636077881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,4,power_law_1.2,0.23700480461120604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,4,power_law_1.01,0.6876607894897461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,4,power_law_1.2,0.26277759075164797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,4,power_law_1.2,0.2807487964630127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,4,power_law_1.01,0.9461183547973633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,4,power_law_1.2,0.3059648036956787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,4,power_law_1.2,0.3384768009185791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,4,power_law_1.01,0.06305279731750488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,4,power_law_1.01,1.3340160369873046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,4,power_law_1.2,0.4260672092437744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,4,power_law_1.2,0.47331838607788085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,4,power_law_1.01,0.06584960222244263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,4,power_law_1.01,1.5874048233032227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,4,power_law_1.2,0.6699391841888428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,4,power_law_1.2,0.7495359897613525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,4,power_law_1.01,0.08106240034103393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,4,power_law_1.01,2.439743995666504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,4,power_law_1.2,1.084550380706787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,4,power_law_1.01,0.06657919883728028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,4,power_law_1.2,1.6395776748657227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,4,power_law_1.01,0.07927039861679078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,4,power_law_1.2,1.7016191482543945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,4,power_law_1.01,0.0911296010017395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,4,power_law_1.01,5.247769546508789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,4,power_law_1.01,0.09608960151672363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,4,power_law_1.2,2.571955108642578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,4,power_law_1.01,0.10823040008544922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,4,power_law_1.01,0.10959359407424926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,4,power_law_1.2,5.021120071411133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,4,power_law_1.01,0.11320320367813111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,4,power_law_1.01,0.11725440025329589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,4,power_law_1.01,0.1286784052848816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,4,power_law_1.01,0.1251263976097107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,4,power_law_1.01,0.12748160362243652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,4,power_law_1.01,0.13891199827194214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,32,4,balanced,0.044064000248909
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,32,4,balanced,0.04621333380540212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,32,4,balanced,0.04737600187460581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,32,4,balanced,0.050288001696268715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,32,4,balanced,0.05856533348560333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,32,4,balanced,0.0738613357146581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,32,4,balanced,0.07562133173147838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,32,4,balanced,0.07889600098133087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,32,4,balanced,0.07880533238252004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,32,4,balanced,0.07971199850241344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,32,4,balanced,0.07673599819342296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,32,4,balanced,0.08323733508586884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,32,4,balanced,0.08307733138402303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,32,4,balanced,0.08956799904505412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,32,4,balanced,0.09745599826176961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,32,4,balanced,0.10031466682751973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,32,4,balanced,0.11000532905260722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,4,power_law_1.01,0.14353920221328736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,32,4,balanced,0.1260533332824707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,32,4,balanced,0.13874133427937826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,32,4,balanced,0.16498133540153503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,32,4,balanced,0.19191465775171915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,32,4,balanced,0.2654719948768616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,32,4,balanced,0.3107573390007019
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,32,4,balanced,0.45421334107716876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,32,4,balanced,0.5734399954477946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,32,4,balanced,0.8367946942647299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,32,4,balanced,1.0917973518371582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,4,power_law_1.01,0.1498944044113159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,32,4,balanced,1.3426507314046223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,32,4,balanced,2.068720022837321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,4,power_law_1.01,0.16636159420013427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,32,4,balanced,4.069936116536458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,32,power_law_1.2,0.041145598888397215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,32,power_law_1.2,0.031699201464653014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,4,power_law_1.01,0.1842560052871704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,32,power_law_1.2,0.03230080008506775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,32,power_law_1.2,0.0335999995470047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,4,power_law_1.01,0.21662719249725343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,32,power_law_1.2,0.034220799803733826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,32,power_law_1.2,0.03488639891147614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,32,power_law_1.2,0.0348800003528595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,4,power_law_1.01,0.24990079402923585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,32,power_law_1.2,0.03813759982585907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,32,power_law_1.2,0.039987200498580934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,4,power_law_1.01,0.32305920124053955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,32,power_law_1.2,0.04169600009918213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,32,power_law_1.2,0.04431999921798706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,32,power_law_1.2,0.04568319916725159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,4,power_law_1.01,0.39086079597473145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,32,power_law_1.2,0.047942399978637695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,32,power_law_1.2,0.047603198885917665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,32,power_law_1.2,0.056364798545837404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,4,power_law_1.01,0.5472832202911377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,32,power_law_1.2,0.0588096022605896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,32,power_law_1.2,0.06313599944114685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,32,power_law_1.2,0.07560319900512695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,4,power_law_1.01,0.7038911819458008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,32,power_law_1.2,0.08101119995117187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,32,power_law_1.2,0.10158720016479492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,2,balanced,0.05416533350944519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,32,power_law_1.2,0.12208000421524048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,2,balanced,0.08312533299128215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,2,balanced,0.09181333581606548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,4,power_law_1.01,0.9528384208679199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,2,balanced,0.13146133224169412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,2,balanced,0.19976532459259033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,2,balanced,0.2634933392206828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,2,balanced,0.2672746578852336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,2,balanced,0.26743467648824054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,32,power_law_1.2,0.16293760538101196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,2,balanced,0.2667466600735982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,2,balanced,0.2695466677347819
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,2,balanced,0.2732800046602885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,2,balanced,0.2758026719093323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,2,balanced,0.27827199300130206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,2,balanced,0.2829333345095317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,2,balanced,0.2927146752675374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,32,power_law_1.2,0.1990847945213318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,2,balanced,0.2972106734911601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,2,balanced,0.3121440013249715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,2,balanced,0.3477120002110799
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,2,balanced,0.36085331439971924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,2,balanced,0.48715198040008545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,2,balanced,0.499941349029541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,32,power_law_1.2,0.29932799339294436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,2,balanced,0.7664639949798584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,4,power_law_1.01,1.2376192092895508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,2,balanced,0.7686026891072592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,2,balanced,1.1386720339457195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,2,balanced,1.2933226426442463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,32,power_law_1.2,0.3636352062225342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,2,balanced,1.9086079597473145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,2,balanced,2.3360692660013833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,32,power_law_1.2,0.5131519794464111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,4,2,balanced,2.963738759358724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,4,power_law_1.01,1.5624704360961914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,32,power_law_1.2,0.6857151985168457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,4,2,balanced,4.446912129720052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,32,power_law_1.2,0.9752320289611817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,4,2,balanced,8.829999923706055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,4,power_law_1.01,2.3330623626708986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,32,power_law_1.2,1.6391040802001953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,32,power_law_1.2,3.073427200317383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,4,power_law_1.01,4.730355072021484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,16,balanced,0.03190399954716364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,16,balanced,0.031354665756225586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,16,balanced,0.029279999434947968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,16,balanced,0.03133333226044973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,16,balanced,0.031328000128269196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,16,balanced,0.03139200061559677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,16,balanced,0.03331200033426285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,16,balanced,0.03186666717131933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,16,balanced,0.03342399994532267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,16,balanced,0.03341866781314214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,16,balanced,0.03164800008138021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,16,balanced,0.03402133285999298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,16,balanced,0.03534399966398875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,16,balanced,0.033717334270477295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,16,balanced,0.037445334096749626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,16,balanced,0.03754666695992152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,16,balanced,0.04187199970086416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,16,balanced,0.04181333382924398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,16,balanced,0.0436160018046697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,16,balanced,0.047775998711586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,16,balanced,0.05036266644795736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,16,balanced,0.06314133107662201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,16,balanced,0.07241599758466084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,16,balanced,0.09222933650016785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,16,balanced,0.1143999993801117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,16,balanced,0.14155200123786926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,16,balanced,0.17619733015696207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,16,balanced,0.21336533625920615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,16,balanced,0.3011253277460734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,16,balanced,0.5473013321558634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,64,balanced,0.05000533163547516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,64,balanced,0.04778666794300079
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,64,balanced,0.045456002155939736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,64,balanced,0.04582933088143667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,64,balanced,0.0476800004641215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,64,balanced,0.04991999765237173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,64,balanced,0.05190933247407278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,64,balanced,0.052111998200416565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,64,balanced,0.05376533170541128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,64,balanced,0.05197866757710775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,64,balanced,0.05395199855168661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,64,balanced,0.0561653325955073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,64,balanced,0.056261335810025535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,64,balanced,0.0580320010582606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,64,balanced,0.06247999767462412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,64,balanced,0.06418666740258534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,64,balanced,0.06829866766929626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,64,balanced,0.07770666480064392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,64,balanced,0.09156266848246257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,64,balanced,0.11545600493748982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,64,balanced,0.13340800007184347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,64,balanced,0.1718026598294576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,64,balanced,0.20755734046300253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,64,balanced,0.28037333488464355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,64,balanced,0.34830931822458905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,64,balanced,0.49542399247487384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,64,balanced,0.6300426721572876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,8,power_law_1.01,0.22188799381256102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,64,balanced,0.7758879661560059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,64,balanced,1.1898826758066814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,64,balanced,2.3459787368774414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,8,power_law_1.2,0.05027840137481689
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,8,power_law_1.01,0.33690879344940183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,8,power_law_1.2,0.06108800172805786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,8,power_law_1.01,0.11140480041503906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,8,power_law_1.2,0.05366399884223938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,8,power_law_1.01,0.1638208031654358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,8,power_law_1.2,0.05406079888343811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,2,balanced,0.04610666632652283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,2,balanced,0.0481279989083608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,2,balanced,0.05263466636339823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,2,balanced,0.08020799855391185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,2,balanced,0.11272000273068745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,2,balanced,0.12825600306193033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,2,balanced,0.1293280025323232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,2,balanced,0.12956800063451132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,2,balanced,0.1283519963423411
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,2,balanced,0.13145066301027933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,8,power_law_1.01,0.2273344039916992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,2,balanced,0.1318666636943817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,2,balanced,0.13451733191808066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,2,balanced,0.13618133465449014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,2,balanced,0.1443946659564972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,2,balanced,0.1502133309841156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,2,balanced,0.15427199999491373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,2,balanced,0.16053332885106406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,2,balanced,0.1835199991861979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,2,balanced,0.19451733430226645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,2,balanced,0.2515520056088765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,2,balanced,0.27542932828267414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,2,balanced,0.3874080181121826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,2,balanced,0.4327840010325114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,8,power_law_1.2,0.0579584002494812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,2,balanced,0.6392159859339396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,2,balanced,0.7388853232065836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,2,balanced,1.170474688212077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,8,power_law_1.01,0.33666560649871824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,2,balanced,1.3654185930887859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,2,balanced,1.6248799959818523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,2,balanced,2.5188533465067544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,8,power_law_1.2,0.06517760157585144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,8,power_law_1.01,0.37017600536346434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,2,balanced,4.857018788655599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,8,power_law_1.2,0.07223039865493774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,8,power_law_1.01,0.3905535936355591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,32,2,power_law_1.01,0.047091200947761536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,8,power_law_1.2,0.07935360074043274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,8,power_law_1.01,0.3776576042175293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,32,2,power_law_1.01,0.04963839948177338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,32,2,power_law_1.01,0.05134720206260681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,8,power_law_1.2,0.08081279993057251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,8,power_law_1.01,0.39864320755004884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,32,2,power_law_1.01,0.060262399911880496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,32,2,power_law_1.01,0.06981120109558106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,8,power_law_1.2,0.08748800158500672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,8,power_law_1.01,0.3987839937210083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,32,2,power_law_1.01,0.08264319896697998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,32,2,power_law_1.01,0.10080000162124633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,8,power_law_1.2,0.08753920197486878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,8,power_law_1.01,0.41998720169067383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,32,2,power_law_1.01,0.10081919431686401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,32,2,power_law_1.01,0.10540159940719604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,8,power_law_1.2,0.08991360068321227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,8,power_law_1.01,0.41710720062255857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,32,2,power_law_1.01,0.10786559581756591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,4,balanced,0.04357333481311798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,4,balanced,0.04355733096599579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,4,balanced,0.05606933434804281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,4,balanced,0.07565333445866902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,4,balanced,0.10498133301734924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,4,balanced,0.11847999691963196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,8,power_law_1.01,0.4213119983673096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,4,balanced,0.12168000141779582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,32,2,power_law_1.01,0.11492480039596557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,8,power_law_1.2,0.08942080140113831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,4,balanced,0.12364266316095988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,4,balanced,0.12170132994651794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,4,balanced,0.12290666500727336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,4,balanced,0.12356799840927124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,4,balanced,0.1250933309396108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,4,balanced,0.12797866264979044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,4,balanced,0.1327839990456899
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,4,balanced,0.1383093297481537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,4,balanced,0.14192533493041992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,4,balanced,0.1485973298549652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,4,balanced,0.16753600041071573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,4,balanced,0.17814934253692627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,4,balanced,0.21052799622217813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,32,2,power_law_1.01,0.11898239850997924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,4,balanced,0.2323840061823527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,4,balanced,0.3038506706555684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,4,balanced,0.35205864906311035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,4,balanced,0.4903093179066976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,4,balanced,0.6057973305384318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,8,power_law_1.01,0.43318400382995603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,4,balanced,0.8495732943216959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,8,power_law_1.2,0.0946943998336792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,32,2,power_law_1.01,0.12309119701385499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,4,balanced,1.0952906608581543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,4,balanced,1.3031466801961262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,4,balanced,2.05515193939209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,32,2,power_law_1.01,0.13281919956207275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,8,power_law_1.01,0.4503744125366211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,4,balanced,3.9429972966512046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,8,power_law_1.2,0.10439039468765259
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,32,2,power_law_1.01,0.14382719993591309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,32,2,power_law_1.01,0.1545408010482788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,8,power_law_1.01,0.48520960807800295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,8,power_law_1.2,0.1060096025466919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,32,2,power_law_1.01,0.1529855966567993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,32,2,power_law_1.01,0.17918720245361328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,8,power_law_1.01,0.5431871891021729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,8,power_law_1.2,0.11374080181121826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,32,2,power_law_1.01,0.20444159507751464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,32,2,power_law_1.01,0.2598400115966797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,8,power_law_1.01,0.5436031818389893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,8,power_law_1.2,0.12840960025787354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,32,2,power_law_1.01,0.3070784091949463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,32,2,power_law_1.01,0.4196415901184082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,8,power_law_1.2,0.13845759630203247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,8,power_law_1.01,0.6220160007476807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,32,2,power_law_1.01,0.5328767776489258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,8,power_law_1.2,0.1606719970703125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,8,power_law_1.01,0.7308159828186035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,32,2,power_law_1.01,0.7173056125640869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,32,2,power_law_1.01,0.9059136390686036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,8,power_law_1.2,0.19898879528045654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,8,power_law_1.01,0.8867072105407715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,4,balanced,0.07014399766921997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,4,balanced,0.12009066343307495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,4,balanced,0.14526399970054626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,32,2,power_law_1.01,1.2978303909301758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,4,balanced,0.24908800919850668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,4,balanced,0.434773325920105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,4,balanced,0.826794703801473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,8,power_law_1.2,0.2527807950973511
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,4,balanced,1.2114506562550862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,4,balanced,1.2129440307617188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,4,balanced,1.2147626876831055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,8,power_law_1.01,0.855628776550293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,4,balanced,1.2169280052185059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,4,balanced,1.2230933507283528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,32,2,power_law_1.01,1.6989887237548829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,4,balanced,1.223530689875285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,4,balanced,1.2316213448842366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,4,balanced,1.2331733703613281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,8,power_law_1.2,0.30852479934692384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,4,balanced,1.2411146958669026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,4,balanced,1.2489546934763591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,4,balanced,1.255568027496338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,4,balanced,1.2736480236053467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,4,balanced,1.2881386280059814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,8,power_law_1.01,1.0870271682739259
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,32,2,power_law_1.01,2.1058303833007814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,4,balanced,1.3221333026885986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,4,balanced,1.3584853808085124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,4,balanced,1.4181440671284993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,8,power_law_1.2,0.41268482208251955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,4,balanced,1.4999465942382812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,4,balanced,1.6014933586120605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,4,balanced,1.791770617167155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,4,balanced,2.0026559829711914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,32,2,power_law_1.01,3.307078552246094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,8,power_law_1.01,1.4343808174133301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,4,balanced,2.649797280629476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,8,power_law_1.2,0.49735679626464846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,4,balanced,3.3252318700154624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,4,balanced,4.175002733866374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,8,power_law_1.2,0.7195263862609863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,4,balanced,8.059439977010092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,8,power_law_1.01,1.9023040771484374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,32,2,power_law_1.01,6.482892608642578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,8,power_law_1.2,0.8602304458618164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,8,power_law_1.01,2.140096092224121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,8,power_law_1.2,1.0720767974853516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,1,8,power_law_1.01,2.540883255004883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,8,power_law_1.2,1.72938232421875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,8,power_law_1.2,3.320403289794922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,1,balanced,0.03374933451414108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,1,balanced,0.03558400024970373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,1,balanced,0.045706664522488914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,1,balanced,0.06821866830190022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,1,balanced,0.10803733269373576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,1,balanced,0.12730133533477783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,1,balanced,0.12869866689046225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,1,8,power_law_1.01,3.7045566558837892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,1,balanced,0.12691733241081238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,1,balanced,0.12782933314641318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,1,balanced,0.12925333778063455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,1,balanced,0.13059733311335245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,1,balanced,0.13385599851608276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,1,balanced,0.13531200091044107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,64,balanced,0.039749334255854286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,1,balanced,0.1363146702448527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,64,balanced,0.03896533449490865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,1,balanced,0.1400266687075297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,1,balanced,0.1413279970486959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,64,balanced,0.035504000882307686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,1,balanced,0.14607999722162882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,64,balanced,0.0354720006386439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,1,balanced,0.17337065935134888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,64,balanced,0.037632000943024956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,1,balanced,0.1662453313668569
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,64,balanced,0.03765333443880081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,64,balanced,0.035775999228159584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,1,balanced,0.2077066699663798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,64,balanced,0.03758399933576584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,1,balanced,0.2166879971822103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,64,balanced,0.035749333600203194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,64,balanced,0.03755199909210205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,1,balanced,0.30305065711339313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,64,balanced,0.03750933210055033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,64,balanced,0.0395413339138031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,1,balanced,0.3162879943847656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,64,balanced,0.03941333293914795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,64,balanced,0.039749334255854286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,64,balanced,0.04370133578777313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,1,balanced,0.4828000068664551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,64,balanced,0.043712000052134194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,64,balanced,0.04390933116277059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,1,balanced,0.5058453480402628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,64,balanced,0.04381866753101349
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,64,balanced,0.04785066843032837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,64,balanced,0.047872001926104225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,1,balanced,0.7510720094045004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,64,balanced,0.05208000044027964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,64,balanced,0.05808533231417338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,64,balanced,0.06630933284759521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,1,balanced,0.9127306938171387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,64,balanced,0.08462933699289958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,64,balanced,0.09661866227785747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,64,balanced,0.12602667013804117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,4,1,balanced,1.1364800135294597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,64,balanced,0.15397333105405173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,64,balanced,0.18292800585428873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,64,balanced,0.2516426642735799
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,4,1,balanced,1.7497013409932454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,64,balanced,0.46107733249664307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,4,1,balanced,3.4426186879475913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,1,8,power_law_1.01,7.1851646423339846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,4,2,power_law_1.2,0.03961600065231323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,4,2,power_law_1.2,0.04761599898338318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,4,2,power_law_1.2,0.04912639856338501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,4,2,power_law_1.2,0.05777279734611511
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,4,2,power_law_1.2,0.0590719997882843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,4,2,power_law_1.2,0.06122879981994629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,4,2,power_law_1.2,0.06094080209732056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,4,2,power_law_1.2,0.06119040250778198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,4,2,power_law_1.2,0.06403840184211732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,4,2,power_law_1.2,0.06533120274543762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,4,2,power_law_1.2,0.06570879817008972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,4,2,power_law_1.2,0.07047680020332336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,4,2,power_law_1.2,0.06848639845848084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,4,2,power_law_1.2,0.0766592025756836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,8,2,power_law_1.01,0.036057600378990175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,4,2,power_law_1.2,0.08589439988136291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,8,2,power_law_1.01,0.0463808000087738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,4,2,power_law_1.2,0.08848639726638793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,8,2,power_law_1.01,0.04017919898033142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,4,2,power_law_1.2,0.08785920143127442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,8,2,power_law_1.01,0.05397760272026062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,8,1,power_law_1.01,0.05358719825744629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,8,2,power_law_1.01,0.05459200143814087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,4,2,power_law_1.2,0.10543359518051147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,8,2,power_law_1.01,0.05707520246505737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,8,1,power_law_1.01,0.06058239936828613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,4,2,power_law_1.2,0.11986559629440308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,8,2,power_law_1.01,0.0582144021987915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,8,1,power_law_1.01,0.0766975998878479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,4,2,power_law_1.2,0.1574079990386963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,8,2,power_law_1.01,0.05852800011634827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,8,1,power_law_1.01,0.11183359622955322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,8,2,power_law_1.01,0.05867519974708557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,4,2,power_law_1.2,0.1651584029197693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,8,1,power_law_1.01,0.11559679508209228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,8,2,power_law_1.01,0.059596800804138185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,8,1,power_law_1.01,0.11832319498062134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,8,2,power_law_1.01,0.06234239935874939
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,4,2,power_law_1.2,0.23960959911346436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,8,1,power_law_1.01,0.1167680025100708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,8,2,power_law_1.01,0.0626688003540039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,4,2,power_law_1.2,0.31162879467010496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,8,2,power_law_1.01,0.06566399931907654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,8,1,power_law_1.01,0.1220352053642273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,8,2,power_law_1.01,0.07340800166130065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,8,1,power_law_1.01,0.12033280134201049
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,4,2,power_law_1.2,0.4153088092803955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,8,2,power_law_1.01,0.07871999740600585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,8,1,power_law_1.01,0.1262719988822937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,8,2,power_law_1.01,0.08202880024909973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,8,1,power_law_1.01,0.12216960191726685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,4,2,power_law_1.2,0.5391551971435546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,8,2,power_law_1.01,0.0946175992488861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,8,1,power_law_1.01,0.1272704005241394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,8,2,power_law_1.01,0.10502400398254394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,8,1,power_law_1.01,0.12903679609298707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,8,2,power_law_1.01,0.11930240392684936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,4,2,power_law_1.2,0.7177984237670898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,8,1,power_law_1.01,0.13871999979019164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,8,1,power_law_1.01,0.15048960447311402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,8,2,power_law_1.01,0.14728319644927979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,8,1,power_law_1.01,0.1543295979499817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,8,2,power_law_1.01,0.17479039430618287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,4,2,power_law_1.2,0.9947648048400879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,8,1,power_law_1.01,0.15585919618606567
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,8,2,power_law_1.01,0.2567487955093384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,8,1,power_law_1.01,0.17880959510803224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,8,2,power_law_1.01,0.29979519844055175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,8,1,power_law_1.01,0.19481600522994996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,4,2,power_law_1.2,1.2758336067199707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,8,2,power_law_1.01,0.43112959861755373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,8,1,power_law_1.01,0.2422719955444336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,8,1,power_law_1.01,0.286028790473938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,8,2,power_law_1.01,0.5116288185119628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,8,1,power_law_1.01,0.3780672073364258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,8,2,power_law_1.01,0.7680255889892578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,8,1,power_law_1.01,0.44576001167297363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,4,2,power_law_1.2,1.8261632919311523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,1,balanced,0.08130666613578796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,1,balanced,0.08896000186602275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,1,balanced,0.09899200002352397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,1,balanced,0.12338667114575703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,8,1,power_law_1.01,0.6051008224487304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,1,balanced,0.1810879906018575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,8,2,power_law_1.01,0.9869759559631348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,1,balanced,0.2693279981613159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,1,balanced,0.2764479915301005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,1,balanced,0.2760266661643982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,8,1,power_law_1.01,0.7828095912933349
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,1,balanced,0.27753599484761554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,1,balanced,0.27720000346501666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,1,balanced,0.28148800134658813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,1,balanced,0.28865599632263184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,8,2,power_law_1.01,1.1640383720397949
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,1,balanced,0.2895359992980957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,1,balanced,0.2973066568374634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,8,1,power_law_1.01,1.1280960083007812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,1,balanced,0.3096160093943278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,1,balanced,0.3141226569811503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,1,balanced,0.32713067531585693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,4,2,power_law_1.2,3.696147155761719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,1,balanced,0.3705386718114217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,1,balanced,0.40301867326100665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,8,1,power_law_1.01,1.4015232086181642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,1,balanced,0.4779893159866333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,8,2,power_law_1.01,1.894118309020996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,1,balanced,0.5421973466873169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,1,balanced,0.7054186662038168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,1,balanced,0.8565599918365479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,8,1,power_law_1.01,1.730342483520508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,1,balanced,1.2667893568674724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,1,balanced,1.605120023091634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,8,2,power_law_1.01,3.616729736328125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,1,balanced,2.3281119664510093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,8,1,power_law_1.01,2.6901119232177733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,1,balanced,3.0447734196980796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,1,balanced,3.8222506841023765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,8,1,power_law_1.01,5.241312026977539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,1,balanced,5.832618713378906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,1,balanced,11.299722035725912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,16,power_law_1.01,0.11973119974136352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,16,power_law_1.2,0.055641597509384154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,16,power_law_1.2,0.036211198568344115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,16,power_law_1.2,0.03383679986000061
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,2,1,power_law_1.2,0.07626240253448487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,2,1,power_law_1.2,0.11644799709320068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,16,power_law_1.2,0.0384768009185791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,2,1,power_law_1.2,0.1293503999710083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,16,power_law_1.01,0.1519744038581848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,2,1,power_law_1.2,0.19715839624404907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,16,power_law_1.2,0.04505600035190582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,2,1,power_law_1.2,0.2050623893737793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,2,1,power_law_1.2,0.2069632053375244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,2,1,power_law_1.2,0.21413118839263917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,16,power_law_1.2,0.05594879984855652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,2,1,power_law_1.2,0.20818560123443602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,2,1,power_law_1.2,0.22181758880615235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,16,power_law_1.2,0.05532159805297852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,2,1,power_law_1.2,0.22335360050201417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,8,4,balanced,0.03932799895604452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,8,4,balanced,0.03366933266321818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,2,1,power_law_1.2,0.21998720169067382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,8,4,balanced,0.035573333501815796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,8,4,balanced,0.03632533301909765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,8,4,balanced,0.035749333600203194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,2,1,power_law_1.2,0.227622389793396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,8,4,balanced,0.035829332967599235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,8,4,balanced,0.03749333322048187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,8,4,balanced,0.036101333796978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,16,power_law_1.2,0.05472639799118042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,8,4,balanced,0.035818666219711304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,2,1,power_law_1.2,0.23173120021820068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,8,4,balanced,0.03585066646337509
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,8,4,balanced,0.03693866729736328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,8,4,balanced,0.03806400050719579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,8,4,balanced,0.03908266623814901
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,8,4,balanced,0.039674667020638786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,2,1,power_law_1.2,0.24931840896606444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,8,4,balanced,0.05105599761009216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,8,4,balanced,0.0535093347231547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,8,4,balanced,0.05402666827042898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,16,power_law_1.01,0.0716480016708374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,8,4,balanced,0.06703466673692067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,2,1,power_law_1.2,0.28409600257873535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,16,power_law_1.2,0.055955201387405396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,8,4,balanced,0.05990933378537496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,8,4,balanced,0.0772159993648529
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,8,4,balanced,0.07729599873224895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,2,1,power_law_1.2,0.2925568103790283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,8,4,balanced,0.10312533378601074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,8,4,balanced,0.11559999982515971
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,8,4,balanced,0.14946132898330688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,8,4,balanced,0.1736639936765035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,2,1,power_law_1.2,0.2942591905593872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,8,4,balanced,0.23253333568572998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,16,power_law_1.2,0.05671679973602295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,8,4,balanced,0.2728000084559123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,2,1,power_law_1.2,0.32855041027069093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,8,4,balanced,0.34112000465393066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,8,4,balanced,0.49456000328063965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,2,1,power_law_1.2,0.36513919830322267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,8,4,balanced,0.9299893379211426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,16,power_law_1.2,0.05857279896736145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,2,1,power_law_1.2,0.4614719867706299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,2,1,power_law_1.2,0.5477568149566651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,16,power_law_1.2,0.05878400206565857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,2,1,power_law_1.2,0.7205696105957031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,16,power_law_1.2,0.0607807993888855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,2,1,power_law_1.2,0.8910271644592285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,16,power_law_1.01,0.09245439767837524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,2,1,power_law_1.2,1.2412927627563477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,16,power_law_1.2,0.064300799369812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,2,1,power_law_1.2,1.6175487518310547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,16,power_law_1.2,0.07203199863433837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,16,power_law_1.2,0.07607679963111877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,2,1,power_law_1.2,2.325836753845215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,16,power_law_1.2,0.07905920147895813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,16,power_law_1.01,0.1034432053565979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,2,1,power_law_1.2,3.1077503204345702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,16,power_law_1.2,0.08725759983062745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,16,power_law_1.2,0.09738879799842834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,2,1,power_law_1.2,3.8335296630859377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,16,power_law_1.2,0.12350720167160034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,16,power_law_1.2,0.13621120452880858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,16,power_law_1.01,0.13099520206451415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,16,power_law_1.2,0.18883199691772462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,2,1,power_law_1.2,5.846649551391602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,16,power_law_1.2,0.23449599742889404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,16,power_law_1.2,0.3123199939727783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,16,power_law_1.01,0.13292160034179687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,16,power_law_1.2,0.43050241470336914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,16,power_law_1.2,0.5909759998321533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,2,1,power_law_1.2,11.466604614257813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,16,power_law_1.2,0.8194047927856445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,16,power_law_1.01,0.13481600284576417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,32,power_law_1.01,0.09500160217285156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,16,power_law_1.2,1.0078463554382324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,32,power_law_1.01,0.08749439716339111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,16,power_law_1.01,0.13822720050811768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,32,power_law_1.01,0.07616639733314515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,16,power_law_1.2,1.3526975631713867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,32,power_law_1.01,0.06072319746017456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,16,power_law_1.01,0.14206720590591432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,32,power_law_1.01,0.06770560145378113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,32,power_law_1.01,0.0711679995059967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,16,power_law_1.2,3.1222591400146484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,16,power_law_1.01,0.14302079677581786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,32,power_law_1.01,0.08057600259780884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,32,power_law_1.01,0.08190720081329346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,16,power_law_1.01,0.14816000461578369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,32,power_law_1.01,0.08199679851531982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,32,power_law_1.01,0.08097280263900757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,16,power_law_1.2,0.10342400074005127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,32,power_law_1.01,0.08306559920310974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,16,power_law_1.01,0.15370880365371703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,16,power_law_1.2,0.05244160294532776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,32,power_law_1.01,0.0844543993473053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,16,power_law_1.2,0.0692799985408783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,16,power_law_1.2,0.07103999853134155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,32,power_law_1.01,0.08504319787025452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,16,power_law_1.01,0.15168639421463012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,16,power_law_1.2,0.08000640273094177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,32,power_law_1.01,0.08669440150260925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,16,power_law_1.2,0.09518719911575317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,32,power_law_1.01,0.0910975992679596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,16,power_law_1.2,0.10760320425033569
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,16,power_law_1.01,0.16001919507980347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,16,power_law_1.2,0.10391039848327636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,32,power_law_1.01,0.09400960206985473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,16,power_law_1.2,0.10616960525512695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,32,power_law_1.01,0.0990015983581543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,16,power_law_1.2,0.10931199789047241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,16,power_law_1.01,0.16172800064086915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,16,power_law_1.2,0.11237759590148926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,32,power_law_1.01,0.11109119653701782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,128,power_law_1.01,0.07457919716835022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,16,power_law_1.2,0.1145792007446289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,32,power_law_1.01,0.12021119594573974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,16,power_law_1.2,0.11228799819946289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,128,power_law_1.01,0.07330560088157653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,16,power_law_1.2,0.1184831976890564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,32,power_law_1.01,0.14033279418945313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,16,power_law_1.01,0.17429759502410888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,128,power_law_1.01,0.07187839746475219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,16,power_law_1.2,0.12789759635925294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,32,power_law_1.01,0.1510848045349121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,16,power_law_1.2,0.1372607946395874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,128,power_law_1.01,0.06949120163917541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,16,power_law_1.2,0.14743679761886597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,32,power_law_1.01,0.19806720018386842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,128,power_law_1.01,0.07401599884033203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,16,power_law_1.01,0.20546560287475585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,16,power_law_1.2,0.17088639736175537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,32,power_law_1.01,0.22833919525146484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,128,power_law_1.01,0.07271680235862732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,16,power_law_1.2,0.18291200399398805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,16,power_law_1.2,0.2263808012008667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,128,power_law_1.01,0.07304959893226623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,32,power_law_1.01,0.3104768037796021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,16,power_law_1.01,0.2099519968032837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,16,power_law_1.2,0.27162880897521974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,128,power_law_1.01,0.07299839854240417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,16,power_law_1.2,0.30492799282073973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,32,power_law_1.01,0.37500159740447997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,128,power_law_1.01,0.07377920150756836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,16,power_law_1.2,0.39073278903961184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,32,power_law_1.01,0.5590911865234375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,128,power_law_1.01,0.07466880083084107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,16,power_law_1.01,0.2583679914474487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,16,power_law_1.2,0.5509952068328857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,128,power_law_1.01,0.07592319846153259
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,16,power_law_1.2,0.47135357856750487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,32,power_law_1.01,0.6283584117889405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,128,power_law_1.01,0.07728639841079712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,16,power_law_1.2,0.8982975959777832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,16,power_law_1.01,0.28495359420776367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,128,power_law_1.01,0.07771520018577575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,32,power_law_1.01,0.7963840007781983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,128,power_law_1.01,0.0776639997959137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,16,power_law_1.2,1.0469568252563477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,128,power_law_1.01,0.0840831995010376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,16,power_law_1.01,0.35318400859832766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,16,power_law_1.2,1.2523648262023925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,32,power_law_1.01,1.25283203125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,128,power_law_1.01,0.0828927993774414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,128,power_law_1.01,0.08483840227127075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,128,power_law_1.01,0.09720960259437561
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,16,power_law_1.2,2.554099273681641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,16,power_law_1.01,0.36253440380096436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,128,power_law_1.01,0.10290559530258178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,32,power_law_1.01,2.644710350036621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,128,power_law_1.01,0.11510399580001832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,128,power_law_1.01,0.1253056049346924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,16,power_law_1.01,0.4618879795074463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,16,power_law_1.2,4.385593414306641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,128,power_law_1.01,0.1606719970703125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,128,power_law_1.01,0.19594240188598633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,16,power_law_1.01,0.5581952095031738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,128,power_law_1.01,0.2444607973098755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,128,power_law_1.01,0.3352576017379761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,128,power_law_1.01,0.47081599235534666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,16,power_law_1.01,0.8066559791564941
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,128,power_law_1.01,0.5798399925231934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,128,power_law_1.01,0.6863232135772706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,16,power_law_1.01,0.922214412689209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,128,power_law_1.01,1.1506943702697754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,16,power_law_1.2,0.044870400428771974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,16,balanced,0.04161066561937332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,16,balanced,0.04375466704368591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,16,power_law_1.2,0.040031999349594116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,16,balanced,0.04370133578777313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,16,balanced,0.0454720010360082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,16,balanced,0.04597333570321401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,16,balanced,0.05206400156021118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,16,balanced,0.05624000231424967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,16,balanced,0.05806399881839752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,16,balanced,0.05811200042565664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,16,balanced,0.06005333364009857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,16,balanced,0.05835199852784475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,16,balanced,0.06217599908510844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,16,balanced,0.06206400195757548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,16,balanced,0.06468800206979115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,16,power_law_1.2,0.035724800825119016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,16,balanced,0.0703359991312027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,16,balanced,0.07172800103823344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,16,balanced,0.07645866771539052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,16,balanced,0.09099200367927551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,16,balanced,0.10449066758155823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,16,balanced,0.12995733817418417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,16,balanced,0.14868266383806863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,16,balanced,0.1898720065752665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,16,power_law_1.2,0.03825919926166534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,16,balanced,0.24093333880106607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,16,balanced,0.3226720094680786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,2,16,power_law_1.01,1.1178175926208496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,16,balanced,0.4026613235473633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,128,power_law_1.01,2.3308671951293944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,16,balanced,0.5790026585261027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,16,balanced,0.743781328201294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,16,power_law_1.2,0.03919999897480011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,16,balanced,0.9226880073547363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,16,balanced,1.4127039909362793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,16,power_law_1.2,0.04291839897632599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,16,balanced,2.7821388244628906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,16,power_law_1.2,0.04469760060310364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,16,power_law_1.2,0.043584001064300534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,16,power_law_1.2,0.04485119879245758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,16,power_law_1.2,0.04549759924411774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,16,power_law_1.2,0.046419200301170346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,16,power_law_1.2,0.048742398619651794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,2,16,power_law_1.01,1.726483154296875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,16,power_law_1.2,0.05225600004196167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,16,power_law_1.2,0.053235197067260744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,16,power_law_1.2,0.05952640175819397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,16,power_law_1.2,0.06178560256958008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,16,power_law_1.2,0.06798719763755798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,16,power_law_1.2,0.07804800271987915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,16,power_law_1.2,0.09673600196838379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,16,power_law_1.2,0.11477760076522828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,16,power_law_1.2,0.1223680019378662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,16,power_law_1.2,0.14403200149536133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,16,power_law_1.2,0.1745471954345703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,16,power_law_1.2,0.2550719976425171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,16,power_law_1.2,0.33912320137023927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,1,64,power_law_1.2,0.026041600108146667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,2,16,power_law_1.01,3.704537582397461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,1,64,power_law_1.2,0.02555519938468933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,16,power_law_1.2,0.47942399978637695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,1,1,balanced,0.2153866688410441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,1,1,balanced,0.43219200770060223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,1,1,balanced,0.7420427004496256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,1,64,power_law_1.2,0.026073598861694337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,1,1,balanced,0.7470986843109131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,1,1,balanced,0.7460693518320719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,1,1,balanced,0.7365012963612875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,1,1,balanced,0.7396799723307291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,1,64,power_law_1.2,0.028575998544692994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,16,power_law_1.2,0.6406400203704834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,1,1,balanced,0.7406933307647705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,1,1,balanced,0.7441333134969076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,1,1,balanced,0.7472586631774902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,1,64,power_law_1.2,0.026572799682617186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,1,1,balanced,0.7881279786427816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,1,1,balanced,0.7606293360392252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,1,1,balanced,0.7670719623565674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,1,1,balanced,0.8169386386871338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,1,64,power_law_1.2,0.02956799864768982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,1,1,balanced,0.813098669052124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,16,power_law_1.2,0.9192959785461425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,1,1,balanced,0.8170239925384521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,1,1,balanced,0.9100746313730875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,1,64,power_law_1.2,0.029363200068473816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,1,1,balanced,0.962394634882609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,1,1,balanced,0.9868853092193604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,1,1,balanced,1.3036692937215169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,1,64,power_law_1.2,0.031167998909950256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,1,1,balanced,1.366165320078532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,1,64,power_law_1.2,0.032543998956680295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,1,1,balanced,1.852666695912679
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,16,power_law_1.2,1.3277952194213867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,1,1,balanced,2.4095253944396973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,1,64,power_law_1.2,0.03307519853115082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,1,64,power_law_1.2,0.033471998572349546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,1,1,balanced,3.530032157897949
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,1,64,power_law_1.2,0.03397760093212128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,1,1,balanced,4.739226659138997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,1,64,power_law_1.2,0.03524479866027832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,1,64,power_law_1.2,0.03272320032119751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,16,power_law_1.2,3.163680076599121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,1,1,balanced,7.021146774291992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,1,64,power_law_1.2,0.036032000184059144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,1,64,power_law_1.2,0.03727999925613403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,1,64,power_law_1.2,0.040755200386047366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,1,1,balanced,9.320485432942709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,1,64,power_law_1.2,0.043007999658584595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,64,power_law_1.2,0.047916799783706665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,64,power_law_1.2,0.05351679921150208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,64,power_law_1.2,0.05575680136680603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,1,1,balanced,11.766613006591797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,64,power_law_1.2,0.06627200245857238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,32,power_law_1.2,0.033471998572349546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,64,power_law_1.2,0.08880640268325805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,64,power_law_1.2,0.11694079637527466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,32,power_law_1.2,0.031007999181747438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,64,power_law_1.2,0.15510400533676147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,32,power_law_1.2,0.030348798632621764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,1,1,balanced,19.311791737874348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,64,power_law_1.2,0.2218303918838501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,32,power_law_1.2,0.030374398827552794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,64,power_law_1.2,0.33126399517059324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,32,power_law_1.2,0.030623999238014222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,64,power_law_1.2,0.4226367950439453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,32,power_law_1.2,0.032441601157188416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,32,power_law_1.2,0.032902398705482484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,64,power_law_1.2,0.6287360191345215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,32,power_law_1.2,0.03258239924907684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,64,power_law_1.2,0.04010879993438721
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,32,power_law_1.2,0.03280639946460724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,64,power_law_1.2,0.03932160139083862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,64,power_law_1.2,0.03374719917774201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,1,1,balanced,37.33179219563802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,32,power_law_1.2,0.032876798510551454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,64,power_law_1.2,1.2096063613891601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,64,power_law_1.2,0.03455359935760498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,64,power_law_1.2,0.03560320138931274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,32,power_law_1.2,0.033792001008987424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,64,power_law_1.2,0.03475199937820435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,32,power_law_1.2,0.03437440097332001
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,64,power_law_1.2,0.03544960021972656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,64,power_law_1.2,0.03506560027599335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,32,power_law_1.2,0.03516159951686859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,64,power_law_1.2,0.03513599932193756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,32,power_law_1.2,0.03644160032272339
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,64,power_law_1.2,0.03548159897327423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,32,power_law_1.2,0.040678399801254275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,64,power_law_1.2,0.03630079925060272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,64,power_law_1.2,0.03720960021018982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,32,power_law_1.2,0.04128639996051788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,64,power_law_1.2,0.03706879913806915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,32,power_law_1.2,0.043756800889968875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,64,power_law_1.2,0.0380160003900528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,32,power_law_1.2,0.05006719827651977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,64,power_law_1.2,0.042080000042915344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,64,power_law_1.2,0.04216319918632507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,32,power_law_1.2,0.05048959851264954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,64,power_law_1.2,0.043756800889968875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,32,power_law_1.2,0.057183998823165896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,64,power_law_1.2,0.04764800071716309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,32,power_law_1.2,0.06945919990539551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,64,power_law_1.2,0.05488640069961548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,64,power_law_1.2,0.060038399696350095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,32,power_law_1.2,0.0893119990825653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,64,power_law_1.2,0.06524159908294677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,32,power_law_1.2,0.10720640420913696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,64,power_law_1.2,0.08343679904937744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,64,power_law_1.2,0.09509119987487794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,32,power_law_1.2,0.1648192048072815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,64,power_law_1.2,0.14573440551757813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,32,power_law_1.2,0.2317568063735962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,64,power_law_1.2,0.16652799844741822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,2,4,power_law_1.2,0.026335999369621277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,1,power_law_1.2,0.05198079943656921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,2,4,power_law_1.2,0.028140801191329955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,64,power_law_1.2,0.24041600227355958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,32,power_law_1.2,0.34019200801849364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,1,power_law_1.2,0.07232000231742859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,2,4,power_law_1.2,0.029529601335525513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,64,power_law_1.2,0.30071680545806884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,1,power_law_1.2,0.07957119941711426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,2,4,power_law_1.2,0.030739200115203858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,32,power_law_1.2,0.45408000946044924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,2,4,power_law_1.2,0.036748799681663516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,2688,1856,6,128,1,64,power_law_1.2,0.42743678092956544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,1,power_law_1.2,0.12197120189666748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,2,4,power_law_1.2,0.04689919948577881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,32,power_law_1.2,0.5105023860931397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,2,4,power_law_1.2,0.04737280011177063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,1,power_law_1.2,0.16054400205612182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,2688,1856,6,128,1,64,power_law_1.2,0.6154111862182617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,2,4,power_law_1.2,0.04886400103569031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,1,power_law_1.2,0.20670719146728517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,2,4,power_law_1.2,0.0502016007900238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,2,4,power_law_1.2,0.05356799960136414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,32,power_law_1.2,0.8463040351867676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,1,power_law_1.2,0.31196160316467286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,2,4,power_law_1.2,0.05541759729385376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,2688,1856,6,128,1,64,power_law_1.2,1.3454463958740235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,2,4,power_law_1.2,0.05750399827957153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,1,power_law_1.2,0.3314239978790283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,2,4,power_law_1.2,0.06376320123672485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,2,4,power_law_1.2,0.05971840023994446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,1,power_law_1.2,0.35043199062347413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,2,4,power_law_1.2,0.06292480230331421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,32,power_law_1.2,2.094528007507324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,1,power_law_1.2,0.3693824052810669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,2,4,power_law_1.2,0.06595839858055115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,2,4,power_law_1.2,0.06829439997673034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,1,power_law_1.2,0.38727679252624514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,2,4,power_law_1.2,0.07775359749794006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,4,power_law_1.2,0.09145600199699402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,1,power_law_1.2,0.3949376106262207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,4,power_law_1.2,0.10065280199050904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,1,power_law_1.2,0.41666560173034667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,4,power_law_1.2,0.1041152000427246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,4,power_law_1.2,0.12872960567474365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,1,power_law_1.2,0.4367936134338379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,4,power_law_1.2,0.177401602268219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,1,power_law_1.2,0.45233922004699706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,4,power_law_1.2,0.20414719581604004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,1,power_law_1.2,0.47054080963134765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,4,power_law_1.2,0.29259519577026366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,1,power_law_1.2,0.5001984119415284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,4,power_law_1.2,0.37271039485931395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,1,power_law_1.2,0.5626944065093994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,4,power_law_1.2,0.4800447940826416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,1,power_law_1.2,0.593996810913086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,4,power_law_1.2,0.589408016204834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,1,power_law_1.2,0.7026368141174316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,2,power_law_1.01,0.0700160026550293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,16,balanced,0.03245333333810171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,4,power_law_1.2,1.0936639785766602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,16,balanced,0.031727999448776245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,16,balanced,0.030970667799313862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,16,balanced,0.03127466638882955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,1,power_law_1.2,0.794975996017456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,16,balanced,0.032960000137488045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,16,balanced,0.0344106654326121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,16,balanced,0.03348266581694285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,16,balanced,0.035301332672437034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,16,balanced,0.03326933334271113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,16,balanced,0.0331839993596077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,2,power_law_1.01,0.09653760194778442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,16,balanced,0.034703999757766724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,16,balanced,0.03457599878311157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,16,balanced,0.03477333237727483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,16,balanced,0.03638399889071783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,16,balanced,0.03878399978081385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,16,balanced,0.038389332592487335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,16,balanced,0.04135466615358988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,1,power_law_1.2,0.9986111640930175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,16,balanced,0.043151999513308205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,16,balanced,0.04528533418973287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,16,balanced,0.04854399959246317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,16,balanced,0.05134400228659312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,16,balanced,0.060789331793785095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,16,balanced,0.06862933437029521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,16,balanced,0.08782933155695598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,16,balanced,0.1088853379090627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,16,balanced,0.1387999951839447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,16,balanced,0.1660426656405131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,16,balanced,0.2035520076751709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,4,power_law_1.2,1.8509567260742188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,16,balanced,0.28487465778986615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,1,power_law_1.2,1.1171456336975099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,2,power_law_1.01,0.1470080018043518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,16,balanced,0.523690660794576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,1,power_law_1.2,1.4447872161865234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,2,power_law_1.01,0.11743359565734864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,1,power_law_1.2,1.776268768310547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,2,power_law_1.01,0.19351680278778077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,1,power_law_1.2,2.4087295532226562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,2,power_law_1.01,0.26766719818115237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,2,power_law_1.01,0.3186559915542603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,1,power_law_1.2,3.044742393493652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,2,power_law_1.01,0.4183551788330078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,1,power_law_1.2,3.7142398834228514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,2,power_law_1.01,0.42427520751953124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,2,power_law_1.01,0.45169920921325685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,1,4,balanced,0.07743466893831889
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,1,4,balanced,0.0981333355108897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,1,4,balanced,0.09763200084368388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,1,4,balanced,0.097871998945872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,1,4,balanced,0.09890666604042053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,1,4,balanced,0.10010666648546855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,1,4,balanced,0.09904000163078308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,1,4,balanced,0.09912000099817912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,1,power_law_1.2,5.5474494934082035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,1,4,balanced,0.10223467151323955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,2,power_law_1.01,0.46524801254272463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,1,4,balanced,0.10167466600735982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,1,4,balanced,0.10242133339246114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,1,4,balanced,0.10241066416104634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,1,4,balanced,0.10296533505121867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,1,4,balanced,0.10719466209411621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,1,4,balanced,0.11733333269755046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,1,4,balanced,0.1184266706307729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,1,4,balanced,0.11973866820335388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,1,4,balanced,0.14127467075983682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,1,4,balanced,0.1444000005722046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,1,4,balanced,0.1709866722424825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,1,4,balanced,0.21092800299326578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,1,4,balanced,0.2664373318354289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,1,4,balanced,0.33530131975809735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,1,4,balanced,0.4566986560821533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,1,4,balanced,0.5693120161692301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,2,power_law_1.01,0.4895040035247803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,1,4,balanced,0.8330612977345785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,1,4,balanced,1.0659626324971516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,1,4,balanced,1.3381013870239258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,1,4,balanced,2.042287985483805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,2,power_law_1.01,0.5074431896209717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,1,4,balanced,3.9369920094807944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,64,power_law_1.01,0.033478400111198424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,2,power_law_1.01,0.5180480003356933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,1,power_law_1.2,10.524940490722656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,64,power_law_1.01,0.03173120021820068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,2,power_law_1.01,0.5441408157348633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,64,power_law_1.01,0.029868799448013305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,2,power_law_1.01,0.561568021774292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,64,power_law_1.01,0.03086079955101013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,64,power_law_1.01,0.03022719919681549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,2,power_law_1.01,0.5849023818969726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,64,power_law_1.01,0.03157120048999786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,2,power_law_1.01,0.6188608169555664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,64,power_law_1.01,0.031763198971748355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,64,power_law_1.01,0.03153280019760132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,2,power_law_1.01,0.6715456008911133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,64,power_law_1.01,0.03167999982833862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,64,power_law_1.01,0.032128000259399415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,2,power_law_1.01,0.7849728107452393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,64,power_law_1.01,0.03338240087032318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,64,power_law_1.01,0.03358719944953918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,2,power_law_1.01,0.7998591899871826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,64,power_law_1.01,0.03379839956760407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,64,power_law_1.01,0.03477759957313538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,2,power_law_1.01,0.9757823944091797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,32,power_law_1.2,0.07167999744415283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,64,power_law_1.01,0.03790079951286316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,32,power_law_1.2,0.05822719931602478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,4,4,power_law_1.2,0.03695360124111176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,4,4,power_law_1.2,0.03672960102558136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,64,power_law_1.01,0.03861120045185089
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,2,power_law_1.01,1.0542783737182617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,4,4,power_law_1.2,0.03454720079898834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,4,4,power_law_1.2,0.03560959994792938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,32,power_law_1.2,0.049004799127578734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,4,4,power_law_1.2,0.035411199927330016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,64,power_law_1.01,0.03933440148830414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,4,4,power_law_1.2,0.0383103996515274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,4,4,power_law_1.2,0.044198399782180785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,4,4,power_law_1.2,0.046367999911308286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,4,4,power_law_1.2,0.047276800870895384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,32,power_law_1.2,0.05160319805145264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,64,power_law_1.01,0.04477440118789673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,4,4,power_law_1.2,0.04749439954757691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,2,power_law_1.01,1.3480768203735352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,4,4,power_law_1.2,0.052019202709198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,4,4,power_law_1.2,0.05322880148887634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,4,4,power_law_1.2,0.05791360139846802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,32,power_law_1.2,0.05135359764099121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,4,4,power_law_1.2,0.05642240047454834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,64,power_law_1.01,0.04612480103969574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,4,4,power_law_1.2,0.06581760048866273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,4,4,power_law_1.2,0.07078400254249573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,4,4,power_law_1.2,0.07979519963264466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,4,4,power_law_1.2,0.10085120201110839
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,64,power_law_1.01,0.05172479748725891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,32,power_law_1.2,0.054655998945236206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,4,4,power_law_1.2,0.10438400506973267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,2,power_law_1.01,1.5032959938049317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,4,4,power_law_1.2,0.1285823941230774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,4,4,power_law_1.2,0.1588736057281494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,4,4,power_law_1.2,0.18293119668960572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,64,power_law_1.01,0.055232000350952146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,32,power_law_1.2,0.05958399772644043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,4,4,power_law_1.2,0.23164799213409423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,4,4,power_law_1.2,0.40244479179382325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,4,4,power_law_1.2,0.475980806350708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,64,power_law_1.01,0.06961280107498169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,32,power_law_1.2,0.0584384024143219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,4,4,power_law_1.2,0.6748735904693604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,2,power_law_1.01,2.010553550720215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,4,4,power_law_1.2,0.8003328323364258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,64,power_law_1.01,0.07837439775466919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,32,power_law_1.2,0.05688959956169128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,4,4,power_law_1.2,1.0945280075073243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,64,power_law_1.01,0.11976319551467896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,32,power_law_1.2,0.060102397203445436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,4,4,power_law_1.2,1.819161605834961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,2,power_law_1.01,2.4322303771972655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,64,power_law_1.01,0.14520319700241088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,32,power_law_1.2,0.061766397953033444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,4,4,power_law_1.2,3.402201461791992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,32,power_law_1.2,0.06442880034446716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,64,power_law_1.01,0.2132159948348999
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,32,power_law_1.2,0.06712960004806519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,2,power_law_1.01,3.015110397338867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,64,power_law_1.01,0.2835200071334839
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,32,power_law_1.2,0.06569600105285645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,32,power_law_1.2,0.07500159740447998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,64,power_law_1.01,0.3176192045211792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,32,power_law_1.2,0.0822272002696991
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,2,power_law_1.01,4.2780414581298825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,64,power_law_1.01,0.5324736118316651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,32,power_law_1.2,0.08584319949150085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,32,power_law_1.2,0.09538559913635254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,32,power_law_1.2,0.105132794380188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,64,power_law_1.01,1.003225612640381
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,32,power_law_1.2,0.13255679607391357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,32,power_law_1.2,0.1410048007965088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,2,power_law_1.01,7.941196441650391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,32,power_law_1.2,0.17336959838867189
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,32,2,balanced,0.0316746657093366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,32,2,balanced,0.03320533285538355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,32,power_law_1.2,0.22624640464782714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,32,2,balanced,0.033615998923778534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,32,2,balanced,0.03501333296298981
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,32,2,balanced,0.03362133353948593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,32,2,balanced,0.034143999218940735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,32,2,balanced,0.03586666782697042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,32,2,balanced,0.035930665830771126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,32,2,balanced,0.03749333322048187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,32,2,balanced,0.03750933210055033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,32,2,balanced,0.037578667203585304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,32,2,balanced,0.03956266740957896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,32,2,balanced,0.0377813329299291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,32,2,balanced,0.039919999738534294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,32,2,balanced,0.047685335079828896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,32,2,balanced,0.04972266654173533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,32,2,balanced,0.05004266897837321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,32,2,balanced,0.058149332801500954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,32,2,balanced,0.07194666564464569
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,32,2,balanced,0.09722666939099629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,32,2,balanced,0.11530133088429768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,32,2,balanced,0.15056000153223673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,32,2,balanced,0.17286932468414307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,32,power_law_1.2,0.2910847902297974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,32,2,balanced,0.24320000410079956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,32,2,balanced,0.3019946614901225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,32,2,balanced,0.4347360134124756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,32,2,balanced,0.5621973276138306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,32,2,balanced,0.6907253265380859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,32,2,balanced,1.0631199677785237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,32,power_law_1.2,0.40007681846618653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,32,2,balanced,2.0756160418192544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,32,power_law_1.2,0.6226240158081054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,32,power_law_1.2,0.7071936130523682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,32,power_law_1.2,0.881497573852539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,8,power_law_1.2,0.0398719996213913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,8,power_law_1.2,0.03537920117378235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,32,power_law_1.2,1.3940544128417969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,8,power_law_1.2,0.0356799989938736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,8,power_law_1.2,0.03825919926166534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,32,power_law_1.2,0.03771519958972931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,8,power_law_1.2,0.03844479918479919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,32,power_law_1.2,0.029203200340270997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,32,power_law_1.2,0.02903040051460266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,8,power_law_1.2,0.04276480078697205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,32,power_law_1.2,2.656716728210449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,32,power_law_1.2,0.02876160144805908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,8,power_law_1.2,0.044012799859046936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,32,power_law_1.2,0.030559998750686646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,32,power_law_1.2,0.031052801012992858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,8,power_law_1.2,0.04410879909992218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,32,power_law_1.2,0.03129599988460541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,8,power_law_1.2,0.043347200751304625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,32,power_law_1.2,0.030752000212669373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,8,power_law_1.2,0.04456959962844849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,32,power_law_1.2,0.0318015992641449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,2,power_law_1.01,0.08486400246620178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,32,power_law_1.2,0.032179200649261476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,8,power_law_1.2,0.04578559994697571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,32,power_law_1.2,0.03383040130138397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,8,power_law_1.2,0.04860160052776337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,32,power_law_1.2,0.03470079898834229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,2,power_law_1.01,0.15638400316238404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,8,power_law_1.2,0.04873600006103516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,32,power_law_1.2,0.03487359881401062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,32,power_law_1.2,0.03619199991226196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,8,power_law_1.2,0.052070397138595584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,32,power_law_1.2,0.03932799994945526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,2,power_law_1.01,0.09964799880981445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,8,power_law_1.2,0.059462398290634155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,32,power_law_1.2,0.039724799990653994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,32,power_law_1.2,0.039904001355171206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,8,power_law_1.2,0.059411197900772095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,32,power_law_1.2,0.0443583995103836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,2,power_law_1.01,0.15608320236206055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,8,power_law_1.2,0.06782720088958741
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,32,power_law_1.2,0.048390400409698484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,8,power_law_1.2,0.07642880082130432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,32,power_law_1.2,0.05724800229072571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,32,power_law_1.2,0.0661184012889862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,8,power_law_1.2,0.08460800051689148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,2,power_law_1.01,0.2074944019317627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,32,power_law_1.2,0.10199040174484253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,8,power_law_1.2,0.10090880393981934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,32,power_law_1.2,0.09436799883842469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,4,balanced,0.05222400029500326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,4,balanced,0.08070399860541026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,4,balanced,0.0906773308912913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,4,balanced,0.1280586620171865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,8,power_law_1.2,0.12643200159072876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,4,balanced,0.19819200038909912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,2,power_law_1.01,0.3258368015289307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,4,balanced,0.2577600081761678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,4,balanced,0.2579466700553894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,4,balanced,0.25839465856552124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,32,power_law_1.2,0.15406719446182252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,4,balanced,0.2590666611989339
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,4,balanced,0.25995200872421265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,4,balanced,0.26478399833043414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,4,balanced,0.26837867498397827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,4,balanced,0.26928534110387164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,4,balanced,0.27246399720509845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,8,power_law_1.2,0.16728960275650023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,4,balanced,0.27933865785598755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,4,balanced,0.2844906648000081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,32,power_law_1.2,0.2070080041885376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,4,balanced,0.2929333249727885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,4,balanced,0.3150773247083028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,4,balanced,0.3300693432490031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,4,balanced,0.4231733481089274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,4,balanced,0.41578133900960285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,4,balanced,0.5939093430836996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,2,power_law_1.01,0.35498239994049074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,4,balanced,0.5682719945907593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,32,power_law_1.2,0.2768127918243408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,4,balanced,0.8178186416625977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,8,power_law_1.2,0.18937599658966064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,4,balanced,0.8998560110727946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,4,balanced,1.3543465932210286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,4,balanced,1.5539306004842122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,32,power_law_1.2,0.3684544086456299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,2,power_law_1.01,0.3764672040939331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,2,4,balanced,1.9790347417195637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,8,power_law_1.2,0.27562880516052246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,2,4,balanced,2.9732373555501304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,32,power_law_1.2,0.5216896057128906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,2,power_law_1.01,0.3831935882568359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,2,4,balanced,5.762933095296224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,8,power_law_1.2,0.357587194442749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,32,power_law_1.2,0.7695551872253418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,2,power_law_1.01,0.39762558937072756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,8,power_law_1.2,0.4776000022888184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,8,power_law_1.2,0.7081344127655029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,2,power_law_1.01,0.4084671974182129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,32,power_law_1.2,2.232512092590332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,8,power_law_1.2,0.8640895843505859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,2,power_law_1.01,0.4249983787536621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,2,power_law_1.01,0.4201536178588867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,8,power_law_1.2,1.469593620300293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,2,power_law_1.01,0.4328000068664551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,2,power_law_1.01,0.4576767921447754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,8,power_law_1.2,2.823334312438965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,2,power_law_1.01,0.47336320877075194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,32,balanced,0.04168533285458883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,32,balanced,0.03994666785001755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,32,balanced,0.043712000052134194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,32,balanced,0.0417546679576238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,32,balanced,0.043663998444875084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,32,balanced,0.04599999884764353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,32,balanced,0.04765866696834564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,32,balanced,0.048021331429481506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,32,balanced,0.049829334020614624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,32,balanced,0.04811733464399973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,32,balanced,0.04966400067011515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,32,balanced,0.0517439991235733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,32,balanced,0.04994133114814758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,32,balanced,0.053247998158137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,32,balanced,0.05717333157857259
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,32,balanced,0.05789333085219065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,32,balanced,0.060090666015942894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,32,balanced,0.06423999865849812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,32,balanced,0.06818666557470958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,32,balanced,0.08462933699289958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,32,balanced,0.10122666756312053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,2,power_law_1.01,0.5047423839569092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,32,balanced,0.1399946709473928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,32,balanced,0.16683199008305868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,32,balanced,0.22007467349370322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,32,balanced,0.26411734024683636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,32,balanced,0.3714880148569743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,32,balanced,0.4740906556447347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,32,balanced,0.5654986699422201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,32,balanced,0.8670240243275961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,32,balanced,1.691450595855713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,2,power_law_1.01,0.5740223884582519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,8,power_law_1.01,0.06369919776916504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,2,power_law_1.01,0.5908544063568115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,8,power_law_1.01,0.08939520120620728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,8,power_law_1.01,0.05867519974708557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,2,power_law_1.01,0.6928832054138183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,8,power_law_1.01,0.07022719979286193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,8,power_law_1.01,0.08190720081329346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,2,power_law_1.01,0.7743231773376464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,8,power_law_1.01,0.08807680010795593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,1,balanced,0.0960053304831187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,1,balanced,0.10146133104960124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,1,balanced,0.12754666805267334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,1,balanced,0.1680319905281067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,1,balanced,0.250165323416392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,1,balanced,0.4081386725107829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,8,power_law_1.01,0.10182399749755859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,1,balanced,0.5680960019429525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,1,balanced,0.5673066775004069
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,1,balanced,0.5698293447494507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,1,balanced,0.5662293434143066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,2,power_law_1.01,0.9567487716674805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,1,balanced,0.5706400076548258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,1,balanced,0.5784639914830526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,8,power_law_1.01,0.10448000431060792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,1,balanced,0.579589327176412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,1,balanced,0.5839306513468424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,1,balanced,0.5873653491338094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,1,balanced,0.5954453150431315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,1,balanced,0.6147733529408773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,8,power_law_1.01,0.10817919969558716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,1,balanced,0.6591093142827352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,1,balanced,0.6968479951222738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,1,balanced,0.7724746863047282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,1,balanced,0.8421813646952311
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,1,balanced,0.997973362604777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,2,power_law_1.01,1.0560192108154296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,8,power_law_1.01,0.107750403881073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,1,balanced,1.1524053414662678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,1,balanced,1.4813332557678223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,8,power_law_1.01,0.10881919860839843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,1,balanced,1.97762664159139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,1,balanced,2.589146614074707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,8,power_law_1.01,0.11528960466384888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,1,balanced,3.5082826614379883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,2,power_law_1.01,1.35348482131958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,1,power_law_1.01,0.08305919766426087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,8,power_law_1.01,0.11719679832458496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,1,balanced,4.396938641866048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,8,power_law_1.01,0.11706240177154541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,1,balanced,6.633024215698242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,1,power_law_1.01,0.11843839883804322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,8,power_law_1.01,0.12985600233078004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,2,power_law_1.01,1.720902442932129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,8,power_law_1.01,0.13393919467926024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,1,power_law_1.01,0.1410048007965088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,1,balanced,12.531183878580729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,8,power_law_1.01,0.14098559617996215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,8,power_law_1.01,0.1545151948928833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,1,power_law_1.01,0.24066560268402098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,2,power_law_1.01,2.4142528533935548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,8,power_law_1.01,0.16918400526046753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,8,power_law_1.01,0.19960960149765014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,1,power_law_1.01,0.3506752014160156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,8,power_law_1.01,0.22539520263671875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,1,power_law_1.01,0.6015679836273193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,8,power_law_1.01,0.2739840030670166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,2,power_law_1.01,2.9921600341796877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,8,power_law_1.01,0.3253119945526123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,1,power_law_1.01,0.6565184116363525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,8,power_law_1.01,0.4256896018981934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,1,power_law_1.01,0.6878208160400391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,8,power_law_1.01,0.526963186264038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,4,2,power_law_1.01,3.606662368774414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,8,power_law_1.01,0.7772352218627929
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,1,power_law_1.01,0.6998400211334228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,8,power_law_1.01,0.9928128242492675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,1,power_law_1.01,0.7367936134338379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,8,power_law_1.01,1.1935104370117187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,1,power_law_1.01,0.750271987915039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,4,2,power_law_1.01,5.512723159790039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,1,power_law_1.01,0.7735295772552491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,2,power_law_1.01,0.11342079639434814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,8,power_law_1.01,1.9492864608764648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,1,power_law_1.01,0.7953536033630371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,1,power_law_1.01,0.8249343872070313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,2,power_law_1.01,0.12449920177459717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,8,power_law_1.01,3.583660888671875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,1,power_law_1.01,0.8527359962463379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,2,power_law_1.01,0.11272959709167481
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,1,power_law_1.01,0.8782143592834473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,4,2,power_law_1.01,10.998239898681641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,1,power_law_1.01,0.9419008255004883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,2,power_law_1.01,0.15649280548095704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,1,power_law_1.01,1.0692928314208985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,2,power_law_1.01,0.19279359579086303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,1,power_law_1.01,1.086336040496826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,2,power_law_1.01,0.2801471948623657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,1,power_law_1.01,1.272550392150879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,32,power_law_1.2,0.06445440053939819
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,2,power_law_1.01,0.3018431901931763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,1,power_law_1.01,1.3155391693115235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,32,power_law_1.2,0.0572160005569458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,32,power_law_1.2,0.053625601530075076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,32,power_law_1.2,0.05483520030975342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,1,power_law_1.01,1.6053951263427735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,2,power_law_1.01,0.3119231939315796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,32,power_law_1.2,0.055244797468185426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,32,power_law_1.2,0.05804160237312317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,32,power_law_1.01,0.04888319969177246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,32,power_law_1.2,0.05868800282478333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,1,power_law_1.01,1.872115135192871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,32,power_law_1.01,0.048281601071357726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,32,power_law_1.2,0.0593280017375946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,2,power_law_1.01,0.3291584014892578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,32,power_law_1.01,0.044268798828125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,32,power_law_1.2,0.059571200609207155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,32,power_law_1.01,0.04431999921798706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,32,power_law_1.2,0.060096001625061034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,32,power_law_1.01,0.045612800121307376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,32,power_law_1.2,0.06095359921455383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,1,power_law_1.01,2.441676712036133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,32,power_law_1.01,0.04712960124015808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,2,power_law_1.01,0.3295167922973633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,32,power_law_1.2,0.060915201902389526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,32,power_law_1.01,0.048563200235366824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,32,power_law_1.2,0.06175360083580017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,32,power_law_1.01,0.049439999461174014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,32,power_law_1.2,0.06343680024147033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,32,power_law_1.01,0.04940159916877747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,32,power_law_1.2,0.06547200083732604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,2,power_law_1.01,0.33799679279327394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,32,power_law_1.01,0.050470399856567386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,1,power_law_1.01,3.1985471725463865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,32,power_law_1.2,0.06821119785308838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,32,power_law_1.01,0.05130879878997803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,32,power_law_1.2,0.07382400035858154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,32,power_law_1.01,0.05435519814491272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,32,power_law_1.2,0.08918399810791015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,32,power_law_1.01,0.05432320237159729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,2,power_law_1.01,0.35011839866638184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,32,power_law_1.2,0.09624320268630981
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,32,power_law_1.01,0.057548797130584715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,32,power_law_1.01,0.06320639848709106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,32,power_law_1.2,0.11419520378112794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,32,power_law_1.01,0.06625279784202576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,1,power_law_1.01,4.364076614379883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,32,power_law_1.2,0.131769597530365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,32,power_law_1.01,0.07047039866447449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,2,power_law_1.01,0.3591552019119263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,32,power_law_1.2,0.16513919830322266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,32,power_law_1.01,0.08631680011749268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,32,power_law_1.2,0.20317440032958983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,32,power_law_1.01,0.09173120260238647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,32,power_law_1.2,0.2850816011428833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,32,power_law_1.01,0.11192320585250855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,2,power_law_1.01,0.37112960815429685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,32,power_law_1.2,0.34967041015625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,32,power_law_1.01,0.14188799858093262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,1,power_law_1.01,5.4502910614013675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,32,power_law_1.01,0.17989759445190429
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,32,power_law_1.2,0.5034304141998291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,32,power_law_1.01,0.21096320152282716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,2,power_law_1.01,0.3880575895309448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,32,power_law_1.2,0.6605823993682861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,32,power_law_1.01,0.2944000005722046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,32,power_law_1.01,0.3598848104476929
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,32,power_law_1.2,0.908569622039795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,2,power_law_1.01,0.412556791305542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,32,power_law_1.01,0.506547212600708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,4,1,power_law_1.01,6.536998748779297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,32,power_law_1.2,1.6488000869750976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,32,power_law_1.01,0.6902783870697021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,2,power_law_1.01,0.43795199394226075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,32,power_law_1.01,0.8402112007141114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,2,power_law_1.01,0.5101696014404297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,32,power_law_1.2,3.153798484802246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,32,power_law_1.01,1.2403008460998535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,2,power_law_1.01,0.5758656024932861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,4,1,power_law_1.01,10.024825286865234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,32,power_law_1.01,2.6644927978515627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,2,power_law_1.01,0.7009727954864502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,2,power_law_1.01,0.8039744377136231
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,2,power_law_1.01,1.061177635192871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,32,8,power_law_1.01,0.03586559891700745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,32,8,power_law_1.01,0.03285120129585266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,2,power_law_1.01,1.2712063789367676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,32,8,power_law_1.01,0.03155199885368347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,32,8,power_law_1.01,0.030630400776863097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,32,8,power_law_1.01,0.03178240060806274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,32,8,power_law_1.01,0.031641599535942075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,32,8,power_law_1.01,0.03283199965953827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,32,8,power_law_1.01,0.032339200377464294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,32,8,power_law_1.01,0.03304960131645203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,4,1,power_law_1.01,19.329779052734374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,32,8,power_law_1.01,0.033164799213409424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,32,8,power_law_1.01,0.033657601475715636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,32,8,power_law_1.01,0.03454079926013946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,32,8,power_law_1.01,0.03468799889087677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,32,8,balanced,0.045834665497144066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,32,8,power_law_1.01,0.03660799860954285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,32,8,balanced,0.045408000548680626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,32,8,balanced,0.047824000318845115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,32,8,balanced,0.04974933465321859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,32,8,power_law_1.01,0.04309119880199432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,32,8,balanced,0.05395199855168661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,32,8,balanced,0.06507200002670288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,32,8,balanced,0.07955199976762135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,32,8,balanced,0.08303466439247131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,32,8,power_law_1.01,0.041094401478767396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,32,8,balanced,0.08275199929873149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,32,8,balanced,0.08406933148701985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,32,8,balanced,0.08273600041866302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,32,8,power_law_1.01,0.044915199279785156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,32,8,balanced,0.08477333188056946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,2,power_law_1.01,1.7243392944335938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,32,8,balanced,0.08599467078844707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,32,8,balanced,0.08686932921409607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,32,8,balanced,0.09147733449935913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,32,8,power_law_1.01,0.052160000801086424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,32,8,balanced,0.09521599610646565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,32,8,balanced,0.09864532947540283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,32,8,balanced,0.11146666606267293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,32,8,power_law_1.01,0.06074879765510559
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,32,8,balanced,0.11945066849390666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,32,8,balanced,0.13492799798647562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,32,8,balanced,0.15059199929237366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,32,8,power_law_1.01,0.07200000286102295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,32,8,balanced,0.18348799149195352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,32,8,balanced,0.23322133223215738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,32,8,power_law_1.01,0.08646399974822998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,32,8,balanced,0.3076533277829488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,32,8,balanced,0.3847413460413615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,32,8,power_law_1.01,0.12223999500274658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,32,8,balanced,0.5464533170064291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,32,8,power_law_1.01,0.1352959990501404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,32,8,balanced,0.6823840141296387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,32,8,balanced,0.8609440326690674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,32,8,power_law_1.01,0.1872063994407654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,32,8,balanced,1.2851093610127766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,32,8,power_law_1.01,0.23292160034179688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,32,8,power_law_1.01,0.30928640365600585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,32,8,balanced,2.4948533376057944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,32,8,power_law_1.01,0.3564608097076416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,32,8,power_law_1.01,0.45530238151550295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,2,power_law_1.01,2.1596927642822266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,32,8,power_law_1.01,0.7157375812530518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,32,8,power_law_1.01,1.342521572113037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,2,power_law_1.01,3.120787239074707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,4,2,power_law_1.2,0.025119999051094057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,2,power_law_1.01,3.9463871002197264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,4,2,power_law_1.2,0.02736000120639801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,4,2,power_law_1.2,0.029971200227737426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,4,2,power_law_1.2,0.03240320086479187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,4,2,power_law_1.2,0.03741439878940582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,4,2,power_law_1.2,0.04796159863471985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,4,2,power_law_1.2,0.05119360089302063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,4,2,power_law_1.2,0.05262719988822937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,4,2,power_law_1.2,0.05280640125274658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,4,power_law_1.2,0.056857597827911374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,4,2,power_law_1.2,0.05542399883270264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,2,2,power_law_1.01,4.797439956665039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,4,2,power_law_1.2,0.05633280277252197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,4,power_law_1.2,0.06213759779930115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,16,8,power_law_1.2,0.03813120126724243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,4,2,power_law_1.2,0.058182400465011594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,16,8,power_law_1.2,0.03648000061511993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,16,8,power_law_1.2,0.034892800450325015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,4,2,power_law_1.2,0.05909119844436646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,16,8,power_law_1.2,0.033958399295806886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,16,8,power_law_1.2,0.035411199927330016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,4,power_law_1.2,0.05607039928436279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,16,8,power_law_1.2,0.0353983998298645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,4,2,power_law_1.2,0.06711040139198303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,16,8,power_law_1.2,0.03578239977359772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,16,8,power_law_1.2,0.036371201276779175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,16,8,power_law_1.2,0.03687680065631867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,4,2,power_law_1.2,0.07186560034751892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,16,8,power_law_1.2,0.03668479919433594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,4,power_law_1.2,0.0648959994316101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,16,8,power_law_1.2,0.037510401010513304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,16,8,power_law_1.2,0.03830400109291077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,4,2,power_law_1.2,0.07626240253448487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,16,8,power_law_1.2,0.0385343998670578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,16,8,power_law_1.2,0.04080640077590943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,16,8,power_law_1.2,0.04577920138835907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,4,power_law_1.2,0.0715071976184845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,4,2,power_law_1.2,0.07895680069923401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,16,8,power_law_1.2,0.0471807986497879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,16,8,power_law_1.2,0.051020801067352295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,4,2,power_law_1.2,0.08588799834251404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,16,8,power_law_1.2,0.0593280017375946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,16,8,power_law_1.2,0.06250879764556885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,4,power_law_1.2,0.08127359747886657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,16,8,power_law_1.2,0.07733759880065919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,2,power_law_1.2,0.10203520059585572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,16,8,power_law_1.2,0.09455999732017517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,16,8,power_law_1.2,0.12074879407882691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,2,power_law_1.2,0.11813119649887086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,16,8,power_law_1.2,0.14814079999923707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,4,power_law_1.2,0.10051200389862061
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,16,8,power_law_1.2,0.2035775899887085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,2,power_law_1.2,0.12613760232925414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,16,8,power_law_1.2,0.24200959205627443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,16,8,power_law_1.2,0.34012160301208494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,4,power_law_1.2,0.10520960092544555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,2,power_law_1.2,0.1506495952606201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,2,2,power_law_1.01,7.690982055664063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,16,8,power_law_1.2,0.4449920177459717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,16,8,power_law_1.2,0.5168000221252441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,2,power_law_1.2,0.20655360221862792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,4,power_law_1.2,0.10514559745788574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,16,8,power_law_1.2,0.803264045715332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,2,power_law_1.2,0.27492480278015136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,4,power_law_1.2,0.11034879684448243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,16,8,power_law_1.2,1.5876992225646973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,2,power_law_1.2,0.35167999267578126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,4,power_law_1.2,0.10920319557189942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,2,power_law_1.2,0.5405119895935059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,4,power_law_1.2,0.11473280191421509
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,2,power_law_1.2,0.6860928058624267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,4,power_law_1.2,0.11763839721679688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,4,power_law_1.2,0.1257024049758911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,2,power_law_1.2,0.8465215682983398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,4,power_law_1.2,0.13515520095825195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,2,power_law_1.2,1.3041728019714356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,4,power_law_1.2,0.1431167960166931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,4,power_law_1.2,0.13822720050811768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,4,power_law_1.2,0.15425920486450195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,2,power_law_1.2,2.500774383544922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,4,power_law_1.2,0.17587200403213502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,2,2,power_law_1.01,14.487884521484375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,4,power_law_1.2,0.20324480533599854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,4,power_law_1.2,0.2336064100265503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,4,power_law_1.2,0.3096447944641113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,4,power_law_1.2,0.3612031936645508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,32,8,power_law_1.01,0.05480960011482239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,32,8,power_law_1.01,0.060838401317596436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,4,power_law_1.2,0.5367167949676513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,32,8,power_law_1.01,0.05324800014495849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,4,power_law_1.2,0.6033088207244873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,32,8,power_law_1.01,0.05639680027961731
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,4,power_law_1.2,0.8887552261352539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,32,8,power_law_1.01,0.060127997398376466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,32,8,power_law_1.01,0.062431997060775755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,4,power_law_1.2,1.303116798400879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,32,8,power_law_1.01,0.06670079827308655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,32,8,power_law_1.01,0.06865919828414917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,4,power_law_1.2,1.4951552391052245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,1,balanced,0.24553066492080688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,32,8,power_law_1.01,0.06824319958686828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,1,balanced,0.32233599821726483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,1,balanced,0.5900693337122599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,1,balanced,1.1270293394724529
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,32,power_law_1.01,0.06369919776916504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,1,balanced,2.164357344309489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,32,8,power_law_1.01,0.07059199810028076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,1,balanced,3.1898721059163413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,4,power_law_1.2,2.306879997253418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,32,8,power_law_1.01,0.0700543999671936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,1,balanced,3.1941067377726235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,32,power_law_1.01,0.04344319999217987
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,1,balanced,3.1957438786824546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,32,8,power_law_1.01,0.07429119944572449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,1,balanced,3.1971254348754883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,32,power_law_1.01,0.039212799072265624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,32,8,power_law_1.01,0.07445759773254394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,1,balanced,3.2103093465169272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,1,balanced,3.2277758916219077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,32,power_law_1.01,0.04307200014591217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,32,8,power_law_1.01,0.07633919715881347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,1,balanced,3.2414881388346353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,4,power_law_1.2,4.455193710327149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,32,power_law_1.01,0.04836480021476745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,1,balanced,3.2534666061401367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,32,8,power_law_1.01,0.08455680012702942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,1,balanced,3.2735039393107095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,32,power_law_1.01,0.05488640069961548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,32,8,power_law_1.01,0.08632959723472595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,1,balanced,3.3026558558146157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,1,balanced,3.322986602783203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,32,power_law_1.01,0.05372160077095032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,32,8,power_law_1.01,0.09621760249137878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,1,balanced,3.3688265482584634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,32,power_law_1.01,0.054079997539520266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,32,8,power_law_1.01,0.1109887957572937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,1,balanced,3.8783626556396484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,32,power_law_1.01,0.053932797908782956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,1,balanced,3.552432060241699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,32,8,power_law_1.01,0.12454400062561036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,32,power_law_1.01,0.05487359762191772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,1,balanced,4.422437349955241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,32,8,power_law_1.01,0.15113600492477416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,1,balanced,3.916597366333008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,32,power_law_1.01,0.055251199007034305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,32,8,power_law_1.01,0.17175040245056153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,32,power_law_1.01,0.05812479853630066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,1,balanced,5.792325337727864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,32,power_law_1.01,0.0588096022605896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,32,8,power_law_1.01,0.2182080030441284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,1,balanced,4.509344100952148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,32,power_law_1.01,0.061868798732757566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,32,8,power_law_1.01,0.2659008026123047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,1,balanced,6.883706410725911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,32,power_law_1.01,0.06606720089912414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,32,8,power_law_1.01,0.3553152084350586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,32,power_law_1.01,0.06753280162811279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,1,balanced,6.729434967041016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,32,power_law_1.01,0.07178879976272583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,32,8,power_law_1.01,0.4376959800720215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,32,power_law_1.01,0.07605119943618774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,16,balanced,0.0580320010582606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,1,balanced,11.817957560221354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,16,balanced,0.09360000491142273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,16,balanced,0.10385066270828247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,16,balanced,0.15709867080052695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,16,balanced,0.22113066911697388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,16,balanced,0.25273066759109497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,32,8,power_law_1.01,0.6562943935394288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,32,power_law_1.01,0.08346239924430847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,16,balanced,0.25439999500910443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,16,balanced,0.2534346580505371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,16,balanced,0.2537013292312622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,16,balanced,0.2540213267008464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,16,balanced,0.2541813254356384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,16,balanced,0.25628799200057983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,16,balanced,0.2592373291651408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,16,balanced,0.25937066475550336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,16,balanced,0.2665013273557027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,16,balanced,0.26874667406082153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,32,power_law_1.01,0.09164800047874451
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,16,balanced,0.26975999275843304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,16,balanced,0.30089600880940753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,16,balanced,0.2947840094566345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,16,balanced,0.387445330619812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,16,balanced,0.343722661336263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,1,balanced,10.339706420898438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,16,balanced,0.48547200361887616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,32,8,power_law_1.01,0.846617603302002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,16,balanced,0.4119573434193929
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,32,power_law_1.01,0.10548479557037353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,16,balanced,0.630181352297465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,16,balanced,0.5900586843490601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,16,balanced,0.9008800188700358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,16,balanced,0.8976693153381348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,32,power_law_1.01,0.12645119428634644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,1,16,balanced,1.1538773377736409
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,1,16,balanced,1.6305707295735676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,32,8,power_law_1.01,0.9765567779541016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,32,power_law_1.01,0.1530303955078125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,1,16,balanced,3.165034612019857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,1,1,balanced,14.20864486694336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,32,power_law_1.01,0.18579839468002318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,32,8,power_law_1.01,1.6113216400146484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,32,power_law_1.01,0.26387200355529783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,32,power_law_1.01,0.3074944019317627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,1,1,balanced,20.07910410563151
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,32,power_law_1.01,0.48730878829956054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,32,8,power_law_1.01,3.0954944610595705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,32,power_law_1.01,0.5439104080200196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,32,power_law_1.01,0.7223616123199463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,32,power_law_1.01,1.3696640014648438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,1,1,balanced,37.66114044189453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,64,power_law_1.2,0.12440320253372192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,64,power_law_1.2,0.14334720373153687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,64,power_law_1.2,0.0995199978351593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,64,power_law_1.2,0.09590399861335755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,64,power_law_1.2,0.10795520544052124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,1,power_law_1.2,0.047040000557899475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,64,power_law_1.2,0.11874560117721558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,1,power_law_1.2,0.057126402854919434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,64,power_law_1.2,0.12165759801864624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,1,power_law_1.2,0.06570879817008972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,64,power_law_1.2,0.1261504054069519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,1,power_law_1.2,0.0818880021572113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,1,power_law_1.2,0.09911680221557617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,64,power_law_1.2,0.1202239990234375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,1,power_law_1.2,0.15228159427642823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,64,power_law_1.2,0.12469120025634765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,1,power_law_1.2,0.15954560041427612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,4,power_law_1.2,0.03268480002880096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,1,power_law_1.2,0.1712000012397766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,64,power_law_1.2,0.12540160417556762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,4,power_law_1.2,0.030828800797462464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,1,power_law_1.2,0.17322239875793458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,4,power_law_1.2,0.029734399914741517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,1,power_law_1.2,0.17523200511932374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,64,power_law_1.2,0.12318079471588135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,4,power_law_1.2,0.030483201146125793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,1,power_law_1.2,0.1827712059020996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,4,power_law_1.2,0.032979199290275575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,64,power_law_1.2,0.12400000095367432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,1,power_law_1.2,0.18865280151367186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,4,power_law_1.2,0.03661440014839172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,1,power_law_1.2,0.19144320487976074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,64,power_law_1.2,0.12341760396957398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,4,power_law_1.2,0.037625598907470706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,1,power_law_1.2,0.19553279876708984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,1,power_law_1.01,0.14968960285186766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,4,power_law_1.2,0.03842560052871704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,64,power_law_1.2,0.13263360261917115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,1,power_law_1.2,0.20716800689697265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,4,power_law_1.2,0.03902080059051514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,1,power_law_1.2,0.21466240882873536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,4,power_law_1.2,0.03883520066738129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,64,power_law_1.2,0.1345088005065918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,4,power_law_1.2,0.04273920059204102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,1,power_law_1.2,0.2265023946762085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,4,power_law_1.2,0.045516800880432126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,64,power_law_1.2,0.15041919946670532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,1,power_law_1.2,0.2642751932144165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,4,power_law_1.2,0.04643200039863586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,1,power_law_1.2,0.28901119232177735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,1,power_law_1.01,0.18586879968643188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,64,power_law_1.2,0.16168960332870483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,4,power_law_1.2,0.04769920110702515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,1,power_law_1.2,0.3530303955078125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,4,power_law_1.2,0.052275198698043826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,64,power_law_1.2,0.174399995803833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,1,power_law_1.2,0.41333761215209963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,4,power_law_1.2,0.05740799903869629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,4,power_law_1.2,0.06304640173912049
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,1,power_law_1.2,0.5521728038787842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,64,power_law_1.2,0.20144639015197754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,4,power_law_1.2,0.06869120001792908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,1,power_law_1.2,0.682425594329834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,1,power_law_1.01,0.2442944049835205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,4,power_law_1.2,0.07854080200195312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,64,power_law_1.2,0.23955199718475342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,4,power_law_1.2,0.10080000162124633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,1,power_law_1.2,0.9311488151550293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,64,power_law_1.2,0.31383678913116453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,4,power_law_1.2,0.11221120357513428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,1,power_law_1.2,1.1810751914978028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,4,power_law_1.2,0.14245760440826416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,64,power_law_1.2,0.37804160118103025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,1,power_law_1.01,0.4387392044067383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,4,power_law_1.2,0.1890112042427063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,1,power_law_1.2,1.687129592895508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,4,power_law_1.2,0.26679039001464844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,64,power_law_1.2,0.49015040397644044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,4,power_law_1.2,0.33060479164123535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,1,power_law_1.2,2.183456039428711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,64,power_law_1.2,0.6691008090972901
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,4,power_law_1.2,0.49755520820617677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,1,power_law_1.01,0.6503551959991455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,1,power_law_1.2,2.6944576263427735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,64,power_law_1.2,0.8343104362487793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,4,power_law_1.2,0.7519040107727051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,4,power_law_1.2,0.8791616439819336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,64,power_law_1.2,1.1666303634643556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,1,power_law_1.01,1.1350784301757812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,1,power_law_1.2,4.149446487426758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,4,power_law_1.2,1.2759936332702637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,64,power_law_1.2,1.4929984092712403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,1,power_law_1.01,1.2277824401855468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,64,power_law_1.2,2.541657638549805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,4,power_law_1.2,2.3681535720825195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,1,power_law_1.2,8.100147247314453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,1,power_law_1.01,1.2905792236328124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,1,power_law_1.01,1.3395008087158202
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,64,power_law_1.2,6.049599838256836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,1,power_law_1.01,1.3842368125915527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,1,power_law_1.01,1.4288127899169922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,1,power_law_1.01,1.4604991912841796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,1,power_law_1.01,1.5085887908935547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,2,8,power_law_1.2,0.056435197591781616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,2,8,power_law_1.2,0.05674239993095398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,16,2,power_law_1.01,0.0364544004201889
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,2,8,power_law_1.2,0.05559039711952209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,16,2,power_law_1.01,0.03669120073318481
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,2,8,power_law_1.2,0.05608320236206055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,16,2,power_law_1.01,0.036345601081848145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,2,8,power_law_1.2,0.05628799796104431
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,16,2,power_law_1.01,0.0383296012878418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,2,8,power_law_1.2,0.05804160237312317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,16,2,power_law_1.01,0.037887999415397645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,2,8,power_law_1.2,0.06298239827156067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,16,2,power_law_1.01,0.0387584000825882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,2,8,power_law_1.2,0.06275200247764587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,16,2,power_law_1.01,0.0392192006111145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,2,8,power_law_1.2,0.06433280110359192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,16,2,power_law_1.01,0.03945600092411041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,2,8,power_law_1.2,0.06581760048866273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,1,power_law_1.01,1.5520959854125977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,16,2,power_law_1.01,0.03984000086784363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,2,8,power_law_1.2,0.06782720088958741
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,16,2,power_law_1.01,0.03991680145263672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,2,8,power_law_1.2,0.07276160120964051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,16,2,power_law_1.01,0.041894400119781496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,2,8,power_law_1.2,0.07511680126190186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,16,2,power_law_1.01,0.04318079948425293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,2,8,power_law_1.2,0.0760703980922699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,16,2,power_law_1.01,0.044121599197387694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,2,8,power_law_1.2,0.0870464026927948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,16,2,power_law_1.01,0.047539201378822324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,16,2,power_law_1.01,0.05475199818611145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,2,8,power_law_1.2,0.0918720006942749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,16,2,power_law_1.01,0.05849599838256836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,2,8,power_law_1.2,0.11455999612808228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,16,2,power_law_1.01,0.05966079831123352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,2,8,power_law_1.2,0.1412160038948059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,16,2,power_law_1.01,0.07136639952659607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,2,8,power_law_1.2,0.1587007999420166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,32,power_law_1.2,0.05987840294837952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,16,2,power_law_1.01,0.08487679958343505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,2,8,power_law_1.2,0.20956799983978272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,16,2,power_law_1.01,0.10517760515213012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,2,8,power_law_1.2,0.24071040153503417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,1,power_law_1.01,1.589087963104248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,16,2,power_law_1.01,0.11948800086975098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,2,8,power_law_1.2,0.3249919891357422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,16,2,power_law_1.01,0.1429759979248047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,32,power_law_1.2,0.053337597846984865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,2,8,power_law_1.2,0.42230401039123533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,16,2,power_law_1.01,0.17452800273895264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,16,2,power_law_1.01,0.250220799446106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,2,8,power_law_1.2,0.6760640144348145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,16,2,power_law_1.01,0.33064320087432864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,2,8,power_law_1.2,0.8336447715759278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,32,power_law_1.2,0.05050879716873169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,16,2,power_law_1.01,0.47246079444885253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,2,8,power_law_1.2,1.1061696052551269
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,16,2,power_law_1.01,0.5091904163360595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,32,power_law_1.2,0.050470399856567386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,16,2,power_law_1.01,0.8212287902832032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,1,power_law_1.01,1.6235776901245118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,2,8,power_law_1.2,1.7550016403198243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,16,2,power_law_1.01,0.9597439765930176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,32,power_law_1.2,0.0530239999294281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,2,8,power_law_1.2,2.316921615600586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,16,2,power_law_1.01,2.315750312805176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,32,power_law_1.2,0.05399680137634277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,2,8,power_law_1.2,3.3151744842529296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,1,power_law_1.01,1.7509376525878906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,32,power_law_1.2,0.05516160130500793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,32,power_law_1.2,0.05562880039215088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,2,8,power_law_1.2,6.041376113891602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,32,power_law_1.2,0.05543680191040039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,1,power_law_1.01,1.969977569580078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,32,power_law_1.2,0.05658239722251892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,32,power_law_1.2,0.057068800926208495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,32,power_law_1.2,0.058880001306533813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,1,power_law_1.01,1.9004928588867187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,32,power_law_1.2,0.05902720093727112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,32,power_law_1.2,0.06168320178985596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,32,power_law_1.2,0.06669440269470214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,1,power_law_1.01,2.176281547546387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,32,power_law_1.2,0.06894720196723939
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,32,power_law_1.2,0.07233279943466187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,1,power_law_1.01,2.1972415924072264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,32,power_law_1.2,0.08297600150108338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,32,power_law_1.2,0.09102079868316651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,32,power_law_1.2,0.1137984037399292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,8,power_law_1.01,0.12125439643859863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,1,power_law_1.01,2.748659133911133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,32,power_law_1.2,0.12511999607086183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,2,power_law_1.2,0.09217280149459839
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,32,power_law_1.2,0.1574463963508606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,8,power_law_1.01,0.20069758892059325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,32,power_law_1.2,0.21144959926605225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,2,power_law_1.2,0.1355712056159973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,8,power_law_1.01,0.08496000170707703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,1,power_law_1.01,2.9247743606567385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,32,power_law_1.2,0.2953536033630371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,2,power_law_1.2,0.126803195476532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,32,power_law_1.2,0.36753919124603274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,8,power_law_1.01,0.11781120300292969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,2,power_law_1.2,0.14758399724960328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,32,power_law_1.2,0.5192639827728271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,2,power_law_1.2,0.21504640579223633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,8,power_law_1.01,0.14754559993743896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,1,power_law_1.01,3.5077182769775392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,32,power_law_1.2,0.632806396484375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,2,power_law_1.2,0.3021120071411133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,8,power_law_1.01,0.19867520332336425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,32,power_law_1.2,0.8043456077575684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,2,power_law_1.2,0.3625663995742798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,8,power_law_1.01,0.20945920944213867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,1,power_law_1.01,4.395935821533203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,2,power_law_1.2,0.4859903812408447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,32,power_law_1.2,1.0905471801757813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,8,power_law_1.01,0.2193471908569336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,2,power_law_1.2,0.5316480159759521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,2,power_law_1.2,0.5160128116607666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,8,power_law_1.01,0.22673280239105226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,32,power_law_1.2,2.8625408172607423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,2,power_law_1.2,0.57259521484375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,1,power_law_1.01,5.97217903137207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,8,power_law_1.01,0.23072640895843505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,2,power_law_1.2,0.5802432060241699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,8,power_law_1.01,0.23635199069976806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,2,power_law_1.2,0.5999360084533691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,8,power_law_1.01,0.240883207321167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,2,power_law_1.2,0.6401023864746094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,1,power_law_1.01,6.806681823730469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,8,power_law_1.01,0.2376960039138794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,2,power_law_1.2,0.6529600143432617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,2,power_law_1.2,0.6866432189941406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,8,power_law_1.01,0.2510463953018188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,2,power_law_1.2,0.6935296058654785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,8,power_law_1.01,0.26117119789123533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,2,power_law_1.2,0.7560256004333497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,8,power_law_1.01,0.26178560256958006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,1,1,power_law_1.01,8.154163360595703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,16,balanced,0.05233600238958994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,2,power_law_1.2,0.8025983810424805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,16,balanced,0.055013333757718406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,16,balanced,0.05397333204746246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,16,balanced,0.062165334820747375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,16,balanced,0.08475200335184734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,16,balanced,0.11385599772135417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,16,balanced,0.11686399579048157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,16,balanced,0.11627733707427979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,16,balanced,0.11853333314259847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,16,balanced,0.11806399623552959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,16,balanced,0.11826133728027344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,16,balanced,0.11964266498883565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,16,balanced,0.12079999844233195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,16,balanced,0.12058132886886597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,8,power_law_1.01,0.2828736066818237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,16,balanced,0.1283199985822042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,16,balanced,0.12854933738708496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,16,balanced,0.13057600458463034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,16,balanced,0.13989333311716715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,16,balanced,0.14641066392262778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,16,balanced,0.15972800056139627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,16,balanced,0.17695466677347818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,2,power_law_1.2,0.9215104103088378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,16,balanced,0.20851733287175497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,16,balanced,0.2328480084737142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,16,balanced,0.30162666241327923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,16,balanced,0.3526293436686198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,16,balanced,0.4840266704559326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,16,balanced,0.591103990872701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,16,balanced,0.7477173010508219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,16,balanced,1.0814133485158284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,8,power_law_1.01,0.3192768096923828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,2,power_law_1.2,1.036678409576416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,16,balanced,2.0866880416870117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,8,power_law_1.01,0.3292351961135864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,2,power_law_1.2,1.2204352378845216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,2,power_law_1.2,1.346463966369629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,8,power_law_1.01,0.3928128004074097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,1,1,power_law_1.01,12.893043518066406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,2,power_law_1.2,1.6466304779052734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,8,power_law_1.01,0.4560512065887451
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,2,power_law_1.2,2.0114559173583983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,8,power_law_1.01,0.5738687992095948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,2,power_law_1.2,2.6557695388793947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,8,power_law_1.01,0.5620800018310547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,8,power_law_1.01,0.7107583999633789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,2,power_law_1.2,3.4672256469726563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,8,power_law_1.01,0.8912320137023926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,2,power_law_1.2,4.059423828125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,1,1,power_law_1.01,23.583570861816405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,8,power_law_1.01,1.247110366821289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,2,power_law_1.2,6.0075328826904295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,8,power_law_1.01,1.4077312469482421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,2,power_law_1.2,11.667884826660156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,1,8,power_law_1.01,1.7549503326416016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,1,8,power_law_1.01,2.7260799407958984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,32,power_law_1.01,0.05740159749984741
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,32,power_law_1.01,0.05434880256652832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,32,power_law_1.01,0.048902401328086854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,32,power_law_1.01,0.05009920001029968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,128,power_law_1.2,0.0689087986946106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,32,power_law_1.01,0.050419199466705325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,1,8,power_law_1.01,5.057363128662109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,128,power_law_1.2,0.06850559711456299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,32,power_law_1.01,0.052832001447677614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,128,power_law_1.2,0.06293119788169861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,32,power_law_1.01,0.052902400493621826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,128,power_law_1.2,0.062009602785110474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,128,power_law_1.2,0.0648639976978302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,32,power_law_1.01,0.052960002422332765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,128,power_law_1.2,0.06365439891815186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,32,power_law_1.01,0.0562175989151001
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,128,power_law_1.2,0.06681600213050842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,32,power_law_1.01,0.05616639852523804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,128,power_law_1.2,0.06783999800682068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,32,power_law_1.01,0.05796480178833008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,128,power_law_1.2,0.06822400093078614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,32,power_law_1.01,0.05973759889602661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,128,power_law_1.2,0.07204480171203613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,32,power_law_1.01,0.06177279949188232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,16,power_law_1.01,0.07654399871826172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,128,power_law_1.2,0.07057279944419861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,32,power_law_1.01,0.06423680186271667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,16,power_law_1.01,0.0796288013458252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,128,power_law_1.2,0.07288960218429566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,16,power_law_1.01,0.05151360034942627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,32,power_law_1.01,0.07527040243148804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,16,power_law_1.01,0.060134398937225345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,128,power_law_1.2,0.0704255998134613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,32,power_law_1.01,0.07808640003204345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,16,power_law_1.01,0.06729599833488464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,16,power_law_1.01,0.07767680287361145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,128,power_law_1.2,0.0716480016708374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,32,power_law_1.01,0.08407679796218873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,16,power_law_1.01,0.0776639997959137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,16,power_law_1.01,0.0778880000114441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,128,power_law_1.2,0.08013439774513245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,32,power_law_1.01,0.09663360118865967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,16,power_law_1.01,0.07961599826812744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,16,power_law_1.01,0.07846400141716003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,128,power_law_1.2,0.0823743999004364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,32,power_law_1.01,0.10462080240249634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,16,power_law_1.01,0.07958400249481201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,16,power_law_1.01,0.0818560004234314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,128,power_law_1.2,0.08563200235366822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,16,power_law_1.01,0.08092799782752991
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,32,power_law_1.01,0.12616959810256959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,16,power_law_1.01,0.0833728015422821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,128,power_law_1.2,0.10039680004119873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,16,power_law_1.01,0.09255040287971497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,32,power_law_1.01,0.14728319644927979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,16,power_law_1.01,0.09809920191764832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,128,power_law_1.2,0.10713599920272827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,16,power_law_1.01,0.10338560342788697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,32,power_law_1.01,0.18948479890823364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,16,power_law_1.01,0.1155776023864746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,128,power_law_1.2,0.1348736047744751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,16,power_law_1.01,0.11486719846725464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,32,power_law_1.01,0.23143041133880615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,16,power_law_1.01,0.13619840145111084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,128,power_law_1.2,0.1536128044128418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,16,power_law_1.01,0.1487488031387329
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,32,power_law_1.01,0.3169663906097412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,128,power_law_1.2,0.1995136022567749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,16,power_law_1.01,0.1800384044647217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,16,power_law_1.01,0.22839040756225587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,32,power_law_1.01,0.40415358543395996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,128,power_law_1.2,0.24469120502471925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,16,power_law_1.01,0.2960576057434082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,16,power_law_1.01,0.356710410118103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,128,power_law_1.2,0.3430975914001465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,32,power_law_1.01,0.5984640121459961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,16,power_law_1.01,0.5158016204833984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,128,power_law_1.2,0.44588160514831543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,16,power_law_1.01,0.6131904125213623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,32,power_law_1.01,0.7753024101257324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,128,power_law_1.2,0.6407040119171142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,16,power_law_1.01,0.78919677734375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,32,power_law_1.01,0.9063488006591797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,16,power_law_1.01,1.0989952087402344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,128,power_law_1.2,0.8309632301330566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,32,power_law_1.01,1.3555007934570313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,128,power_law_1.2,1.045740795135498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,16,power_law_1.01,2.4123519897460937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,128,power_law_1.2,1.7067968368530273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,32,power_law_1.01,2.856012725830078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,128,power_law_1.2,3.607334518432617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,4,balanced,0.05659199754397074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,4,balanced,0.058287998040517174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,4,balanced,0.0599839985370636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,4,balanced,0.06645866731802623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,4,balanced,0.07653866708278656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,4,balanced,0.10133333007494609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,4,balanced,0.10147200028101604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,4,balanced,0.10571733117103577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,4,balanced,0.10796800255775452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,4,balanced,0.10739200313886006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,4,balanced,0.10748266180356343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,4,balanced,0.1111893355846405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,4,balanced,0.11314666271209717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,4,balanced,0.11559999982515971
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,4,balanced,0.12315733234087627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,4,balanced,0.1251573363939921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,4,balanced,0.1276479959487915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,4,balanced,0.14246933658917746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,4,balanced,0.15076800187428793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,4,balanced,0.1758400003115336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,4,balanced,0.19926933447519937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,4,balanced,0.24740266799926758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,4,balanced,0.29785066843032837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,4,balanced,0.42903467019399005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,4,balanced,0.5347573359807333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,4,balanced,0.7744906743367513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,4,balanced,1.0173973242441814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,4,4,power_law_1.01,0.05146239995956421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,4,balanced,1.2575146357218425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,4,4,power_law_1.01,0.04973439872264862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,4,balanced,1.9176054000854492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,4,4,power_law_1.01,0.050195199251174924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,4,4,power_law_1.01,0.05486720204353333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,4,balanced,3.7401707967122397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,4,4,power_law_1.01,0.05559679865837097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,4,4,power_law_1.01,0.05829120278358459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,4,4,power_law_1.01,0.060499197244644164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,4,4,power_law_1.01,0.05983359813690185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,4,4,power_law_1.01,0.06310399770736694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,4,4,power_law_1.01,0.06421120166778564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,2,4,power_law_1.2,0.06106879711151123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,4,4,power_law_1.01,0.06512640118598938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,2,4,power_law_1.2,0.0794048011302948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,2,4,power_law_1.2,0.07371519804000855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,4,4,power_law_1.01,0.06816639900207519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,2,4,power_law_1.2,0.07251200079917908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,4,4,power_law_1.01,0.07009919881820678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,2,4,power_law_1.2,0.07706239819526672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,2,4,power_law_1.2,0.07530879974365234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,4,4,power_law_1.01,0.07329279780387879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,2,4,power_law_1.2,0.08586239814758301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,4,4,power_law_1.01,0.0787775993347168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,2,4,power_law_1.2,0.08232960104942322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,2,4,power_law_1.2,0.09000959992408752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,4,4,power_law_1.01,0.08398079872131348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,2,4,power_law_1.2,0.09428480267524719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,4,4,power_law_1.01,0.09682559967041016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,2,4,power_law_1.2,0.08655999898910523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,2,4,power_law_1.2,0.08988800048828124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,4,4,power_law_1.01,0.10880639553070068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,2,4,power_law_1.2,0.09966080188751221
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,4,4,power_law_1.01,0.13544319868087767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,2,4,power_law_1.2,0.10288000106811523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,2,4,power_law_1.2,0.11256959438323974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,4,4,power_law_1.01,0.1602560043334961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,2,4,power_law_1.2,0.1248960018157959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,2,4,power_law_1.2,0.13274879455566407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,4,4,power_law_1.01,0.20774400234222412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,2,4,power_law_1.2,0.16465280055999756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,4,4,power_law_1.01,0.24809598922729492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,2,4,power_law_1.2,0.19073280096054077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,2,4,power_law_1.2,0.27305600643157957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,4,4,power_law_1.01,0.32312960624694825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,2,4,power_law_1.2,0.38355839252471924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,4,4,power_law_1.01,0.5152768135070801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,2,4,power_law_1.2,0.5067455768585205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,2,4,power_law_1.2,0.5846079826354981
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,4,4,power_law_1.01,0.60033278465271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,2,4,power_law_1.2,0.8563008308410645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,4,4,power_law_1.01,0.8470848083496094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,2,4,power_law_1.2,1.270867156982422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,4,4,power_law_1.01,1.0736127853393556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,2,4,power_law_1.2,1.7574016571044921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,4,4,power_law_1.01,1.619001579284668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,2,4,power_law_1.2,2.036729621887207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,4,4,power_law_1.01,2.4043455123901367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,2,4,power_law_1.2,2.759775924682617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,2,4,power_law_1.2,3.4481407165527345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,4,4,power_law_1.01,4.899987030029297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,8,power_law_1.2,0.0723583996295929
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,8,power_law_1.2,0.09289600253105164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,2,4,power_law_1.2,9.468844604492187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,8,power_law_1.2,0.056006401777267456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,8,power_law_1.2,0.06431999802589417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,8,power_law_1.2,0.08403840065002441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,8,power_law_1.2,0.0916159987449646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,8,power_law_1.2,0.11796480417251587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,8,power_law_1.2,0.11999360322952271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,8,power_law_1.2,0.12668800354003906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,8,power_law_1.2,0.125600004196167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,8,power_law_1.2,0.13638399839401244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,8,power_law_1.2,0.13598079681396485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,8,power_law_1.2,0.14309120178222656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,8,power_law_1.2,0.15420160293579102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,4,power_law_1.2,0.048198398947715757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,8,power_law_1.2,0.16297600269317628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,2,power_law_1.01,0.09296640157699584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,8,power_law_1.2,0.17151999473571777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,4,power_law_1.2,0.058771198987960814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,8,power_law_1.2,0.165664005279541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,2,power_law_1.01,0.1899072051048279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,8,power_law_1.2,0.1926144003868103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,2,power_law_1.01,0.13176319599151612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,4,power_law_1.2,0.0520576000213623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,8,power_law_1.2,0.18922239542007446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,2,power_law_1.01,0.17997440099716186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,4,power_law_1.2,0.0564736008644104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,8,power_law_1.2,0.22467200756072997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,2,power_law_1.01,0.2964287996292114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,8,power_law_1.2,0.2668544054031372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,2,power_law_1.01,0.41339521408081054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,4,power_law_1.2,0.06840959787368775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,8,power_law_1.2,0.3208832025527954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,2,power_law_1.01,0.5946752071380615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,4,power_law_1.2,0.07637760043144226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,8,power_law_1.2,0.36895360946655276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,2,power_law_1.01,0.5878592014312745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,2,power_law_1.01,0.6279232025146484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,8,power_law_1.2,0.48137598037719725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,4,power_law_1.2,0.09445760250091553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,2,power_law_1.01,0.6535488128662109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,8,power_law_1.2,0.5493760108947754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,4,power_law_1.2,0.0997439980506897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,2,power_law_1.01,0.6832704067230224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,8,power_law_1.2,0.8038016319274902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,2,power_law_1.01,0.7031744003295899
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,4,power_law_1.2,0.1012992024421692
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,2,power_law_1.01,0.7002048015594482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,8,power_law_1.2,0.9980480194091796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,4,power_law_1.2,0.10440959930419921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,2,power_law_1.01,0.7109824180603027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,4,power_law_1.2,0.1087615966796875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,2,power_law_1.01,0.7415552139282227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,8,power_law_1.2,1.3425087928771973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,2,power_law_1.01,0.7778560161590576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,4,power_law_1.2,0.11437439918518066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,2,power_law_1.01,0.809331226348877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,8,power_law_1.2,2.2497215270996094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,4,power_law_1.2,0.11849600076675415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,2,power_law_1.01,0.8469568252563476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,4,power_law_1.2,0.12961280345916748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,2,power_law_1.01,0.9008959770202637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,4,power_law_1.2,0.13876479864120483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,2,power_law_1.01,0.9818047523498535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,8,power_law_1.2,4.104870223999024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,2,power_law_1.01,1.0076352119445802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,4,power_law_1.2,0.14471039772033692
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,2,power_law_1.01,1.1942975997924805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,4,power_law_1.2,0.1542207956314087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,2,power_law_1.01,1.3003583908081056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,4,power_law_1.2,0.17003519535064698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,2,power_law_1.01,1.6015296936035157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,4,power_law_1.2,0.1885823965072632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,4,power_law_1.2,0.22792320251464843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,2,power_law_1.01,1.8081855773925781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,4,power_law_1.2,0.2813760042190552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,2,power_law_1.01,2.3791936874389648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,4,power_law_1.2,0.35508480072021487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,2,power_law_1.01,2.929350471496582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,4,power_law_1.2,0.44377598762512205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,4,power_law_1.2,0.6103616237640381
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,2,power_law_1.01,3.56297607421875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,8,power_law_1.01,0.08704000115394592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,4,power_law_1.2,0.7931903839111328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,8,power_law_1.01,0.12113280296325683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,8,power_law_1.01,0.07621759772300721
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,2,power_law_1.01,5.271500778198242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,4,power_law_1.2,1.1060416221618652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,8,power_law_1.01,0.07994880080223084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,8,power_law_1.01,0.09813759922981262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,4,power_law_1.2,1.3734272003173829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,8,power_law_1.01,0.11674879789352417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,8,power_law_1.01,0.14790400266647338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,8,power_law_1.01,0.15221760272979737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,4,power_law_1.2,1.8223232269287108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,2,power_law_1.01,9.77738265991211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,8,power_law_1.01,0.1509503960609436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,8,power_law_1.01,0.15842560529708863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,4,power_law_1.2,2.620684814453125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,8,power_law_1.01,0.1627071976661682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,8,power_law_1.01,0.1678272008895874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,8,power_law_1.01,0.1602687954902649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,8,power_law_1.01,0.17550719976425172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,8,power_law_1.01,0.17701120376586915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,4,power_law_1.2,5.55689582824707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,8,power_law_1.01,0.17826559543609619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,8,power_law_1.01,0.1941375970840454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,8,power_law_1.01,0.20358400344848632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,8,power_law_1.01,0.21557118892669677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,8,balanced,0.04380266865094503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,8,balanced,0.04571733375390371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,8,power_law_1.01,0.25459840297698977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,8,balanced,0.046223998069763184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,8,balanced,0.05197866757710775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,8,balanced,0.07041066884994507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,8,balanced,0.10130133231480916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,8,balanced,0.13127467036247253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,8,balanced,0.12890666723251343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,8,balanced,0.12877333164215088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,8,balanced,0.1320853332678477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,8,balanced,0.12972266475359598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,8,balanced,0.13185066978136697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,8,balanced,0.13537599643071493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,8,balanced,0.13609600067138672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,8,balanced,0.14062933127085367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,8,balanced,0.1421226660410563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,8,balanced,0.1529706617196401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,8,balanced,0.15809067090352377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,8,balanced,0.1696959932645162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,8,power_law_1.01,0.29377920627593995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,8,balanced,0.18766933679580688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,8,balanced,0.2053920030593872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,8,balanced,0.2449386715888977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,8,balanced,0.2882026632626851
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,8,balanced,0.3630026578903198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,8,balanced,0.4456426699956258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,8,balanced,0.5980853239695231
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,8,balanced,0.7459200223286947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,8,power_law_1.01,0.345849609375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,8,balanced,0.9250400066375732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,8,balanced,1.3648692766825359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,8,balanced,2.6302879651387534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,8,power_law_1.01,0.4246784210205078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,8,power_law_1.01,0.5323200225830078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,8,power_law_1.01,0.6617216110229492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,8,power_law_1.01,0.8722559928894043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,128,power_law_1.01,0.03814400136470795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,8,power_law_1.01,1.2072896003723144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,128,power_law_1.01,0.03648000061511993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,128,power_law_1.01,0.036748799681663516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,8,power_law_1.01,1.3317119598388671
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,128,power_law_1.01,0.03672960102558136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,128,power_law_1.01,0.03706879913806915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,8,power_law_1.01,2.1185535430908202
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,128,power_law_1.01,0.037529599666595456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,128,power_law_1.01,0.03814400136470795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,8,power_law_1.01,0.03858560025691986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,8,power_law_1.01,0.035366401076316833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,128,power_law_1.01,0.037376001477241516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,8,power_law_1.01,0.03478400111198425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,128,power_law_1.01,0.03805440068244934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,8,power_law_1.01,0.036883199214935304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,8,power_law_1.01,0.03937920033931732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,8,power_law_1.01,4.453004837036133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,128,power_law_1.01,0.037836799025535585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,8,power_law_1.01,0.0414463996887207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,128,power_law_1.01,0.037887999415397645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,8,power_law_1.01,0.04255360066890716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,8,power_law_1.01,0.042777600884437564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,128,power_law_1.01,0.039263999462127684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,8,power_law_1.01,0.0434112012386322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,128,power_law_1.01,0.03873279988765717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,8,power_law_1.01,0.043296000361442565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,8,power_law_1.01,0.0444927990436554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,128,power_law_1.01,0.03986560106277466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,8,power_law_1.01,0.04542720019817352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,128,power_law_1.01,0.0416128009557724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,8,power_law_1.01,0.047302401065826415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,8,power_law_1.01,0.04915199875831604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,128,power_law_1.01,0.04243200123310089
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,8,power_law_1.01,0.05497599840164184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,128,power_law_1.01,0.043833601474761966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,8,power_law_1.01,0.05905280113220215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,128,power_law_1.01,0.044012799859046936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,8,power_law_1.01,0.06272000074386597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,8,power_law_1.01,0.07269759774208069
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,128,power_law_1.01,0.045228800177574156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,8,power_law_1.01,0.0782912015914917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,128,power_law_1.01,0.04734080135822296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,8,power_law_1.01,0.09596160054206848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,8,power_law_1.01,0.10867199897766114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,128,power_law_1.01,0.05059199929237366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,8,power_law_1.01,0.13512320518493653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,128,power_law_1.01,0.060627198219299315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,8,power_law_1.01,0.1793023943901062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,128,power_law_1.01,0.06836479902267456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,8,power_law_1.01,0.23530240058898927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,32,balanced,0.0330079992612203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,32,balanced,0.030581332743167877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,32,balanced,0.03166933357715607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,32,balanced,0.031680000325044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,32,balanced,0.03359466542800268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,32,balanced,0.03393599887688955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,128,power_law_1.01,0.08763520121574402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,32,balanced,0.03569599986076355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,8,power_law_1.01,0.28086400032043457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,32,balanced,0.03397866586844126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,32,balanced,0.034048000971476235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,32,balanced,0.03578133384386698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,32,balanced,0.03565866748491923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,32,balanced,0.03585600107908249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,32,balanced,0.03625066578388214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,32,balanced,0.03748266647259394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,32,balanced,0.0418453315893809
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,32,balanced,0.04005866746107737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,32,balanced,0.041834667325019836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,32,balanced,0.04394133388996124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,32,balanced,0.046037331223487854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,32,balanced,0.0499946673711141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,32,balanced,0.05607999861240387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,32,balanced,0.0762613316377004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,32,balanced,0.09518399834632874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,128,power_law_1.01,0.11638400554656983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,32,balanced,0.12410133083661397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,8,power_law_1.01,0.4349055767059326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,32,balanced,0.1574079990386963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,32,balanced,0.21176000436147055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,32,balanced,0.2641119956970215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,32,balanced,0.3163413405418396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,32,balanced,0.469157338142395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,32,balanced,0.8812959988911947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,128,power_law_1.01,0.1414080023765564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,8,power_law_1.01,0.591206407546997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,128,power_law_1.01,0.18757760524749756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,8,power_law_1.01,0.7403071880340576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,2,128,power_law_1.01,0.2407167911529541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,8,power_law_1.01,0.9918720245361328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,2,128,power_law_1.01,0.39779839515686033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,8,power_law_1.01,2.102681541442871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,2,128,power_law_1.01,0.7590015888214111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,16,balanced,0.05420266588528951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,16,balanced,0.05824000140031179
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,16,balanced,0.05994133154551188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,16,balanced,0.07250133156776428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,16,balanced,0.0913866659005483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,16,balanced,0.0956213374932607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,16,balanced,0.09691199660301208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,16,balanced,0.09745599826176961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,16,balanced,0.0981173316637675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,16,balanced,0.09930133819580078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,16,balanced,0.09968533118565877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,16,balanced,0.10083733002344768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,16,balanced,0.10285866260528564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,16,balanced,0.10604799787203471
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,16,balanced,0.11101333300272624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,16,balanced,0.1135093371073405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,16,balanced,0.11886399984359741
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,16,balanced,0.13147200147310892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,16,balanced,0.1425333321094513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,16,balanced,0.18732800086339316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,16,balanced,0.19138665994008383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,16,balanced,0.25831466913223267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,16,balanced,0.26587732632954914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,16,balanced,0.3545920054117839
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,16,balanced,0.4139680067698161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,16,balanced,0.5821813344955444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,16,balanced,0.70906662940979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,2,16,balanced,0.8693013191223145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,2,16,balanced,1.3098666667938232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,2,16,balanced,2.67250124613444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,32,1,power_law_1.01,0.03496319949626923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,32,1,power_law_1.01,0.035257598757743834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,32,1,power_law_1.01,0.03644160032272339
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,32,1,power_law_1.01,0.03930880129337311
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,32,1,power_law_1.01,0.039084801077842714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,32,1,power_law_1.01,0.040582400560379026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,32,1,power_law_1.01,0.04193280041217804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,32,1,power_law_1.01,0.04224640130996704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,32,1,power_law_1.01,0.04185599982738495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,32,1,power_law_1.01,0.04296959936618805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,32,1,power_law_1.01,0.044147199392318724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,32,1,power_law_1.01,0.045900800824165346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,32,1,power_law_1.01,0.04850560128688812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,32,1,power_law_1.01,0.04826880097389221
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,32,1,power_law_1.01,0.05494400262832642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,32,1,power_law_1.01,0.05676800012588501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,32,power_law_1.2,0.07671679854393006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,32,1,power_law_1.01,0.06254720091819763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,32,1,power_law_1.01,0.07397760152816772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,32,1,power_law_1.01,0.09000319838523865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,32,1,power_law_1.01,0.10805759429931641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,32,1,power_law_1.01,0.11953920125961304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,32,1,power_law_1.01,0.16051199436187744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,32,1,power_law_1.01,0.19258879423141478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,32,1,power_law_1.01,0.27080960273742677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,32,1,power_law_1.01,0.34650239944458006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,32,1,power_law_1.01,0.49395198822021485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,32,power_law_1.2,0.08762879967689514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,32,1,power_law_1.01,0.6400447845458984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,32,1,power_law_1.01,0.7864448070526123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,32,1,power_law_1.01,1.2227328300476075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,32,power_law_1.2,0.08496000170707703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,32,1,power_law_1.01,2.3913984298706055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,16,power_law_1.01,0.05319679975509643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,32,power_law_1.2,0.06585599780082703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,16,power_law_1.01,0.07331200242042542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,16,power_law_1.01,0.07063040137290955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,32,power_law_1.2,0.06707199811935424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,16,power_law_1.01,0.04824959933757782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,32,power_law_1.2,0.07193599939346314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,16,power_law_1.01,0.051660799980163576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,16,power_law_1.01,0.05663359761238098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,32,power_law_1.2,0.0748799979686737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,16,power_law_1.01,0.05923200249671936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,32,power_law_1.2,0.07724159955978394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,16,power_law_1.01,0.06516479849815368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,64,power_law_1.01,0.0834176003932953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,16,power_law_1.01,0.06588799953460693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,32,power_law_1.2,0.07799680233001709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,64,power_law_1.01,0.09007999897003174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,16,power_law_1.01,0.06679040193557739
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,64,power_law_1.01,0.08054400086402894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,32,power_law_1.2,0.08103039860725403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,16,power_law_1.01,0.07077119946479797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,64,power_law_1.01,0.07690879702568054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,16,power_law_1.01,0.07143679857254029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,32,power_law_1.2,0.08218880295753479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,64,power_law_1.01,0.07451519966125489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,16,power_law_1.01,0.07423359751701356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,64,power_law_1.01,0.0868607997894287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,32,power_law_1.2,0.0864960014820099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,64,power_law_1.01,0.08716800212860107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,16,power_law_1.01,0.07793279886245727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,64,power_law_1.01,0.08893439769744874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,16,power_law_1.01,0.08710399866104127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,32,power_law_1.2,0.08309760093688964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,64,power_law_1.01,0.09080320000648498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,64,power_law_1.01,0.08906239867210389
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,16,power_law_1.01,0.08961920142173767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,32,power_law_1.2,0.08691840171813965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,64,power_law_1.01,0.09287679791450501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,16,power_law_1.01,0.1016319990158081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,64,power_law_1.01,0.09318400025367737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,32,power_law_1.2,0.09224320054054261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,64,power_law_1.01,0.09631360173225403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,16,power_law_1.01,0.11424000263214111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,64,power_law_1.01,0.09134719967842102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,16,power_law_1.01,0.11992319822311401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,64,power_law_1.01,0.09976959824562073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,32,power_law_1.2,0.09451519846916198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,16,power_law_1.01,0.14787839651107787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,64,power_law_1.01,0.10071680545806885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,32,power_law_1.2,0.1004032015800476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,64,power_law_1.01,0.10574079751968384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,16,power_law_1.01,0.17279360294342042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,64,power_law_1.01,0.11886080503463745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,32,power_law_1.2,0.1133504033088684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,16,power_law_1.01,0.22906239032745362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,64,power_law_1.01,0.13102079629898072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,64,power_law_1.01,0.16102399826049804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,16,power_law_1.01,0.29602560997009275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,32,power_law_1.2,0.12711679935455322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,64,power_law_1.01,0.18643200397491455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,16,power_law_1.01,0.3777472019195557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,64,power_law_1.01,0.24343039989471435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,32,power_law_1.2,0.15317120552062988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,64,power_law_1.01,0.27946879863739016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,16,power_law_1.01,0.4737919807434082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,64,power_law_1.01,0.38356480598449705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,32,power_law_1.2,0.1838912010192871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,16,power_law_1.01,0.7123199939727783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,64,power_law_1.01,0.4887296199798584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,32,power_law_1.2,0.23124480247497559
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,64,power_law_1.01,0.7253568172454834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,16,power_law_1.01,0.8777664184570313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,32,power_law_1.2,0.29743359088897703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,64,power_law_1.01,0.9520575523376464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,16,power_law_1.01,1.1044159889221192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,64,power_law_1.01,1.1449983596801758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,32,power_law_1.2,0.39441280364990233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,16,power_law_1.01,1.6854080200195312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,64,power_law_1.01,1.7811008453369142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,32,power_law_1.2,0.5109504222869873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,32,power_law_1.2,0.7329919815063477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,16,power_law_1.01,3.631302261352539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,64,power_law_1.01,3.7513919830322267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,32,power_law_1.2,0.983737564086914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,32,power_law_1.2,1.2021823883056642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,32,power_law_1.2,1.882271957397461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,32,power_law_1.2,3.822111892700195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,128,balanced,0.04160533348719279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,128,balanced,0.04179200033346812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,128,balanced,0.04243200023969015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,128,balanced,0.04146133363246918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,8,power_law_1.2,0.034169599413871765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,128,balanced,0.041562666495641075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,128,balanced,0.041637333730856575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,128,balanced,0.043280000487963356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,128,balanced,0.04363733530044556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,128,balanced,0.041482667128245033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,128,balanced,0.04351999859015147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,128,balanced,0.04387733340263367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,128,balanced,0.04558933277924856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,128,balanced,0.044293334086736046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,128,balanced,0.04376000165939331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,128,balanced,0.04868799944718679
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,8,power_law_1.2,0.028281599283218384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,128,balanced,0.04804266492525736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,128,balanced,0.04933333396911621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,128,balanced,0.05037866532802582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,128,balanced,0.05409066875775655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,128,balanced,0.05780800183614095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,128,balanced,0.06032533446947733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,128,balanced,0.07050666709740956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,128,balanced,0.0786186655362447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,128,balanced,0.10231467088063557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,8,power_law_1.2,0.028755199909210206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,128,balanced,0.11788800358772278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,128,balanced,0.15130666891733804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,128,balanced,0.18894400199254355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,128,balanced,0.2242506742477417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,128,balanced,0.3285760084788005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,8,power_law_1.2,0.028575998544692994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,128,balanced,0.60753067334493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,8,power_law_1.2,0.029542401432991028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,8,power_law_1.2,0.03239679932594299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,8,power_law_1.2,0.033107200264930726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,8,power_law_1.2,0.03511680066585541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,8,power_law_1.2,0.03351039886474609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,8,power_law_1.2,0.03497599959373474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,8,power_law_1.2,0.034745600819587705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,8,power_law_1.2,0.03591040074825287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,8,power_law_1.2,0.0367935985326767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,8,power_law_1.2,0.03804160058498383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,8,power_law_1.2,0.04255360066890716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,8,power_law_1.2,0.044582399725914004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,8,power_law_1.2,0.04688000082969666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,8,power_law_1.2,0.051923197507858274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,8,power_law_1.2,0.059596800804138185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,8,power_law_1.2,0.06888319849967957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,8,power_law_1.2,0.07788159847259521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,4,power_law_1.2,0.06960639953613282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,8,power_law_1.2,0.10637439489364624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,8,power_law_1.2,0.13166719675064087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,1,balanced,0.052442664901415505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,8,power_law_1.2,0.18272000551223755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,1,balanced,0.06081066528956095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,1,balanced,0.08213333288828532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,1,balanced,0.11341333389282227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,1,balanced,0.17292267084121704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,1,balanced,0.20171199242273966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,1,balanced,0.20784533023834229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,1,balanced,0.20962133010228476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,4,power_law_1.2,0.10012160539627075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,1,balanced,0.21157866716384888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,1,balanced,0.21093867222468057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,8,power_law_1.2,0.23708159923553468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,1,balanced,0.21649599075317383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,1,balanced,0.22080532709757486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,1,balanced,0.22630399465560913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,1,balanced,0.23441066344579062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,1,balanced,0.24053333202997842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,1,balanced,0.2535039981206258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,1,balanced,0.2637439966201782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,1,balanced,0.2972693244616191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,1,balanced,0.32103999455769855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,1,balanced,0.4289066791534424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,1,balanced,0.45028265317281085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,8,power_law_1.2,0.33182721138000487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,1,balanced,0.6598453521728516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,1,balanced,0.730522632598877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,1,balanced,1.1438826719919841
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,4,power_law_1.2,0.0675711989402771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,1,balanced,1.3053226470947266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,8,power_law_1.2,0.4700160026550293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,1,balanced,2.0106773376464844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,1,balanced,2.468143939971924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,1,balanced,2.8745654424031577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,4,power_law_1.2,0.0885312020778656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,8,power_law_1.2,0.5294976234436035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,1,balanced,4.583050727844238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,8,power_law_1.2,1.136019229888916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,4,power_law_1.2,0.10022399425506592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,1,balanced,8.810469309488932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,4,power_law_1.2,0.11834239959716797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,8,power_law_1.2,2.171776008605957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,4,power_law_1.2,0.14562560319900514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,4,power_law_1.2,0.15453439950942993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,4,power_law_1.2,0.15222400426864624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,4,power_law_1.2,0.16550400257110595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,4,power_law_1.2,0.16299519538879395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,4,power_law_1.2,0.17062400579452514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,4,power_law_1.2,0.17776639461517335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,4,power_law_1.2,0.1838719964027405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,4,4,power_law_1.2,0.043859198689460754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,4,power_law_1.2,0.19903359413146973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,4,4,power_law_1.2,0.062483197450637816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,4,4,power_law_1.2,0.05060480237007141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,4,4,power_law_1.2,0.053401601314544675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,4,4,power_law_1.2,0.05445759892463684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,4,4,power_law_1.2,0.05672320127487183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,4,4,power_law_1.2,0.058847999572753905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,4,4,power_law_1.2,0.059084802865982056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,4,power_law_1.2,0.20136959552764894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,4,4,power_law_1.2,0.060140800476074216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,4,4,power_law_1.2,0.06336640119552613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,4,4,power_law_1.2,0.06439679861068726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,4,4,power_law_1.2,0.06681600213050842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,4,4,power_law_1.2,0.06870399713516236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,16,power_law_1.01,0.05868160128593445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,4,4,power_law_1.2,0.0713536024093628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,4,4,power_law_1.2,0.07882239818572997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,4,power_law_1.2,0.21241600513458253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,4,4,power_law_1.2,0.08940799832344055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,4,4,power_law_1.2,0.09356799721717834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,4,4,power_law_1.2,0.11066240072250366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,16,power_law_1.01,0.1105728030204773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,4,4,power_law_1.2,0.13334399461746216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,4,4,power_law_1.2,0.1866368055343628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,4,4,power_law_1.2,0.22527999877929689
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,4,power_law_1.2,0.24090878963470458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,4,4,power_law_1.2,0.28782079219818113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,16,power_law_1.01,0.05649920105934143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,4,4,power_law_1.2,0.40264320373535156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,4,4,power_law_1.2,0.5425983905792237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,16,power_law_1.01,0.05767040252685547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,4,4,power_law_1.2,0.6103871822357178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,4,power_law_1.2,0.258022403717041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,4,4,power_law_1.2,0.9500800132751465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,16,power_law_1.01,0.06227840185165405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,4,4,power_law_1.2,1.340659236907959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,4,power_law_1.2,0.3335103988647461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,4,4,power_law_1.2,1.4256192207336427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,16,power_law_1.01,0.06885759830474854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,4,4,power_law_1.2,2.4998464584350586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,16,power_law_1.01,0.077183997631073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,4,power_law_1.2,0.3460671901702881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,16,power_law_1.01,0.07923840284347534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,4,power_law_1.2,0.4353792190551758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,4,4,power_law_1.2,5.404415893554687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,16,power_law_1.01,0.08154240250587463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,16,power_law_1.01,0.08170239925384522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,4,power_law_1.2,0.5045119762420655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,16,power_law_1.01,0.08340479731559754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,16,power_law_1.01,0.08524799942970276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,4,power_law_1.2,0.7379007816314698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,16,power_law_1.01,0.08622080087661743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,4,power_law_1.2,0.8258367538452148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,16,power_law_1.01,0.08854399919509888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,16,power_law_1.01,0.09503359794616699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,4,power_law_1.2,1.2204416275024415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,16,power_law_1.01,0.09619200229644775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,16,power_law_1.01,0.10257279872894287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,4,power_law_1.2,1.5423744201660157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,16,power_law_1.01,0.11715199947357177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,16,power_law_1.01,0.12300800085067749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,16,power_law_1.01,0.14512640237808228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,4,power_law_1.2,2.009440040588379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,16,power_law_1.01,0.17104640007019042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,8,power_law_1.01,0.056729602813720706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,16,power_law_1.01,0.21468799114227294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,4,power_law_1.2,3.2078014373779298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,16,power_law_1.01,0.2518336057662964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,8,power_law_1.01,0.06216959953308106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,16,power_law_1.01,0.3474560022354126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,8,power_law_1.01,0.0720192015171051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,16,power_law_1.01,0.40517759323120117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,8,power_law_1.01,0.05724160075187683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,16,power_law_1.01,0.5766784191131592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,8,power_law_1.01,0.06305919885635376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,4,power_law_1.2,6.125459289550781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,16,power_law_1.01,0.7544960021972656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,8,power_law_1.01,0.06892799735069274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,8,power_law_1.01,0.07093120217323304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,16,power_law_1.01,0.9310144424438477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,8,power_law_1.01,0.07957119941711426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,16,power_law_1.01,1.3660032272338867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,8,power_law_1.01,0.0804032027721405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,8,power_law_1.01,0.08123520016670227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,8,power_law_1.01,0.0828224003314972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,16,power_law_1.01,2.9876991271972657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,8,power_law_1.01,0.08657280206680298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,8,power_law_1.01,0.08757759928703308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,1,2,power_law_1.2,0.08064640164375306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,8,power_law_1.01,0.08956159949302674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,1,2,power_law_1.2,0.10087039470672607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,1,2,power_law_1.2,0.10060800313949585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,1,2,power_law_1.2,0.14010239839553834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,8,power_law_1.01,0.0989247977733612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,1,2,power_law_1.2,0.13936640024185182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,1,2,power_law_1.2,0.1413632035255432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,1,2,power_law_1.2,0.1503167986869812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,8,power_law_1.01,0.0998528003692627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,1,2,power_law_1.2,0.14880000352859496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,1,2,power_law_1.2,0.15234559774398804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,1,2,power_law_1.2,0.15466879606246947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,1,2,power_law_1.2,0.15242880582809448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,8,power_law_1.01,0.1085312008857727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,1,2,power_law_1.2,0.16254080533981324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,1,2,power_law_1.2,0.16407040357589722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,1,2,power_law_1.2,0.17994240522384644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,1,2,power_law_1.2,0.2020479917526245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,8,power_law_1.01,0.12643840312957763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,1,2,power_law_1.2,0.21003520488739014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,1,2,power_law_1.2,0.23900799751281737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,1,2,power_law_1.2,0.2655168056488037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,1,2,power_law_1.2,0.3144256114959717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,8,power_law_1.01,0.14209280014038086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,1,2,power_law_1.2,0.3823296070098877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,1,2,power_law_1.2,0.4551743984222412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,8,1,power_law_1.01,0.03587839901447296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,8,1,power_law_1.01,0.03787519931793213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,1,2,power_law_1.2,0.6467455863952637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,8,power_law_1.01,0.17662719488143921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,8,1,power_law_1.01,0.038073599338531494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,8,1,power_law_1.01,0.043263998627662656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,1,2,power_law_1.2,0.9317184448242187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,8,1,power_law_1.01,0.04429439902305603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,8,1,power_law_1.01,0.0440447986125946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,8,1,power_law_1.01,0.04537599980831146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,1,2,power_law_1.2,1.2556672096252441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,8,1,power_law_1.01,0.046777600049972536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,8,1,power_law_1.01,0.047916799783706665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,8,power_law_1.01,0.18736000061035157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,8,1,power_law_1.01,0.05002239942550659
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,1,2,power_law_1.2,1.4775872230529785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,8,1,power_law_1.01,0.05087360143661499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,8,1,power_law_1.01,0.05371519923210144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,8,1,power_law_1.01,0.05674239993095398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,8,1,power_law_1.01,0.06303359866142273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,8,1,power_law_1.01,0.06692479848861695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,1,2,power_law_1.2,2.3602624893188477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,8,1,power_law_1.01,0.07120000123977661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,8,power_law_1.01,0.23004798889160155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,8,1,power_law_1.01,0.07297279834747314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,8,1,power_law_1.01,0.08517760038375854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,8,1,power_law_1.01,0.09861760139465332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,8,1,power_law_1.01,0.12017279863357544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,1,2,power_law_1.2,3.1524288177490236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,8,1,power_law_1.01,0.13877760171890258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,8,1,power_law_1.01,0.17934720516204833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,8,power_law_1.01,0.298144006729126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,8,1,power_law_1.01,0.21660799980163575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,8,1,power_law_1.01,0.30210559368133544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,8,1,power_law_1.01,0.3791039943695068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,1,2,power_law_1.2,3.861113739013672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,8,1,power_law_1.01,0.5406527996063233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,8,1,power_law_1.01,0.6975872039794921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,8,power_law_1.01,0.40570878982543945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,8,1,power_law_1.01,0.8567872047424316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,1,2,power_law_1.2,6.106335830688477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,8,1,power_law_1.01,1.3223999977111816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,8,power_law_1.01,0.5214015960693359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,8,1,power_law_1.01,2.612646484375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,8,power_law_1.01,0.7222976207733154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,1,2,power_law_1.2,12.605241394042968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,8,power_law_1.01,0.9096128463745117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,8,power_law_1.01,1.1322431564331055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,8,power_law_1.01,1.7833343505859376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,8,power_law_1.01,3.437625503540039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,32,power_law_1.2,0.03224320113658905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,32,power_law_1.2,0.028908801078796387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,32,power_law_1.2,0.02869119942188263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,32,power_law_1.2,0.028863999247550964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,32,2,power_law_1.2,0.03282560110092163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,32,power_law_1.2,0.030060800909996032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,32,2,power_law_1.2,0.03361279964447021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,32,power_law_1.2,0.030969598889350893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,32,2,power_law_1.2,0.03278720080852508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,32,2,power_law_1.2,0.03219839930534363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,32,power_law_1.2,0.031046399474143983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,32,2,power_law_1.2,0.03315840065479279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,32,power_law_1.2,0.032467201352119446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,32,2,power_law_1.2,0.033318400382995605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,32,2,power_law_1.2,0.03381119966506958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,32,power_law_1.2,0.032339200377464294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,32,2,power_law_1.2,0.034841600060462954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,32,2,power_law_1.2,0.03491199910640717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,32,power_law_1.2,0.034143999218940735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,32,2,power_law_1.2,0.03462400138378143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,32,power_law_1.2,0.03558399975299835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,32,2,power_law_1.2,0.035308799147605895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,32,2,power_law_1.2,0.0367680013179779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,32,power_law_1.2,0.03559040129184723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,32,2,power_law_1.2,0.03779839873313904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,32,2,power_law_1.2,0.039366400241851805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,32,power_law_1.2,0.03637759983539581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,32,2,power_law_1.2,0.0470335990190506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,32,power_law_1.2,0.036723199486732486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,32,2,power_law_1.2,0.04802559912204742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,32,2,power_law_1.2,0.05021439790725708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,32,power_law_1.2,0.040633600950241086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,32,2,power_law_1.2,0.05605120062828064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,32,2,power_law_1.2,0.06622719764709473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,32,power_law_1.2,0.04238080084323883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,32,2,power_law_1.2,0.08057600259780884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,32,2,power_law_1.2,0.09840000271797181
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,32,power_law_1.2,0.043935999274253845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,32,2,power_law_1.2,0.13734400272369385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,32,power_law_1.2,0.048537600040435794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,32,2,power_law_1.2,0.16931840181350707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,32,2,power_law_1.2,0.21911039352416992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,32,power_law_1.2,0.05268480181694031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,32,2,power_law_1.2,0.2736128091812134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,32,power_law_1.2,0.062412798404693604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,32,2,power_law_1.2,0.4279871940612793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,32,4,power_law_1.2,0.03356159925460815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,32,4,power_law_1.2,0.030694401264190672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,32,power_law_1.2,0.08247680068016053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,32,2,power_law_1.2,0.5499839782714844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,32,4,power_law_1.2,0.029318401217460634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,32,4,power_law_1.2,0.02942720055580139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,32,4,power_law_1.2,0.02937600016593933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,32,power_law_1.2,0.11661440134048462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,32,2,power_law_1.2,0.6359424114227294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,32,4,power_law_1.2,0.030527999997138976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,32,4,power_law_1.2,0.030374398827552794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,32,4,power_law_1.2,0.03027839958667755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,32,power_law_1.2,0.1373247981071472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,32,4,power_law_1.2,0.030847999453544616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,32,2,power_law_1.2,1.0824640274047852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,32,4,power_law_1.2,0.031148800253868104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,32,4,power_law_1.2,0.03187200129032135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,32,4,power_law_1.2,0.03206399977207184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,32,power_law_1.2,0.20266239643096923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,32,4,power_law_1.2,0.03225600123405457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,32,4,power_law_1.2,0.03546879887580871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,32,4,power_law_1.2,0.03971840143203735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,32,2,power_law_1.2,2.2694976806640623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,32,4,power_law_1.2,0.04471679925918579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,32,power_law_1.2,0.2851583957672119
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,32,4,power_law_1.2,0.04472959935665131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,32,4,power_law_1.2,0.052095997333526614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,32,4,power_law_1.2,0.05261440277099609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,32,4,power_law_1.2,0.08013439774513245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,32,4,power_law_1.2,0.08536319732666016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,32,power_law_1.2,0.40674557685852053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,32,4,power_law_1.2,0.10599039793014527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,32,4,power_law_1.2,0.12981760501861572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,32,4,power_law_1.2,0.1996608018875122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,32,4,power_law_1.2,0.2521087884902954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,32,power_law_1.2,0.4992063999176025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,32,4,power_law_1.2,0.3940608024597168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,32,4,power_law_1.2,0.3957695960998535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,32,4,power_law_1.2,0.5713856220245361
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,32,power_law_1.2,0.6203839778900146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,32,4,power_law_1.2,0.755731201171875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,32,4,power_law_1.2,1.48156795501709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,32,power_law_1.2,1.2302528381347657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,32,power_law_1.2,2.589241600036621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,8,balanced,0.054287999868392944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,8,balanced,0.07734400033950806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,8,balanced,0.09616000453631084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,8,balanced,0.14102933804194132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,8,balanced,0.20604799191157022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,8,balanced,0.25596799453099567
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,8,balanced,0.25831466913223267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,8,balanced,0.25870933135350543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,8,balanced,0.2590773304303487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,8,balanced,0.25967466831207275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,8,balanced,0.2617280085881551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,8,balanced,0.2642880082130432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,8,balanced,0.2675093412399292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,8,balanced,0.26868265867233276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,8,balanced,0.27397332588831586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,8,balanced,0.27803200483322144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,8,balanced,0.28402666250864667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,8,balanced,0.32171199719111127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,8,balanced,0.3126666744550069
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,8,balanced,0.3905920187632243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,8,balanced,0.37354131539662677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,8,balanced,0.5789546569188436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,8,balanced,0.4789919853210449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,8,balanced,0.7015893459320068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,8,balanced,0.7128053506215414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,8,balanced,1.033903996149699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,8,balanced,1.1379146575927734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,1,8,balanced,1.427994728088379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,1,8,balanced,2.091557343800863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,1,8,balanced,4.128746668497722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,1,power_law_1.2,0.08640639781951905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,1,power_law_1.2,0.09771519899368286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,32,power_law_1.2,0.059539198875427246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,32,power_law_1.2,0.048307201266288756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,1,power_law_1.2,0.11175040006637574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,32,power_law_1.2,0.047712001204490664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,1,power_law_1.2,0.1395840048789978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,32,power_law_1.2,0.047391998767852786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,32,power_law_1.2,0.05139840245246887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,1,power_law_1.2,0.18773119449615477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,32,power_law_1.2,0.06094719767570496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,32,power_law_1.2,0.059487998485565186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,1,power_law_1.2,0.24230399131774902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,32,power_law_1.2,0.06033920049667359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,32,power_law_1.2,0.06509439945220948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,1,power_law_1.2,0.2841792106628418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,32,power_law_1.2,0.06520320177078247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,32,power_law_1.2,0.06613759994506836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,1,power_law_1.2,0.3693056106567383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,32,power_law_1.2,0.07012479901313781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,32,power_law_1.2,0.0717631995677948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,1,power_law_1.2,0.3886271953582764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,32,power_law_1.2,0.07447680234909057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,32,power_law_1.2,0.08236799836158752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,2,power_law_1.01,0.07251200079917908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,1,power_law_1.2,0.39470078945159914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,32,power_law_1.2,0.08533120155334473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,32,power_law_1.2,0.09151359796524047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,1,power_law_1.2,0.4233856201171875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,32,power_law_1.2,0.10430719852447509
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,2,power_law_1.01,0.12138880491256714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,32,power_law_1.2,0.1093824028968811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,1,power_law_1.2,0.4389823913574219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,32,power_law_1.2,0.1454848051071167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,32,power_law_1.2,0.1598080039024353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,2,power_law_1.01,0.1459328055381775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,1,power_law_1.2,0.442412805557251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,32,power_law_1.2,0.22773120403289795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,32,power_law_1.2,0.26041600704193113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,2,power_law_1.01,0.11850880384445191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,1,power_law_1.2,0.45934720039367677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,32,power_law_1.2,0.34939520359039306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,1,power_law_1.2,0.49169921875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,32,power_law_1.2,0.5058559894561767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,2,power_law_1.01,0.20533759593963624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,32,power_law_1.2,0.7811264038085938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,1,power_law_1.2,0.5010111808776856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,2,power_law_1.01,0.26847999095916747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,32,power_law_1.2,0.8478336334228516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,1,power_law_1.2,0.521772813796997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,32,power_law_1.2,1.0292736053466798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,2,power_law_1.01,0.3173311948776245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,1,power_law_1.2,0.5891007900238037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,1,power_law_1.2,0.6344319820404053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,32,power_law_1.2,1.7954368591308594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,2,power_law_1.01,0.41547517776489257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,1,power_law_1.2,0.7551936149597168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,2,power_law_1.01,0.43118720054626464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,1,power_law_1.2,0.8428095817565918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,32,power_law_1.2,4.530982589721679
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,2,power_law_1.01,0.4488383769989014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,1,power_law_1.2,1.047391986846924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,2,power_law_1.01,0.47920641899108884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,1,power_law_1.2,1.2230976104736329
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,1,power_law_1.2,1.60949764251709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,2,power_law_1.01,0.5008255958557128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,2,power_law_1.01,0.5094207763671875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,1,power_law_1.2,1.9950784683227538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,2,power_law_1.01,0.5369408130645752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,1,power_law_1.2,2.755206489562988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,2,power_law_1.01,0.5709824085235595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,1,power_law_1.2,3.546342468261719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,8,4,power_law_1.01,0.03172479867935181
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,8,4,power_law_1.01,0.031200000643730165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,8,4,power_law_1.01,0.030272001028060914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,2,power_law_1.01,0.5760767936706543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,8,4,power_law_1.01,0.030694401264190672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,8,4,power_law_1.01,0.03188480138778686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,8,4,power_law_1.01,0.03160960078239441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,8,4,power_law_1.01,0.03409920036792755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,8,4,power_law_1.01,0.03455359935760498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,8,4,power_law_1.01,0.03650560081005096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,8,4,power_law_1.01,0.036550399661064145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,8,4,power_law_1.01,0.035769599676132205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,8,4,power_law_1.01,0.03760640025138855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,1,power_law_1.2,4.332883071899414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,8,4,power_law_1.01,0.0395904004573822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,2,power_law_1.01,0.6022719860076904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,8,4,power_law_1.01,0.039212799072265624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,8,4,power_law_1.01,0.04913919866085052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,8,4,power_law_1.01,0.05509759783744812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,8,4,power_law_1.01,0.060102397203445436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,8,4,power_law_1.01,0.0722815990447998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,8,4,power_law_1.01,0.08042880296707153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,8,4,power_law_1.01,0.09259520173072815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,8,4,power_law_1.01,0.10944000482559205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,8,4,power_law_1.01,0.1315392017364502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,8,4,power_law_1.01,0.18435839414596558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,2,power_law_1.01,0.6497151851654053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,8,4,power_law_1.01,0.2629568099975586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,8,4,power_law_1.01,0.33841280937194823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,8,4,power_law_1.01,0.4449151992797852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,8,4,power_law_1.01,0.6165887832641601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,1,power_law_1.2,6.5452415466308596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,8,4,power_law_1.01,0.8193599700927734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,2,power_law_1.01,0.6693120002746582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,8,4,power_law_1.01,1.4019136428833008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,8,4,power_law_1.01,2.3256959915161133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,2,power_law_1.01,0.7586175918579101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,2,power_law_1.01,0.8388735771179199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,1,power_law_1.2,12.630790710449219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,2,power_law_1.01,1.024403190612793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,2,power_law_1.01,1.151033592224121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,2,power_law_1.01,1.5144639968872071
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,2,power_law_1.01,1.6934080123901367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,2,4,power_law_1.2,0.08266879916191101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,2,4,power_law_1.2,0.10003199577331542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,2,4,power_law_1.2,0.07474560141563416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,2,4,power_law_1.2,0.0797760009765625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,2,power_law_1.01,2.251807975769043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,2,4,power_law_1.2,0.07891839742660522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,2,4,power_law_1.2,0.08158720135688782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,2,4,power_law_1.2,0.0828544020652771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,2,4,power_law_1.2,0.10349440574645996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,2,4,power_law_1.2,0.10922880172729492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,2,4,power_law_1.2,0.10975359678268433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,2,4,power_law_1.2,0.09416959881782531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,2,4,power_law_1.2,0.10189440250396728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,2,4,power_law_1.2,0.10103679895401001
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,2,4,power_law_1.2,0.11568000316619872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,2,4,power_law_1.2,0.14543360471725464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,2,4,power_law_1.2,0.14797439575195312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,2,4,power_law_1.2,0.1530303955078125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,16,4,balanced,0.033333333830038704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,2,4,power_law_1.2,0.17701760530471802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,16,4,balanced,0.03369600077470144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,16,4,balanced,0.033626665671666466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,16,4,balanced,0.03332266708215078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,2,power_law_1.01,2.8072256088256835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,16,4,balanced,0.03374933451414108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,2,4,power_law_1.2,0.24941439628601075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,16,4,balanced,0.035717333356539406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,16,4,balanced,0.035674666364987694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,16,4,balanced,0.03566399961709976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,16,4,balanced,0.03549866626660029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,16,4,balanced,0.037418665985266365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,2,4,power_law_1.2,0.32325119972229005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,16,4,balanced,0.03746666759252548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,16,4,balanced,0.037685332198937736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,16,4,balanced,0.037802666425704956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,16,4,balanced,0.03955733279387156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,16,4,balanced,0.045968001087506614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,16,4,balanced,0.04742933313051859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,2,4,power_law_1.2,0.3035968065261841
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,16,4,balanced,0.046021332343419395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,16,4,balanced,0.05414933462937673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,16,4,balanced,0.06010666489601135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,16,4,balanced,0.0727946658929189
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,16,4,balanced,0.09768533706665039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,2,4,power_law_1.2,0.5310527801513671
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,16,4,balanced,0.12961600224177042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,16,4,balanced,0.15411200126012167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,16,4,balanced,0.20586133003234863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,2,4,power_law_1.2,0.4468224048614502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,16,4,balanced,0.25474133094151813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,16,4,balanced,0.35442666212717694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,16,4,balanced,0.45478399594624835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,2,4,power_law_1.2,0.7882304191589355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,16,4,balanced,0.5565919876098633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,16,4,balanced,0.8459146817525228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,2,4,power_law_1.2,0.9286144256591797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,16,4,balanced,1.6366292635599773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,2,4,power_law_1.2,1.7300479888916016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,2,power_law_1.01,3.3666366577148437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,2,4,power_law_1.2,2.313145637512207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,2,4,power_law_1.2,2.517568016052246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,2,4,power_law_1.2,4.221247863769531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,2,power_law_1.01,4.946860885620117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,2,4,power_law_1.2,9.48551025390625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,2,power_law_1.01,9.73214111328125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,16,power_law_1.2,0.04444800019264221
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,16,power_law_1.2,0.03498240113258362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,16,power_law_1.2,0.029337599873542786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,16,power_law_1.2,0.03017599880695343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,16,power_law_1.2,0.03017599880695343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,16,power_law_1.2,0.032576000690460204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,16,power_law_1.2,0.03265919983386993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,16,power_law_1.2,0.033241599798202515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,16,power_law_1.2,0.03411200046539307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,16,power_law_1.2,0.0345984011888504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,16,power_law_1.2,0.03463039994239807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,16,1,power_law_1.2,0.030502399802207945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,16,power_law_1.2,0.035097599029541016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,16,1,power_law_1.2,0.03126400113105774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,16,1,power_law_1.2,0.03222399950027466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,16,power_law_1.2,0.03705599904060364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,16,1,power_law_1.2,0.037222400307655334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,16,1,power_law_1.2,0.0358271986246109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,16,1,power_law_1.2,0.0369024008512497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,16,power_law_1.2,0.03849599957466125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,16,1,power_law_1.2,0.04133760035037994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,16,1,power_law_1.2,0.03861120045185089
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,16,power_law_1.2,0.043084800243377686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,16,1,power_law_1.2,0.04066559970378876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,16,1,power_law_1.2,0.03950720131397247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,16,1,power_law_1.2,0.04362240135669708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,16,power_law_1.2,0.04332799911499023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,16,1,power_law_1.2,0.04849919974803925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,16,1,power_law_1.2,0.0507968008518219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,16,1,power_law_1.2,0.05516800284385681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,16,power_law_1.2,0.04588159918785095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,16,1,power_law_1.2,0.06401280164718628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,16,1,power_law_1.2,0.06661760210990905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,16,1,power_law_1.2,0.0682752013206482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,16,power_law_1.2,0.04952960014343262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,16,1,power_law_1.2,0.08542079925537109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,16,1,power_law_1.2,0.09157760143280029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,16,power_law_1.2,0.05252479910850525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,16,1,power_law_1.2,0.1167296051979065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,32,balanced,0.06651199857393901
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,16,1,power_law_1.2,0.1294144034385681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,32,balanced,0.06693333387374878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,32,balanced,0.06633066634337108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,16,1,power_law_1.2,0.1696768045425415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,16,power_law_1.2,0.05839359760284424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,32,balanced,0.0771679977575938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,32,balanced,0.09335466225941975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,16,1,power_law_1.2,0.21137280464172364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,32,balanced,0.12053866187731425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,32,balanced,0.1602720022201538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,16,1,power_law_1.2,0.2945087909698486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,32,balanced,0.15406399965286255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,16,power_law_1.2,0.07219200134277344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,32,balanced,0.15617066621780396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,32,balanced,0.15682133038838705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,16,1,power_law_1.2,0.37998080253601074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,32,balanced,0.1711733341217041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,32,balanced,0.1677173376083374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,32,balanced,0.16726400454839072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,16,1,power_law_1.2,0.5483776092529297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,16,power_law_1.2,0.0886847972869873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,32,balanced,0.16175466775894165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,32,balanced,0.16572800278663635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,32,balanced,0.16851200660069784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,32,balanced,0.1718133290608724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,16,1,power_law_1.2,0.7065408229827881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,32,balanced,0.17686933279037476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,32,balanced,0.18303465843200684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,16,power_law_1.2,0.11879040002822876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,32,balanced,0.19667200247446695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,16,1,power_law_1.2,0.8652671813964844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,32,balanced,0.20695465803146362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,32,balanced,0.23519466320673624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,32,balanced,0.2609226703643799
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,32,balanced,0.3177280028661092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,16,power_law_1.2,0.14531199932098388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,32,balanced,0.40561068058013916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,16,1,power_law_1.2,1.344428825378418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,32,balanced,0.494922677675883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,32,balanced,0.6350666681925455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,16,power_law_1.2,0.17701760530471802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,32,balanced,0.8191946347554525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,32,balanced,1.2177546819051106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,16,1,power_law_1.2,2.623660850524902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,16,power_law_1.2,0.2794624090194702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,32,balanced,2.245685259501139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,16,power_law_1.2,0.3242687940597534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,2,16,power_law_1.2,0.3957695960998535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,2,16,power_law_1.2,0.6175551891326905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,2,16,power_law_1.2,1.4735424041748046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,16,4,power_law_1.2,0.03191039860248566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,16,4,power_law_1.2,0.02884480059146881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,16,4,power_law_1.2,0.02927359938621521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,16,4,power_law_1.2,0.029523199796676634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,16,4,power_law_1.2,0.0304639995098114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,16,4,power_law_1.2,0.030700799822807313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,16,4,power_law_1.2,0.031737598776817325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,16,4,power_law_1.2,0.03283199965953827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,16,4,power_law_1.2,0.033580800890922545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,16,4,power_law_1.2,0.03416320085525513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,16,4,power_law_1.2,0.03431040048599243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,16,4,power_law_1.2,0.0358271986246109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,2,power_law_1.01,0.07153919935226441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,16,4,power_law_1.2,0.03597440123558045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,16,4,power_law_1.2,0.04164479970932007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,16,4,power_law_1.2,0.04552960097789764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,16,4,power_law_1.2,0.04798719882965088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,16,4,power_law_1.2,0.05076479911804199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,16,4,power_law_1.2,0.05596799850463867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,16,4,power_law_1.2,0.06510720252990723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,16,4,power_law_1.2,0.09059839844703674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,16,4,power_law_1.2,0.09747200012207032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,16,4,power_law_1.2,0.13038719892501832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,16,4,power_law_1.2,0.16494719982147216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,16,4,power_law_1.2,0.25357439517974856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,2,power_law_1.01,0.08941439986228943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,16,4,power_law_1.2,0.2892224073410034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,16,4,power_law_1.2,0.4639743804931641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,16,4,power_law_1.2,0.6245952129364014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,2,1,power_law_1.2,0.07960960268974304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,2,1,power_law_1.2,0.11285760402679443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,2,1,power_law_1.2,0.1483456015586853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,16,4,power_law_1.2,0.8119808197021484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,2,1,power_law_1.2,0.22961280345916749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,2,1,power_law_1.2,0.23505918979644774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,2,1,power_law_1.2,0.23539199829101562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,16,4,power_law_1.2,1.1727295875549317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,2,1,power_law_1.2,0.23850879669189454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,2,1,power_law_1.2,0.23367679119110107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,2,1,power_law_1.2,0.24508159160614013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,2,1,power_law_1.2,0.24973440170288086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,2,1,power_law_1.2,0.25500800609588625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,16,4,power_law_1.2,2.1336896896362303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,2,1,power_law_1.2,0.26426880359649657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,2,1,power_law_1.2,0.2634176015853882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,2,1,power_law_1.2,0.26243200302124026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,2,1,power_law_1.2,0.2927103996276855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,2,power_law_1.01,0.07980160117149353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,2,1,power_law_1.2,0.29745919704437257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,2,1,power_law_1.2,0.32286078929901124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,2,1,power_law_1.2,0.35278079509735105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,2,1,power_law_1.2,0.39127678871154786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,2,1,power_law_1.2,0.4875391960144043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,2,1,power_law_1.2,0.5725503921508789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,2,1,power_law_1.2,0.743500804901123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,2,1,power_law_1.2,0.9164287567138671
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,2,1,power_law_1.2,1.2574336051940918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,2,power_law_1.01,0.10899200439453124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,2,1,power_law_1.2,1.6274112701416015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,2,1,power_law_1.2,2.369094467163086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,2,1,power_law_1.2,3.075391960144043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,2,power_law_1.01,0.13764480352401734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,2,1,power_law_1.2,3.8922496795654298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,2,1,power_law_1.2,5.970969772338867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,2,power_law_1.01,0.19938559532165528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,2,2,power_law_1.01,0.03139840066432953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,2,1,power_law_1.2,11.640409851074219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,2,2,power_law_1.01,0.0328575998544693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,2,power_law_1.01,0.20840959548950194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,2,2,power_law_1.01,0.03560959994792938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,2,2,power_law_1.01,0.04140160083770752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,2,2,power_law_1.01,0.048179200291633605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,2,2,power_law_1.01,0.06723840236663818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,2,2,power_law_1.01,0.07079039812088013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,2,power_law_1.01,0.22157440185546876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,2,2,power_law_1.01,0.07527679800987244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,2,2,power_law_1.01,0.07490559816360473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,2,2,power_law_1.01,0.0750976026058197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,2,2,power_law_1.01,0.07717120051383972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,2,power_law_1.01,0.2294975996017456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,2,2,power_law_1.01,0.08079360127449035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,2,2,power_law_1.01,0.081523197889328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,2,2,power_law_1.01,0.08223999738693237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,2,power_law_1.01,0.22624640464782714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,2,2,power_law_1.01,0.08695039749145508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,2,2,power_law_1.01,0.08935040235519409
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,2,2,power_law_1.01,0.09696000218391418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,2,2,power_law_1.01,0.10647679567337036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,2,power_law_1.01,0.23964159488677977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,2,power_law_1.01,0.11818239688873292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,2,power_law_1.01,0.14244480133056642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,2,power_law_1.01,0.1486143946647644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,2,power_law_1.01,0.24920320510864258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,2,power_law_1.01,0.18497920036315918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,2,power_law_1.01,0.22363519668579102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,2,power_law_1.01,0.31553919315338136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,32,balanced,0.05497066676616669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,32,balanced,0.04403733213742574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,32,balanced,0.04571199913819631
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,32,balanced,0.04770666857560476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,32,balanced,0.047872001926104225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,32,balanced,0.05378133555253347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,2,power_law_1.01,0.2547584056854248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,32,balanced,0.06910933554172516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,32,balanced,0.07015466690063477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,32,balanced,0.06853866577148438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,32,balanced,0.07195733487606049
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,32,balanced,0.0705386648575465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,32,balanced,0.07267733414967854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,2,power_law_1.01,0.3769792079925537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,32,balanced,0.07222400108973186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,32,balanced,0.07267733414967854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,32,balanced,0.07939200103282928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,32,balanced,0.07974400122960408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,32,balanced,0.08339732885360718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,32,balanced,0.09150399764378865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,32,balanced,0.0951039989789327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,32,balanced,0.1092800001303355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,32,balanced,0.11920533577601115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,32,balanced,0.1428053379058838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,32,balanced,0.1697173317273458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,32,balanced,0.21781333287556967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,32,balanced,0.2635413408279419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,2,power_law_1.01,0.5111360073089599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,32,balanced,0.37042665481567383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,32,balanced,0.45605866114298504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,32,balanced,0.5600373347600301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,32,balanced,0.8265386422475179
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,2,power_law_1.01,0.2642816066741943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,32,balanced,1.6103199323018391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,2,power_law_1.01,0.7004992008209229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,2,power_law_1.01,0.8175871849060059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,2,power_law_1.01,0.27538559436798093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,2,power_law_1.01,1.2837120056152345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,2,power_law_1.01,0.28517119884490966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,2,power_law_1.01,2.561625671386719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,2,power_law_1.01,0.30543999671936034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,2,power_law_1.01,0.34690558910369873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,2,power_law_1.01,0.38297600746154786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,32,balanced,0.0765066643555959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,32,balanced,0.052245333790779114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,32,balanced,0.06339733302593231
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,32,balanced,0.07764266431331635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,32,balanced,0.10772266983985901
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,32,balanced,0.11520533760388692
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,32,balanced,0.11618666847546895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,32,balanced,0.11726933717727661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,32,balanced,0.11618133385976155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,32,balanced,0.11661866307258606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,32,balanced,0.11776533722877502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,32,balanced,0.11905066172281902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,32,balanced,0.11779200037320454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,32,balanced,0.11975466211636861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,32,balanced,0.12476266423861186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,32,balanced,0.12473066647847493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,32,balanced,0.12717333436012268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,32,balanced,0.13391466935475668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,32,balanced,0.13411200046539307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,32,balanced,0.15955199797948202
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,32,balanced,0.1529866655667623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,32,balanced,0.21100266774495444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,32,balanced,0.19751467307408652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,32,balanced,0.2587253252665202
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,32,balanced,0.28014399607976276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,32,balanced,0.36556800206502277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,32,balanced,0.46023468176523846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,2,power_law_1.01,0.4755648136138916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,32,balanced,0.5246719916661581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,32,balanced,0.7936480045318604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,32,balanced,1.4839040438334148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,2,power_law_1.01,0.5876287937164306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,128,balanced,0.05413866539796194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,128,balanced,0.05389333268006643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,128,balanced,0.05213333169619242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,128,balanced,0.054192001620928444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,128,balanced,0.05590933561325073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,128,balanced,0.055733333031336464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,128,balanced,0.055813332398732506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,2,power_law_1.01,0.7414976119995117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,128,balanced,0.05596266686916351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,128,balanced,0.05597866574923197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,128,balanced,0.0580320010582606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,128,balanced,0.05789866546789805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,128,balanced,0.059989333152770996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,128,balanced,0.06025599936644236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,128,balanced,0.0601440022389094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,128,balanced,0.06647466619809468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,128,balanced,0.06614933411280315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,128,balanced,0.07008000214894612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,128,balanced,0.0783786674340566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,128,balanced,0.08474666873613994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,128,balanced,0.1032319962978363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,128,balanced,0.11754133303960164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,128,balanced,0.15639999508857727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,128,balanced,0.1525973379611969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,128,balanced,0.19909866650899252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,128,balanced,0.2220159967740377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,128,balanced,0.2940799991289775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,128,balanced,0.3532586495081584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,2,128,balanced,0.4224426746368408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,2,128,balanced,0.6119946638743082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,2,power_law_1.01,0.8842559814453125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,2,128,balanced,1.336330731709798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,2,power_law_1.01,1.1891584396362305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,4,balanced,0.04822400212287903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,4,balanced,0.047968000173568726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,4,balanced,0.054042667150497437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,4,balanced,0.07307733098665874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,4,balanced,0.10200533270835876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,4,balanced,0.15320533514022827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,2,power_law_1.01,1.4960127830505372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,4,balanced,0.15425599614779154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,4,balanced,0.15465066830317178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,4,balanced,0.15946666399637857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,4,balanced,0.1586720049381256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,4,balanced,0.1583466629187266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,4,balanced,0.15902400016784668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,4,balanced,0.16482667128245035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,4,balanced,0.16309866309165955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,4,balanced,0.1702559987703959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,4,balanced,0.17196265856424967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,4,balanced,0.17781867583592734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,4,balanced,0.19378666083017984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,4,balanced,0.20719999074935913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,4,balanced,0.23587733507156372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,4,balanced,0.2530933419863383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,4,balanced,0.31726400057474774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,4,balanced,0.3625333309173584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,4,balanced,0.4682026704152425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,4,balanced,0.5555306673049927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,4,balanced,0.78001602490743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,4,balanced,0.9814559618631998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,2,power_law_1.2,0.12227840423583984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,4,balanced,1.2001226743062336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,4,balanced,1.803434689839681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,4,balanced,3.4901866912841797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,2,power_law_1.01,2.0711231231689453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,2,power_law_1.2,0.14167040586471558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,2,power_law_1.2,0.13616000413894652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,2,balanced,0.06005333364009857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,2,balanced,0.0625493327776591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,2,balanced,0.07453866799672444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,2,balanced,0.09834667046864827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,2,balanced,0.14332800110181174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,2,balanced,0.21782400210698447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,2,balanced,0.2221013307571411
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,2,balanced,0.2244373361269633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,2,balanced,0.22814400990804037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,2,balanced,0.23080533742904663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,2,balanced,0.23227733373641968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,2,balanced,0.23643199602762857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,2,balanced,0.23645333449045816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,2,balanced,0.23586134115854898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,2,balanced,0.2457759976387024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,2,balanced,0.24909865856170654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,2,balanced,0.25758934020996094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,2,balanced,0.2788319985071818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,2,power_law_1.01,2.684979248046875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,2,power_law_1.2,0.19456640481948853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,2,balanced,0.29947733879089355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,2,balanced,0.33816532293955487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,2,balanced,0.37459735075632733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,2,balanced,0.4625493288040161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,2,balanced,0.5478880008061727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,2,balanced,0.7683093547821045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,2,balanced,0.9311839739481608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,2,balanced,1.352666695912679
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,2,balanced,1.7481279373168945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,2,power_law_1.2,0.2559936046600342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,2,balanced,2.1574559211730957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,2,balanced,3.271514574686686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,2,power_law_1.2,0.4197824001312256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,2,balanced,6.396528244018555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,8,2,power_law_1.01,3.2667263031005858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,2,power_law_1.2,0.47664642333984375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,128,balanced,0.03974399964014689
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,2,power_law_1.2,0.493177604675293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,128,balanced,0.03994666785001755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,128,balanced,0.040192000567913055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,128,balanced,0.0415786678592364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,128,balanced,0.040074666341145836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,128,balanced,0.04171200096607208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,128,balanced,0.04177066683769226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,128,balanced,0.04161600023508072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,128,balanced,0.041989331444104515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,128,balanced,0.04171733558177948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,128,balanced,0.04353066782156626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,128,balanced,0.04359466830889384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,128,balanced,0.042810668547948204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,128,balanced,0.0436160018046697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,128,balanced,0.0476800004641215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,128,balanced,0.04879466692606608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,128,balanced,0.05017066498597463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,128,balanced,0.05005866785844167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,128,balanced,0.054383998115857445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,128,balanced,0.05605866511662801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,128,balanced,0.05997333427270254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,128,balanced,0.06817600131034851
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,128,balanced,0.07791466514269511
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,128,balanced,0.09897599617640178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,128,balanced,0.1156160036722819
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,128,balanced,0.1497599979241689
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,128,balanced,0.18418665726979574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,128,balanced,0.21613333622614542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,2,power_law_1.2,0.5044288158416748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,128,balanced,0.31964800755182904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,128,balanced,0.5867040157318115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,8,2,power_law_1.01,5.007404708862305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,2,power_law_1.2,0.5281536102294921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,2,power_law_1.2,0.5434688091278076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,2,power_law_1.2,0.5842944145202636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,16,8,power_law_1.01,0.030195200443267824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,16,8,power_law_1.01,0.029894399642944335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,16,8,power_law_1.01,0.030508801341056824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,16,8,power_law_1.01,0.0290367990732193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,2,power_law_1.2,0.5829823970794678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,16,8,power_law_1.01,0.030502399802207945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,16,8,power_law_1.01,0.030847999453544616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,16,8,power_law_1.01,0.03188480138778686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,16,8,power_law_1.01,0.03243519961833954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,16,8,power_law_1.01,0.033670398592948916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,16,8,power_law_1.01,0.03442560136318207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,16,8,power_law_1.01,0.0343423992395401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,16,8,power_law_1.01,0.03588480055332184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,16,8,power_law_1.01,0.03699199855327606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,2,power_law_1.2,0.6037119865417481
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,16,8,power_law_1.01,0.03885439932346344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,16,8,power_law_1.01,0.04776319861412048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,16,8,power_law_1.01,0.04865919947624207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,16,8,power_law_1.01,0.05175039768218994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,16,8,power_law_1.01,0.06312959790229797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,16,8,power_law_1.01,0.06395519971847534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,16,8,power_law_1.01,0.0870527982711792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,16,8,power_law_1.01,0.11716480255126953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,8,2,power_law_1.01,9.426918029785156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,2,power_law_1.2,0.6363647937774658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,16,8,power_law_1.01,0.15564800500869752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,16,8,power_law_1.01,0.1830847978591919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,16,8,power_law_1.01,0.2605504035949707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,16,8,power_law_1.01,0.32471039295196535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,16,8,power_law_1.01,0.4664192199707031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,2,power_law_1.2,0.04708479940891266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,16,8,power_law_1.01,0.6052480220794678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,2,power_law_1.2,0.6545792102813721
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,16,8,power_law_1.01,0.7700223922729492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,2,power_law_1.2,0.05379199981689453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,16,8,power_law_1.01,1.1655488014221191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,2,power_law_1.2,0.05768960118293762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,2,power_law_1.2,0.6888383865356446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,16,8,power_law_1.01,2.16876163482666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,2,power_law_1.2,0.06766719818115234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,2,power_law_1.2,0.08723840117454529
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,2,power_law_1.2,0.7511807918548584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,2,power_law_1.2,0.10309120416641235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,2,power_law_1.2,0.13356800079345704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,2,power_law_1.2,0.8318592071533203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,2,power_law_1.2,0.14062080383300782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,2,power_law_1.2,0.1493631958961487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,2,power_law_1.2,1.0129728317260742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,2,power_law_1.2,0.1546880006790161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,2,power_law_1.2,0.15750399827957154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,2,power_law_1.2,1.1133248329162597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,2,power_law_1.2,0.1688256025314331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,2,power_law_1.2,0.17155840396881103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,64,power_law_1.01,0.12647680044174195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,2,power_law_1.2,1.392793560028076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,2,power_law_1.2,0.18483200073242187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,2,power_law_1.2,0.20032000541687012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,64,power_law_1.01,0.12257280349731445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,16,power_law_1.2,0.034764799475669864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,2,power_law_1.2,0.21368319988250734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,2,power_law_1.2,1.6277759552001954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,16,power_law_1.2,0.0281792014837265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,2,power_law_1.2,0.21379199028015136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,16,power_law_1.2,0.029209598898887634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,64,power_law_1.01,0.11688319444656373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,16,power_law_1.2,0.027750399708747864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,2,power_law_1.2,0.24559359550476073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,16,power_law_1.2,0.029209598898887634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,2,power_law_1.2,2.1551231384277343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,2,power_law_1.2,0.2866624116897583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,16,power_law_1.2,0.030022400617599487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,64,power_law_1.01,0.09916160106658936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,16,power_law_1.2,0.03054719865322113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,2,power_law_1.2,0.35072638988494875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,16,power_law_1.2,0.031033599376678468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,16,power_law_1.2,0.030976000428199767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,64,power_law_1.01,0.10641920566558838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,2,power_law_1.2,0.3983488082885742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,16,power_law_1.2,0.032179200649261476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,2,power_law_1.2,2.760588836669922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,16,power_law_1.2,0.032128000259399415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,2,power_law_1.2,0.5222400188446045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,16,power_law_1.2,0.03333759903907776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,64,power_law_1.01,0.1130944013595581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,16,power_law_1.2,0.03454720079898834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,2,power_law_1.2,0.6040703773498535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,16,power_law_1.2,0.03457280099391937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,64,power_law_1.01,0.12868479490280152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,16,power_law_1.2,0.039750400185585025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,2,power_law_1.2,0.8637568473815918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,16,power_law_1.2,0.040627199411392215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,16,power_law_1.2,0.042080000042915344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,2,power_law_1.2,3.849932861328125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,64,power_law_1.01,0.11989120244979859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,16,power_law_1.2,0.04485119879245758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,2,power_law_1.2,1.0153535842895507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,16,power_law_1.2,0.051795202493667605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,16,power_law_1.2,0.05985280275344849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,64,power_law_1.01,0.12433279752731323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,16,power_law_1.2,0.06338559985160827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,2,power_law_1.2,1.500819206237793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,16,power_law_1.2,0.08241279721260071
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,64,power_law_1.01,0.13040000200271606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,16,power_law_1.2,0.09820799827575684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,2,power_law_1.2,1.9845056533813477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,2,power_law_1.2,4.872608184814453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,16,power_law_1.2,0.1457856059074402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,64,power_law_1.01,0.12399359941482543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,16,power_law_1.2,0.1852671980857849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,64,power_law_1.01,0.12970240116119386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,2,power_law_1.2,2.3785919189453124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,16,power_law_1.2,0.27178239822387695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,16,power_law_1.2,0.4575808048248291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,64,power_law_1.01,0.12561919689178466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,16,power_law_1.2,0.5842879772186279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,2,power_law_1.2,3.7852222442626955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,1,2,power_law_1.2,5.969535827636719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,64,power_law_1.01,0.12400640249252319
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,16,power_law_1.2,0.987609577178955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,64,power_law_1.01,0.12607359886169434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,64,power_law_1.01,0.1300287961959839
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,16,power_law_1.2,1.8160127639770507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,64,power_law_1.01,0.12934399843215943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,2,power_law_1.2,6.871135711669922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,64,power_law_1.01,0.167193603515625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,1,2,power_law_1.2,9.109568023681641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,64,power_law_1.01,0.16766719818115233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,64,power_law_1.01,0.17747199535369873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,64,power_law_1.01,0.22829439640045165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,64,power_law_1.01,0.26600959300994875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,64,power_law_1.01,0.32127358913421633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,64,power_law_1.01,0.43167362213134763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,64,power_law_1.01,0.514086389541626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,64,power_law_1.01,0.7086400032043457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,16,8,balanced,0.03409066547950109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,16,8,balanced,0.031925333042939506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,1,2,power_law_1.2,18.480812072753906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,16,8,balanced,0.03176533430814743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,16,8,balanced,0.03182400017976761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,16,8,balanced,0.03200533241033554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,16,8,balanced,0.03383466601371765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,16,8,balanced,0.03196800003449122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,16,8,balanced,0.032127998769283295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,16,8,balanced,0.03164266546567281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,16,8,balanced,0.03391999999682108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,16,8,balanced,0.03372266640265783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,16,8,balanced,0.03576533248027166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,16,8,balanced,0.034058667719364166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,4,power_law_1.01,0.14689919948577881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,16,8,balanced,0.03357866654793421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,16,8,balanced,0.03972266614437103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,16,8,balanced,0.03952533255020777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,16,8,balanced,0.040005333721637726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,64,power_law_1.01,0.9220095634460449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,16,8,balanced,0.041946664452552795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,16,8,balanced,0.044010668992996216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,16,8,balanced,0.050794666012128196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,16,8,balanced,0.05451733370621999
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,4,power_law_1.01,0.23737599849700927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,16,8,balanced,0.07149866720040639
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,16,8,balanced,0.08920533458391826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,16,8,balanced,0.13030933340390524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,16,8,balanced,0.15893866618474325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,16,8,balanced,0.2251733342806498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,16,8,balanced,0.26343466838200885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,16,8,balanced,0.3181013266245524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,16,8,balanced,0.4799413283665975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,4,power_law_1.01,0.12993279695510865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,16,8,balanced,0.9104373455047607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,64,power_law_1.01,1.1777664184570313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,4,power_law_1.01,0.19785599708557128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,4,power_law_1.01,0.31306240558624265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,4,power_law_1.01,0.4158016204833984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,64,power_law_1.01,1.8028287887573242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,4,power_law_1.01,0.5987904071807861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,4,power_law_1.01,0.6488512039184571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,4,power_law_1.01,0.6467008113861084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,64,power_law_1.01,3.6581249237060547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,4,power_law_1.01,0.6626239776611328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,4,power_law_1.01,0.7106495857238769
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,4,power_law_1.01,0.7092991828918457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,4,power_law_1.01,0.7094783782958984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,2,power_law_1.2,0.09752960205078125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,4,power_law_1.01,0.6884416103363037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,2,power_law_1.2,0.14632320404052734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,4,power_law_1.01,0.7605760097503662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,4,power_law_1.01,0.7517695903778077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,2,power_law_1.2,0.15816960334777833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,32,power_law_1.01,0.1889791965484619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,4,power_law_1.01,0.8013567924499512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,2,power_law_1.2,0.1852671980857849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,32,power_law_1.01,0.2837951898574829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,4,power_law_1.01,0.835916805267334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,2,power_law_1.2,0.2965248107910156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,32,power_law_1.01,0.24002559185028077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,4,power_law_1.01,0.904319953918457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,4,power_law_1.01,1.015500831604004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,32,power_law_1.01,0.1115839958190918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,2,power_law_1.2,0.42892160415649416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,4,8,balanced,0.03827733298142751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,4,8,balanced,0.04095466683308283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,4,8,balanced,0.03726933399836222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,4,power_law_1.01,0.9846336364746093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,4,8,balanced,0.03930133332808813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,4,8,balanced,0.03849066545565923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,4,8,balanced,0.03711466739575068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,4,8,balanced,0.0386559988061587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,4,8,balanced,0.038959999879201256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,4,8,balanced,0.03994666785001755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,32,power_law_1.01,0.1461951971054077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,4,8,balanced,0.039093332986036934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,4,8,balanced,0.04273599882920583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,2,power_law_1.2,0.5341184139251709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,4,8,balanced,0.043893332282702126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,4,8,balanced,0.04298666616280874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,4,8,balanced,0.04292800029118856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,4,8,balanced,0.05301866432030996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,4,8,balanced,0.05481066803137461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,4,power_law_1.01,1.132096004486084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,4,8,balanced,0.05950933198134104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,4,8,balanced,0.0713973343372345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,4,8,balanced,0.0755626658598582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,4,8,balanced,0.09108799695968628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,4,8,balanced,0.10452266534169515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,4,8,balanced,0.12890133261680603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,4,8,balanced,0.1593173344930013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,32,power_law_1.01,0.17831679582595825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,4,8,balanced,0.2092746694882711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,4,8,balanced,0.2635519901911418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,2,power_law_1.2,0.7369664192199707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,4,8,balanced,0.36434133847554523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,4,8,balanced,0.4591946601867676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,4,power_law_1.01,1.290003204345703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,4,8,balanced,0.5603146553039551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,4,8,balanced,0.8544267018636068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,4,8,balanced,1.625882625579834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,32,power_law_1.01,0.19825279712677002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,2,power_law_1.2,0.7811583995819091
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,4,power_law_1.01,1.58023681640625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,32,power_law_1.01,0.22466559410095216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,2,power_law_1.2,0.830355167388916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,4,power_law_1.01,1.7458112716674805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,32,power_law_1.01,0.2190783977508545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,2,power_law_1.2,0.879475212097168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,4,power_law_1.01,2.4208959579467773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,32,power_law_1.01,0.22067840099334718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,2,power_law_1.2,0.8777983665466309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,32,power_law_1.01,0.22042880058288575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,4,power_law_1.01,2.8176448822021483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,2,power_law_1.2,0.9143744468688965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,32,power_law_1.01,0.23243520259857178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,2,power_law_1.2,0.9565695762634278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,32,power_law_1.01,0.22088320255279542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,4,power_law_1.01,3.3778560638427733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,32,power_law_1.01,0.2294015884399414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,2,power_law_1.2,0.9958016395568847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,32,power_law_1.01,0.2405951976776123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,4,power_law_1.01,4.940172958374023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,2,power_law_1.2,1.034553623199463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,1,balanced,0.08062399923801422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,1,balanced,0.09634666641553243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,32,power_law_1.01,0.22527360916137695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,1,balanced,0.11528533697128296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,1,balanced,0.15754133462905884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,1,balanced,0.23813333113988241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,1,balanced,0.39212266604105633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,1,balanced,0.3977546691894531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,2,power_law_1.2,1.064192008972168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,1,balanced,0.40563734372456867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,1,balanced,0.4074133237202962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,1,balanced,0.41023464997609455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,1,balanced,0.40809067090352374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,32,power_law_1.01,0.2440704107284546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,1,balanced,0.4119733174641927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,1,balanced,0.4132586717605591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,1,balanced,0.4241439898808797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,1,balanced,0.4289120038350423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,1,balanced,0.4323519865671794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,2,power_law_1.2,1.1735103607177735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,1,balanced,0.44893332322438556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,1,balanced,0.4910506804784139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,1,balanced,0.5196053187052408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,32,power_law_1.01,0.2576128005981445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,1,balanced,0.6047893365224203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,1,balanced,0.6593279838562012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,1,balanced,0.816490650177002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,1,balanced,0.9694186846415201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,4,power_law_1.01,8.888614654541016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,1,balanced,1.3810133934020996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,2,power_law_1.2,1.2918208122253418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,32,power_law_1.01,0.26168320178985593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,1,balanced,1.6702027320861816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,1,balanced,2.4333653450012207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,1,balanced,3.1578613917032876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,2,power_law_1.2,1.4612416267395019
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,32,power_law_1.01,0.2843839883804321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,1,balanced,3.917109489440918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,32,power_law_1.01,0.3539200067520142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,2,power_law_1.2,1.5027775764465332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,1,balanced,5.979040145874023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,32,power_law_1.01,0.384115195274353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,2,power_law_1.2,1.795212745666504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,1,balanced,11.568730672200521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,32,power_law_1.01,0.45854721069335935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,2,power_law_1.2,2.0085248947143555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,32,power_law_1.01,0.5925568103790283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,2,power_law_1.2,2.4728960037231444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,32,power_law_1.01,0.7179647922515869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,32,power_law_1.01,1.035801601409912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,2,power_law_1.2,2.631328010559082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,4,power_law_1.2,0.07125120162963867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,32,power_law_1.01,1.2319295883178711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,2,power_law_1.2,3.362073516845703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,4,power_law_1.2,0.11445120573043824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,4,power_law_1.2,0.06286720037460328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,32,power_law_1.01,1.5896767616271972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,4,power_law_1.2,0.08123520016670227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,2,power_law_1.2,4.010182571411133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,4,power_law_1.2,0.10859520435333252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,1,balanced,0.06340266764163971
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,1,balanced,0.09486933549245198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,1,balanced,0.10368000467618306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,4,power_law_1.2,0.1305343985557556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,1,balanced,0.1570186714331309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,1,balanced,0.25408534208933514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,1,balanced,0.45125333468119305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,32,power_law_1.01,2.2824447631835936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,1,balanced,0.6493653456370035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,1,balanced,0.6575466791788737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,1,balanced,0.6702880064646403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,4,power_law_1.2,0.18347519636154175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,1,balanced,0.6669493516286215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,1,balanced,0.6765279769897461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,1,balanced,0.6799946626027426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,1,balanced,0.6872266928354899
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,1,balanced,0.6924959818522135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,2,power_law_1.2,4.86242561340332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,4,power_law_1.2,0.19610879421234131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,1,balanced,0.7017066478729248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,1,balanced,0.707050641377767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,1,balanced,0.7319839795430502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,1,balanced,0.7735679944356283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,1,balanced,0.8053173224131266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,4,power_law_1.2,0.19850239753723145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,1,balanced,0.8704266548156738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,1,balanced,0.9427253405253092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,1,balanced,1.0862666765848796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,4,power_law_1.2,0.19050240516662598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,1,balanced,1.2736319700876872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,1,balanced,1.6031626065572102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,1,balanced,1.9939200083414714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,32,power_law_1.01,4.69793930053711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,4,power_law_1.2,0.2121216058731079
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,1,balanced,2.590288003285726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,2,power_law_1.2,6.692582702636718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,4,power_law_1.2,0.21363840103149415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,1,balanced,3.4473066329956055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,4,power_law_1.2,0.22308480739593506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,1,balanced,4.206021308898926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,4,power_law_1.2,0.23625600337982178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,1,balanced,6.130949020385742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,4,power_law_1.2,0.2452415943145752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,4,power_law_1.2,0.24595839977264405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,1,balanced,11.596234639485678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,4,power_law_1.2,0.26800639629364015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,4,power_law_1.2,0.2912447929382324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,2,power_law_1.2,13.123175048828125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,4,power_law_1.2,0.3140352010726929
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,4,power_law_1.2,0.37678720951080324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,4,power_law_1.2,0.40909438133239745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,4,power_law_1.2,0.5173759937286377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,4,power_law_1.2,0.5346816062927247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,4,power_law_1.2,0.03442560136318207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,4,power_law_1.2,0.7044159889221191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,4,power_law_1.2,0.030099201202392577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,4,power_law_1.2,0.03281280100345611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,4,power_law_1.2,0.8327360153198242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,4,power_law_1.2,0.03675520122051239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,4,power_law_1.2,0.03722879886627197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,4,power_law_1.2,1.2990079879760743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,4,power_law_1.2,0.05023360252380371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,4,power_law_1.2,0.05180799961090088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,4,power_law_1.2,0.052147197723388675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,32,power_law_1.01,0.06307839751243591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,4,power_law_1.2,0.05283839702606201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,4,power_law_1.2,1.541420841217041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,4,power_law_1.2,0.05249279737472534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,32,power_law_1.01,0.06368640065193176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,4,power_law_1.2,0.05553280115127564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,4,power_law_1.2,0.05588480234146118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,4,power_law_1.2,1.8505472183227538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,32,power_law_1.01,0.04947839975357056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,4,power_law_1.2,0.058412802219390866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,4,power_law_1.2,0.061273598670959474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,32,power_law_1.01,0.052262401580810545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,4,power_law_1.2,0.06330239772796631
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,4,power_law_1.2,0.06869760155677795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,32,power_law_1.01,0.05454720258712768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,4,power_law_1.2,2.7134527206420898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,4,power_law_1.2,0.0695680022239685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,4,power_law_1.2,0.07534080147743225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,32,power_law_1.01,0.07282559871673584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,4,power_law_1.2,0.07749760150909424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,32,power_law_1.01,0.0638592004776001
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,4,power_law_1.2,0.09054719805717468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,4,power_law_1.2,0.10259840488433838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,32,power_law_1.01,0.07867519855499268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,32,power_law_1.01,0.06365439891815186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,4,power_law_1.2,0.12557439804077147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,4,power_law_1.2,0.15062400102615356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,4,power_law_1.2,5.713663864135742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,32,power_law_1.01,0.06247040033340454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,4,power_law_1.2,0.1879807949066162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,32,power_law_1.01,0.058303999900817874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,4,power_law_1.2,0.23637759685516357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,32,power_law_1.01,0.06437119841575623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,4,power_law_1.2,0.31848959922790526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,32,power_law_1.01,0.05950080156326294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,32,power_law_1.01,0.06521599888801574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,4,power_law_1.2,0.3902719974517822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,32,power_law_1.01,0.06962559819221496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,32,power_law_1.01,0.06427519917488098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,4,4,power_law_1.2,0.48611841201782224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,32,power_law_1.01,0.07002879977226258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,4,4,power_law_1.2,0.6443967819213867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,32,power_law_1.01,0.07063680291175842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,32,power_law_1.01,0.07303040027618408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,32,power_law_1.01,0.07530879974365234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,32,power_law_1.01,0.06915199756622314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,4,4,power_law_1.2,1.3668992042541503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,32,power_law_1.01,0.08206719756126404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,32,power_law_1.01,0.08414720296859741
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,32,power_law_1.01,0.06943359971046448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,32,power_law_1.01,0.09039999842643738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,32,power_law_1.01,0.09978880286216736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,32,power_law_1.01,0.07274879813194275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,16,balanced,0.04383466641108195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,16,balanced,0.0439573327700297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,16,balanced,0.044906665881474815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,16,balanced,0.05013866722583771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,32,power_law_1.01,0.11004799604415894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,16,balanced,0.07072533170382182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,16,balanced,0.09661866227785747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,16,balanced,0.09914666414260864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,16,balanced,0.0978559950987498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,16,balanced,0.09851200381914775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,16,balanced,0.09788800279299419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,16,balanced,0.10003200173377991
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,16,balanced,0.09908800323804219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,16,balanced,0.10141332944234212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,16,balanced,0.10145599643389384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,16,balanced,0.10912533601125081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,16,balanced,0.10758933424949646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,16,balanced,0.11098666985829671
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,32,power_law_1.01,0.07226880192756653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,16,balanced,0.117658664782842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,16,balanced,0.12040000160535176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,16,balanced,0.13431466619173685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,32,power_law_1.01,0.1309440016746521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,16,balanced,0.14853866895039877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,16,balanced,0.1732800006866455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,16,balanced,0.19524266322453818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,16,balanced,0.24034667015075684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,16,balanced,0.2935146689414978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,16,balanced,0.3963786760965983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,16,balanced,0.48603200912475586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,16,balanced,0.6031359831492106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,32,power_law_1.01,0.1486143946647644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,16,balanced,0.8616053263346354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,32,power_law_1.01,0.07329919934272766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,16,balanced,1.634112040201823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,32,power_law_1.01,0.1835904002189636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,32,power_law_1.01,0.07548159956932068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,32,power_law_1.01,0.2342207908630371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,32,power_law_1.01,0.30335359573364257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,32,power_law_1.01,0.07923200130462646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,32,power_law_1.01,0.4069568157196045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,16,balanced,0.04407466451327006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,16,balanced,0.04517333209514618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,16,balanced,0.04373333354791006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,16,balanced,0.04603200157483419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,16,balanced,0.04978133241335551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,16,balanced,0.05020800232887268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,16,balanced,0.05209066470464071
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,16,balanced,0.05233600238958994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,16,balanced,0.052154665191968284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,32,power_law_1.01,0.07950720191001892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,16,balanced,0.05193600058555603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,16,balanced,0.053904001911481224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,16,balanced,0.05403199791908264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,16,balanced,0.05194133520126343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,16,balanced,0.05589333176612854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,16,balanced,0.058117335041364036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,32,power_law_1.01,0.6313983917236328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,16,balanced,0.060047999024391174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,16,balanced,0.060047999024391174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,16,balanced,0.06841066479682922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,16,balanced,0.06676800052324931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,16,balanced,0.08853333195050557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,16,balanced,0.09902399778366089
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,16,balanced,0.12339733044306438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,16,balanced,0.1381439963976542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,16,balanced,0.18259199460347494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,16,balanced,0.20754132668177286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,16,balanced,0.2892640034357707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,16,balanced,0.34855465094248456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,16,balanced,0.41470932960510254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,16,balanced,0.6231946547826132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,32,power_law_1.01,0.0869376003742218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,32,power_law_1.01,0.7361279964447022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,16,balanced,1.1852693557739258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,32,power_law_1.01,0.08968960046768189
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,32,power_law_1.01,0.9158847808837891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,32,power_law_1.01,0.09747200012207032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,32,power_law_1.01,1.3722623825073241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,32,power_law_1.01,0.11296639442443848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,1,power_law_1.01,0.05422080159187317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,32,power_law_1.01,0.12476799488067628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,1,power_law_1.01,0.0862272024154663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,32,power_law_1.01,2.8093568801879885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,1,power_law_1.01,0.09843840003013611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,32,power_law_1.01,0.15446399450302123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,1,power_law_1.01,0.1229632019996643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,32,power_law_1.01,0.18736640214920045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,1,power_law_1.01,0.20052480697631836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,1,power_law_1.2,0.07390080094337463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,1,power_law_1.01,0.26920959949493406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,32,power_law_1.01,0.23296639919281006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,1,power_law_1.2,0.08354560136795045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,1,power_law_1.01,0.3212671995162964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,1,power_law_1.2,0.11358079910278321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,1,power_law_1.2,0.13717119693756102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,32,power_law_1.01,0.26133759021759034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,1,power_law_1.01,0.42401919364929197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,1,power_law_1.2,0.18172800540924072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,1,power_law_1.2,0.3056704044342041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,1,power_law_1.01,0.4405183792114258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,1,power_law_1.2,0.33717119693756104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,32,power_law_1.01,0.33203840255737305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,1,power_law_1.2,0.3529344081878662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,1,power_law_1.01,0.4612095832824707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,1,power_law_1.2,0.3644864082336426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,1,power_law_1.2,0.36978559494018554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,1,power_law_1.01,0.48108158111572263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,1,power_law_1.2,0.38673279285430906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,32,power_law_1.01,0.39650559425354004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,1,power_law_1.2,0.3978879928588867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,4,power_law_1.2,0.04910080134868622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,1,power_law_1.01,0.505779218673706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,1,power_law_1.2,0.40500478744506835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,1,power_law_1.2,0.42931838035583497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,4,power_law_1.2,0.047251200675964354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,1,power_law_1.01,0.5257728099822998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,1,power_law_1.2,0.44323201179504396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,32,power_law_1.01,0.5550784111022949
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,1,power_law_1.2,0.44600958824157716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,4,power_law_1.2,0.053011202812194826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,1,power_law_1.01,0.54552321434021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,1,power_law_1.2,0.4795392036437988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,4,power_law_1.2,0.05856000185012818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,1,power_law_1.2,0.5386559963226318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,1,power_law_1.01,0.5741824150085449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,1,power_law_1.2,0.6048128128051757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,32,power_law_1.01,0.709830379486084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,4,power_law_1.2,0.06403200030326843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,1,power_law_1.01,0.589299201965332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,1,power_law_1.2,0.7335999965667724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,4,power_law_1.2,0.08136320114135742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,1,power_law_1.2,0.7872960090637207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,1,power_law_1.01,0.6130047798156738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,4,power_law_1.2,0.08410239815711976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,1,power_law_1.2,0.9927295684814453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,1,power_law_1.2,1.1790080070495605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,1,power_law_1.01,0.6788159847259522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,4,power_law_1.2,0.08856319785118102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,2,32,power_law_1.01,0.9081791877746582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,4,power_law_1.2,0.09040639996528625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,1,power_law_1.2,1.5468031883239746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,1,power_law_1.01,0.7422463893890381
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,4,power_law_1.2,0.09216639995574952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,1,power_law_1.2,1.8945152282714843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,1,power_law_1.01,0.8461376190185547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,4,power_law_1.2,0.09575679898262024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,1,power_law_1.2,2.6184640884399415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,4,power_law_1.2,0.10084480047225952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,1,power_law_1.01,0.918943977355957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,2,32,power_law_1.01,1.36364803314209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,4,power_law_1.2,0.10014079809188843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,1,power_law_1.2,3.3969406127929687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,1,power_law_1.01,1.1224767684936523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,4,power_law_1.2,0.10558719635009765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,4,power_law_1.2,0.11192320585250855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,1,power_law_1.01,1.3791423797607423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,1,power_law_1.2,4.104166412353516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,4,power_law_1.2,0.11416319608688355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,4,power_law_1.2,0.1243008017539978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,1,power_law_1.01,1.7696704864501953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,4,power_law_1.2,0.14402559995651246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,1,power_law_1.2,6.288972854614258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,1,power_law_1.01,2.0559360504150392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,4,power_law_1.2,0.16108160018920897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,4,power_law_1.2,0.20935680866241455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,2,32,power_law_1.01,2.6942207336425783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,1,power_law_1.01,2.710016059875488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,4,power_law_1.2,0.23733758926391602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,4,power_law_1.2,0.31535360813140867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,4,power_law_1.2,0.4051392078399658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,1,power_law_1.2,11.985676574707032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,1,power_law_1.01,3.4620033264160157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,4,power_law_1.2,0.5301568031311035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,4,power_law_1.2,0.6845056056976319
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,1,power_law_1.01,4.196038436889649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,4,power_law_1.2,0.9945280075073242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,4,power_law_1.2,1.3103296279907226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,1,power_law_1.01,6.146156692504883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,4,power_law_1.2,1.7309823989868165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,4,power_law_1.2,2.6394943237304687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,1,power_law_1.01,0.05185920000076294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,1,power_law_1.01,0.07330560088157653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,1,power_law_1.01,0.08208640217781067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,1,power_law_1.01,11.490118408203125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,32,power_law_1.01,0.08790400028228759
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,1,power_law_1.01,0.1273792028427124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,32,power_law_1.01,0.055078399181365964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,4,power_law_1.2,5.512960052490234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,32,power_law_1.01,0.04906240105628967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,1,power_law_1.01,0.1788800001144409
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,32,power_law_1.01,0.053651201725006106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,1,power_law_1.01,0.2337023973464966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,32,power_law_1.01,0.050220799446105954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,32,power_law_1.01,0.057760000228881836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,1,power_law_1.01,0.3197439908981323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,32,power_law_1.01,0.06007680296897888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,1,power_law_1.01,0.34559359550476076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,32,power_law_1.01,0.06301440000534057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,32,power_law_1.01,0.06047359704971313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,1,power_law_1.01,0.35923199653625487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,32,power_law_1.01,0.0611519992351532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,1,power_law_1.01,0.3665472030639648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,32,power_law_1.01,0.06356480121612548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,32,power_law_1.01,0.06420480012893677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,1,power_law_1.01,0.3857088088989258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,32,power_law_1.01,0.0644864022731781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,1,power_law_1.01,0.3917311906814575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,32,power_law_1.01,0.06999679803848266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,32,power_law_1.01,0.07582079768180847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,1,power_law_1.01,0.41220479011535643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,32,power_law_1.01,0.07704960107803345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,1,power_law_1.01,0.4234943866729736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,32,power_law_1.01,0.07975040078163147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,4,balanced,0.045941332976023354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,4,balanced,0.048063998421033226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,32,power_law_1.01,0.08744320273399353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,4,balanced,0.050383999943733215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,1,power_law_1.01,0.4309247970581055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,4,balanced,0.07220800220966339
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,4,balanced,0.10198932886123657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,4,balanced,0.1525706648826599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,4,balanced,0.20090667406717935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,4,balanced,0.20335467656453451
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,4,balanced,0.20414932568868002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,4,balanced,0.2083146572113037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,4,balanced,0.2063466707865397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,4,balanced,0.20958399772644043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,32,power_law_1.01,0.09039999842643738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,4,balanced,0.2135253349939982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,4,balanced,0.21497599283854166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,4,balanced,0.2216213345527649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,4,balanced,0.22643733024597168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,1,power_law_1.01,0.45754241943359375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,4,balanced,0.2362826665242513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,4,balanced,0.24958399931589761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,4,balanced,0.26551999648412067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,4,balanced,0.2894773284594218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,32,power_law_1.01,0.10694400072097779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,4,balanced,0.31921066840489704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,4,balanced,0.3699626525243123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,4,balanced,0.4336533149083455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,32,power_law_1.01,0.12067840099334717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,4,balanced,0.5421919822692871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,4,balanced,0.673093318939209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,1,power_law_1.01,0.4869120121002197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,32,power_law_1.01,0.12080639600753784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,4,balanced,0.9100266297658285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,4,balanced,1.1293066342671711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,32,power_law_1.01,0.12003840208053589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,4,balanced,1.4211999575297039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,32,power_law_1.01,0.14517120122909546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,1,power_law_1.01,0.5372543811798096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,4,balanced,2.089306672414144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,32,power_law_1.01,0.18031359910964967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,32,power_law_1.01,0.08440960049629212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,4,balanced,3.99289608001709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,1,power_law_1.01,0.576204776763916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,32,power_law_1.01,0.22480640411376954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,32,power_law_1.01,0.08551679849624634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,1,power_law_1.01,0.6728320121765137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,32,power_law_1.01,0.2867455959320068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,32,power_law_1.01,0.1022528052330017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,1,power_law_1.01,0.7649407863616944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,32,power_law_1.01,0.4206655979156494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,32,power_law_1.01,0.11257599592208863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,1,power_law_1.01,0.9591487884521485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,32,power_law_1.01,0.11717120409011841
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,32,power_law_1.01,0.4921728134155273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,1,power_law_1.01,1.0963583946228028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,32,power_law_1.01,0.7214399814605713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,32,power_law_1.01,0.11503360271453858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,1,power_law_1.01,1.4194496154785157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,32,power_law_1.01,0.11193599700927734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,32,power_law_1.01,0.9750016212463379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,1,power_law_1.01,1.750752067565918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,32,power_law_1.01,0.1174015998840332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,32,power_law_1.01,0.12044800519943237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,1,power_law_1.01,2.399577522277832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,32,power_law_1.01,1.9859455108642579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,32,power_law_1.01,0.12783360481262207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,32,power_law_1.01,0.1282688021659851
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,1,power_law_1.01,3.0403135299682615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,4,power_law_1.01,0.05496320128440857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,32,power_law_1.01,0.12620160579681397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,4,power_law_1.01,0.06967039704322815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,32,power_law_1.01,0.13768960237503053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,1,power_law_1.01,3.6836288452148436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,4,power_law_1.01,0.06259840130805969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,32,power_law_1.01,0.14392319917678834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,4,power_law_1.01,0.06666880249977111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,32,power_law_1.01,0.13994239568710326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,4,power_law_1.01,0.08152959942817688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,32,power_law_1.01,0.16001919507980347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,1,power_law_1.01,5.523852920532226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,4,power_law_1.01,0.09828479886054993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,32,power_law_1.01,0.17671040296554566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,4,power_law_1.01,0.11825920343399048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,32,power_law_1.01,0.22358400821685792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,4,power_law_1.01,0.11630719900131226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,32,power_law_1.01,0.22777600288391114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,4,power_law_1.01,0.11959680318832397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,32,power_law_1.01,0.3140032052993774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,4,power_law_1.01,0.12168320417404174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,4,power_law_1.01,0.1253376007080078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,32,power_law_1.01,0.3201855897903442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,4,power_law_1.01,0.03284479975700379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,1,power_law_1.01,10.510835266113281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,4,power_law_1.01,0.13062399625778198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,4,power_law_1.01,0.031091201305389404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,32,power_law_1.01,0.39765119552612305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,4,power_law_1.01,0.032307198643684386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,4,power_law_1.01,0.13212159872055054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,4,power_law_1.01,0.03442560136318207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,32,power_law_1.01,0.4884031772613525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,4,power_law_1.01,0.1386623978614807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,4,power_law_1.01,0.03558399975299835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,4,power_law_1.01,0.040140798687934874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,4,power_law_1.01,0.15173120498657228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,32,power_law_1.01,0.6900864124298096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,4,power_law_1.01,0.04190720021724701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,4,power_law_1.01,0.15839999914169312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,4,power_law_1.01,0.04234879910945892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,4,power_law_1.01,0.043161600828170776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,32,power_law_1.01,0.9846783638000488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,4,power_law_1.01,0.15946880578994752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,4,power_law_1.01,0.043910399079322815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,4,power_law_1.01,0.04689280092716217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,4,power_law_1.01,0.1762112021446228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,4,power_law_1.01,0.04973439872264862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,32,power_law_1.01,1.1021247863769532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,4,power_law_1.01,0.19744640588760376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,4,power_law_1.01,0.05073919892311096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,4,power_law_1.01,0.05198079943656921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,4,power_law_1.01,0.23105919361114502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,4,power_law_1.01,0.05596799850463867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,4,power_law_1.01,0.05858560204505921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,32,power_law_1.01,2.0455360412597656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,4,power_law_1.01,0.26153600215911865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,4,power_law_1.01,0.06492159962654113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,4,power_law_1.01,0.0748799979686737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,4,power_law_1.01,0.33231360912323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,4,power_law_1.01,0.08479359745979309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,4,power_law_1.01,0.10744960308074951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,4,power_law_1.01,0.39409921169281004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,4,power_law_1.01,0.13207679986953735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,4,power_law_1.01,0.523199987411499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,4,power_law_1.01,0.17052160501480101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,32,power_law_1.01,3.547520065307617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,4,power_law_1.01,0.22687358856201173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,1,power_law_1.01,0.1458176016807556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,4,power_law_1.01,0.6558720111846924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,4,power_law_1.01,0.3192768096923828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,1,power_law_1.01,0.3107392072677612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,4,power_law_1.01,0.39626240730285645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,4,power_law_1.01,0.9331583976745605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,4,power_law_1.01,0.6
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,1,power_law_1.01,0.4364352226257324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,4,power_law_1.01,1.2180607795715332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,4,power_law_1.01,0.7469567775726318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,1,power_law_1.01,0.5891327857971191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,4,power_law_1.01,0.9342144012451172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,4,power_law_1.01,1.5356608390808106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,1,power_law_1.01,1.0982912063598633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,4,power_law_1.01,1.5409600257873535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,1,power_law_1.01,1.805036735534668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,4,power_law_1.01,2.2871103286743164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,1,power_law_1.01,2.080614471435547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,4,power_law_1.01,2.84899845123291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,1,power_law_1.01,2.806163215637207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,4,power_law_1.01,4.54961929321289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,1,power_law_1.01,3.028620719909668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,8,power_law_1.01,0.10748159885406494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,1,power_law_1.01,3.1662271499633787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,1,power_law_1.01,3.258015823364258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,8,power_law_1.01,0.1498944044113159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,1,power_law_1.01,3.3758399963378904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,8,power_law_1.01,0.08371840119361877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,1,power_law_1.01,3.5915904998779298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,8,balanced,0.045824001232783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,8,balanced,0.04381866753101349
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,8,balanced,0.0476693312327067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,8,balanced,0.05272533496220907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,8,balanced,0.0784693310658137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,8,balanced,0.08442667126655579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,8,balanced,0.08637866377830505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,8,balanced,0.0867199997107188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,8,balanced,0.08721599976221721
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,8,balanced,0.08913600444793701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,8,balanced,0.08924800157546997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,8,balanced,0.0909546713034312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,8,balanced,0.09126399954160054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,8,balanced,0.09513599673906963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,8,balanced,0.10099732875823975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,8,balanced,0.10406933228174846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,8,balanced,0.10646399855613708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,8,balanced,0.11652800440788269
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,8,balanced,0.12589866916338602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,8,balanced,0.15050133069356283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,8,balanced,0.1671253244082133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,8,balanced,0.2215893268585205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,16,balanced,0.043824002146720886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,8,balanced,0.25196266174316406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,1,power_law_1.01,3.5639873504638673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,8,power_law_1.01,0.1002303957939148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,16,balanced,0.04191466669241587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,8,balanced,0.37273601690928143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,16,balanced,0.04156800111134847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,16,balanced,0.04377600053946177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,16,balanced,0.04780800143877665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,8,balanced,0.4169333378473918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,16,balanced,0.049914668003718056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,16,balanced,0.05386666456858317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,8,balanced,0.5768266518910726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,16,balanced,0.05440000196297964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,16,balanced,0.053957333167394005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,16,balanced,0.05387733379999796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,16,balanced,0.05588266750176748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,8,balanced,0.7514506975809733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,16,balanced,0.056405335664749146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,16,balanced,0.056287998954455055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,16,balanced,0.060090666015942894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,16,balanced,0.062261333068211876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,8,balanced,0.895477294921875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,16,balanced,0.06619733572006226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,16,balanced,0.06830400228500366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,16,balanced,0.0763733337322871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,16,balanced,0.08288533488909404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,16,balanced,0.10302933057149251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,8,balanced,1.3917120297749836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,16,balanced,0.13690666357676187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,16,balanced,0.1751520037651062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,16,balanced,0.20382932821909586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,16,balanced,0.2797866662343343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,16,balanced,0.3495359818140666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,8,balanced,2.672250747680664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,16,balanced,0.5001173416773478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,16,balanced,0.6401866674423218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,16,balanced,0.7899839878082275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,1,power_law_1.01,3.8578369140625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,16,balanced,1.2064800262451172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,8,power_law_1.01,0.11917439699172974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,16,balanced,2.380256017049154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,1,power_law_1.01,3.8246463775634765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,8,power_law_1.01,0.15649280548095704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,1,power_law_1.01,3.869036865234375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,8,power_law_1.01,0.16435199975967407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,1,power_law_1.01,4.254796981811523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,8,power_law_1.01,0.1715648055076599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,1,power_law_1.01,4.478112030029297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,8,power_law_1.01,0.1759168028831482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,1,power_law_1.01,5.020927810668946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,8,power_law_1.01,0.17317119836807252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,1,power_law_1.2,0.04615679979324341
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,1,power_law_1.01,5.199513626098633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,4,power_law_1.01,0.07414399981498718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,1,power_law_1.2,0.06754559874534607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,1,power_law_1.2,0.07617920041084289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,8,power_law_1.01,0.17905919551849364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,1,power_law_1.2,0.10568959712982177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,4,power_law_1.01,0.08813440203666686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,1,power_law_1.2,0.152838397026062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,1,power_law_1.01,5.834617614746094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,1,power_law_1.2,0.28010880947113037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,1,power_law_1.2,0.31486079692840574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,8,power_law_1.01,0.19118080139160157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,4,power_law_1.01,0.06421120166778564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,1,power_law_1.2,0.3309376001358032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,1,power_law_1.2,0.32565760612487793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,1,power_law_1.2,0.35961599349975587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,1,power_law_1.01,5.548230361938477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,4,power_law_1.01,0.08152959942817688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,1,power_law_1.2,0.3572864055633545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,8,power_law_1.01,0.19088000059127808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,1,power_law_1.2,0.36563839912414553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,1,power_law_1.2,0.37531518936157227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,4,power_law_1.01,0.09910399913787842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,1,power_law_1.2,0.39232640266418456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,1,power_law_1.2,0.40899200439453126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,1,power_law_1.01,6.537241363525391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,8,power_law_1.01,0.19407999515533447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,1,power_law_1.2,0.4216320037841797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,4,power_law_1.01,0.1293951988220215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,1,power_law_1.2,0.4335360050201416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,1,power_law_1.2,0.49239678382873536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,8,power_law_1.01,0.20384640693664552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,4,power_law_1.01,0.13477120399475098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,1,power_law_1.2,0.478656005859375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,1,power_law_1.2,0.5498879909515381
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,1,power_law_1.01,7.725772857666016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,1,power_law_1.2,0.6274432182312012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,4,power_law_1.01,0.1372032046318054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,1,power_law_1.2,0.7431424140930176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,8,power_law_1.01,0.21071360111236573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,1,power_law_1.2,0.7719935894012451
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,4,power_law_1.01,0.14190720319747924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,1,power_law_1.2,0.9243712425231934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,8,power_law_1.01,0.22071681022644044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,1,power_law_1.01,9.559442901611328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,1,power_law_1.2,1.0466624259948731
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,4,power_law_1.01,0.14494719505310058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,1,power_law_1.2,1.3652607917785644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,4,power_law_1.01,0.14872959852218628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,8,power_law_1.01,0.24209918975830078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,1,power_law_1.2,1.6859840393066405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,4,power_law_1.01,0.15276800394058226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,1,1,power_law_1.2,2.031315231323242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,1,power_law_1.01,11.186713409423827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,8,power_law_1.01,0.2764287948608398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,4,power_law_1.01,0.1587007999420166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,1,1,power_law_1.2,3.039910316467285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,4,power_law_1.01,0.16725120544433594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,8,power_law_1.01,0.3344575881958008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,4,power_law_1.01,0.1752128005027771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,1,1,power_law_1.2,5.667577743530273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,1,power_law_1.01,13.7278076171875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,8,power_law_1.01,0.3807231903076172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,4,power_law_1.01,0.1791424036026001
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,4,power_law_1.01,0.19063040018081664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,8,power_law_1.01,0.48818559646606446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,4,power_law_1.01,0.22136321067810058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,1,power_law_1.01,16.745082092285156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,8,power_law_1.01,0.5697728157043457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,4,power_law_1.01,0.23972480297088622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,4,power_law_1.01,0.2939136028289795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,8,power_law_1.01,0.7246208190917969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,4,power_law_1.01,0.36471679210662844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,8,power_law_1.01,0.9041919708251953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,4,power_law_1.01,0.4723008155822754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,4,power_law_1.01,0.18385920524597169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,4,power_law_1.01,0.5415552139282227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,4,power_law_1.01,0.3179647922515869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,8,power_law_1.01,1.25982723236084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,1,power_law_1.01,31.619430541992188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,4,power_law_1.01,0.7217152118682861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,4,power_law_1.01,0.1399616003036499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,4,power_law_1.01,0.26410880088806155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,4,power_law_1.01,0.8827712059020996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,8,power_law_1.01,1.6708799362182618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,4,power_law_1.01,0.38235518932342527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,4,power_law_1.01,0.5976384162902832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,4,power_law_1.01,1.1959360122680665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,1,8,power_law_1.01,1.9675392150878905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,4,power_law_1.01,0.6677504062652588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,4,power_law_1.01,0.6556863784790039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,4,power_law_1.01,1.5940863609313964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,4,power_law_1.01,0.7084671974182128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,64,power_law_1.2,0.034969601035118106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,64,power_law_1.2,0.032076799869537355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,4,power_law_1.01,0.7099520206451416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,8,4,power_law_1.01,1.9162431716918946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,1,8,power_law_1.01,3.186124801635742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,64,power_law_1.2,0.03301759958267212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,64,power_law_1.2,0.035392001271247864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,64,power_law_1.2,0.03369599878787995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,4,power_law_1.01,0.7403007984161377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,64,power_law_1.2,0.03330560028553009
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,64,power_law_1.2,0.03323520123958588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,64,power_law_1.2,0.03399679958820343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,4,power_law_1.01,0.746995210647583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,64,power_law_1.2,0.03400959968566895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,64,power_law_1.2,0.03452160060405731
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,64,power_law_1.2,0.03501439988613129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,8,4,power_law_1.01,2.862278366088867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,4,power_law_1.01,0.7809152126312255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,64,power_law_1.2,0.03576320111751556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,64,power_law_1.2,0.036422398686408994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,64,power_law_1.2,0.036524799466133115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,4,power_law_1.01,0.7700160026550293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,64,power_law_1.2,0.03956480026245117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,64,power_law_1.2,0.04078719913959503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,64,power_law_1.2,0.040659201145172116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,4,power_law_1.01,0.8152447700500488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,64,power_law_1.2,0.04402559995651245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,1,8,power_law_1.01,6.2289470672607425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,64,power_law_1.2,0.04542720019817352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,64,power_law_1.2,0.05201280117034912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,4,power_law_1.01,0.8339648246765137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,64,power_law_1.2,0.05616000294685364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,64,power_law_1.2,0.06937599778175355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,4,power_law_1.01,0.8901247978210449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,64,power_law_1.2,0.09344000220298768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,64,power_law_1.2,0.13662079572677613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,8,4,power_law_1.01,5.797919845581054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,64,power_law_1.2,0.14872959852218628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,4,power_law_1.01,1.0119296073913575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,64,power_law_1.2,0.22867200374603272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,64,power_law_1.2,0.32280960083007815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,4,power_law_1.01,0.9839872360229492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,64,power_law_1.2,0.44437761306762696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,4,power_law_1.01,1.156383991241455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,64,power_law_1.2,0.5837376117706299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,4,power_law_1.01,1.232032012939453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,64,power_law_1.2,1.2903807640075684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,4,power_law_1.01,1.4572799682617188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,1,power_law_1.2,0.09498879909515381
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,1,power_law_1.2,0.09660800099372864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,4,power_law_1.01,1.574015998840332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,4,32,power_law_1.2,0.03832319974899292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,4,32,power_law_1.2,0.033344000577926636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,1,power_law_1.2,0.11191040277481079
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,4,32,power_law_1.2,0.033215999603271484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,4,power_law_1.01,1.8739583969116211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,4,32,power_law_1.2,0.03216640055179596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,4,32,power_law_1.2,0.03272959887981415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,4,32,power_law_1.2,0.03429119884967804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,4,32,power_law_1.2,0.03430399894714355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,4,32,power_law_1.2,0.03454720079898834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,4,power_law_1.01,2.428268814086914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,1,power_law_1.2,0.12694400548934937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,4,32,power_law_1.2,0.034668800234794614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,4,32,power_law_1.2,0.03518719971179962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,4,32,power_law_1.2,0.03513599932193756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,4,32,power_law_1.2,0.036723199486732486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,4,32,power_law_1.2,0.037215998768806456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,1,power_law_1.2,0.15936000347137452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,4,4,balanced,0.039818666875362396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,4,4,balanced,0.039962666730086006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,4,4,balanced,0.039919999738534294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,4,4,balanced,0.03886399914820989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,4,4,balanced,0.041706666350364685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,4,4,balanced,0.0403413325548172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,4,4,balanced,0.04033066580692927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,4,32,power_law_1.2,0.03749760091304779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,4,4,balanced,0.040106666584809623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,4,4,balanced,0.04197866717974345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,4,4,balanced,0.04016000032424927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,4,power_law_1.01,3.1969791412353517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,4,4,balanced,0.03994133323431015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,4,4,balanced,0.04211199780305227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,4,4,balanced,0.041834667325019836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,4,4,balanced,0.04419200122356415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,4,4,balanced,0.04771199822425842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,4,4,balanced,0.0483893354733785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,4,4,balanced,0.05374933282534281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,4,32,power_law_1.2,0.0403328001499176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,4,4,balanced,0.06206933160622915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,4,4,balanced,0.06427733103434245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,4,4,balanced,0.07698133091131847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,4,4,balanced,0.08612266182899475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,4,4,balanced,0.10332266489664714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,4,4,balanced,0.1220746636390686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,4,4,balanced,0.15563199917475382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,4,32,power_law_1.2,0.04178560078144074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,4,4,balanced,0.1848693291346232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,4,4,balanced,0.2518293261528015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,4,4,balanced,0.3211680054664612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,4,4,balanced,0.38791465759277344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,4,4,balanced,0.5807786782582601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,4,32,power_law_1.2,0.04289279878139496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,4,4,balanced,1.0986986955006917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,1,power_law_1.2,0.19185919761657716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,4,32,power_law_1.2,0.048921599984169006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,4,32,power_law_1.2,0.04948480129241943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,4,32,power_law_1.2,0.057183998823165896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,4,power_law_1.01,3.4972160339355467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,4,32,power_law_1.2,0.06112639904022217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,1,power_law_1.2,0.20952320098876953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,4,32,power_law_1.2,0.06938880085945129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,4,32,power_law_1.2,0.09054080247879029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,4,32,power_law_1.2,0.14246400594711303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,1,power_law_1.2,0.26058239936828614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,4,32,power_law_1.2,0.1528447985649109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,4,32,power_law_1.2,0.23143041133880615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,1,4,power_law_1.01,4.332121658325195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,4,32,power_law_1.2,0.312608003616333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,1,power_law_1.2,0.2811072111129761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,4,32,power_law_1.2,0.4683135986328125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,1,power_law_1.2,0.28832640647888186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,4,32,power_law_1.2,0.7984447956085206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,1,power_law_1.2,0.3008064031600952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,1,16,power_law_1.01,0.02892799973487854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,1,4,power_law_1.01,6.77215347290039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,1,16,power_law_1.01,0.027270400524139406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,4,32,power_law_1.2,1.302393627166748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,1,16,power_law_1.01,0.026214399933815004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,1,16,power_law_1.01,0.026604801416397095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,1,power_law_1.2,0.3101248025894165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,1,16,power_law_1.01,0.029836800694465638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,1,16,power_law_1.01,0.031744000315666196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,1,16,power_law_1.01,0.02963840067386627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,1,power_law_1.2,0.31626880168914795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,1,16,power_law_1.01,0.035411199927330016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,1,16,power_law_1.01,0.034796801209449765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,1,16,power_law_1.01,0.03640959858894348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,1,16,power_law_1.01,0.038150399923324585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,1,power_law_1.2,0.33496320247650146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,1,16,power_law_1.01,0.039238399267196654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,1,16,power_law_1.01,0.039955198764801025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,1,16,power_law_1.01,0.039776000380516055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,1,power_law_1.2,0.3511296033859253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,1,16,power_law_1.01,0.04535039961338043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,1,16,power_law_1.01,0.04661760032176972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,1,16,power_law_1.01,0.04602240025997162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,1,power_law_1.2,0.3608319997787476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,1,16,power_law_1.01,0.052019202709198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,1,4,power_law_1.01,12.614559936523438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,16,power_law_1.01,0.05801600217819214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,16,power_law_1.01,0.06268799901008607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,1,power_law_1.2,0.38837759494781493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,16,power_law_1.01,0.07325440049171447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,16,power_law_1.01,0.08323839902877808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,8,1,power_law_1.2,0.0345984011888504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,8,1,power_law_1.2,0.03988479971885681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,8,1,power_law_1.2,0.037920001149177554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,16,power_law_1.01,0.09595519900321961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,8,1,power_law_1.2,0.057171201705932616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,1,power_law_1.2,0.4449151992797852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,8,1,power_law_1.2,0.05807999968528747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,8,1,power_law_1.2,0.05917440056800842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,8,1,power_law_1.2,0.05864959955215454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,16,power_law_1.01,0.12558720111846924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,8,1,power_law_1.2,0.059334397315979004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,8,1,power_law_1.2,0.059487998485565186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,8,1,power_law_1.2,0.060601598024368285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,8,1,power_law_1.2,0.06292480230331421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,16,power_law_1.01,0.1642240047454834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,8,1,power_law_1.2,0.06636800169944763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,8,1,power_law_1.2,0.06801279783248901
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,8,1,power_law_1.2,0.07132800221443177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,1,power_law_1.2,0.4867136001586914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,8,1,power_law_1.2,0.0803264021873474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,8,1,power_law_1.2,0.08323839902877808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,16,power_law_1.01,0.20698881149291992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,8,1,power_law_1.2,0.08855040073394775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,8,1,power_law_1.2,0.1063040018081665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,8,1,power_law_1.2,0.11301759481430054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,8,1,power_law_1.2,0.14141440391540527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,8,1,power_law_1.2,0.16699520349502564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,16,power_law_1.01,0.2657088041305542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,8,1,power_law_1.2,0.22149760723114015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,8,1,power_law_1.2,0.26431999206542967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,1,power_law_1.2,0.5928063869476319
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,8,1,power_law_1.2,0.3662976026535034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,8,1,power_law_1.2,0.4618175983428955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,16,power_law_1.01,0.32807040214538574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,8,1,power_law_1.2,0.6571712017059326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,8,1,power_law_1.2,0.8515263557434082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,1,power_law_1.2,0.6698495864868164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,8,1,power_law_1.2,1.039027214050293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,16,power_law_1.01,0.48935680389404296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,8,1,power_law_1.2,1.6095872879028321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,1,power_law_1.2,0.864134407043457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,8,1,power_law_1.2,3.1395135879516602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,1,balanced,0.0727946658929189
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,1,balanced,0.12475200494130452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,16,power_law_1.01,0.9451968193054199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,1,balanced,0.15227199594179788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,1,balanced,0.25521600246429443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,1,balanced,0.47095998128255206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,1,balanced,0.5756800174713135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,1,balanced,0.5793600082397461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,1,balanced,0.5824906826019287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,1,balanced,0.5864746570587158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,1,balanced,0.5899519920349121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,1,balanced,0.5984373490015665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,1,power_law_1.2,1.0528512001037598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,1,balanced,0.6002613306045532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,1,balanced,0.6077013413111368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,1,balanced,0.6132266521453857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,1,balanced,0.6269280115763346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,1,balanced,0.6355093320210775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,1,balanced,0.655898650487264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,1,balanced,0.7247839768727621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,1,balanced,0.7282826900482178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,1,balanced,0.8909760316212972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,1,balanced,0.879962682723999
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,1,balanced,1.3039733568827312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,1,balanced,1.2190133730570476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,1,power_law_1.2,1.4155712127685547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,1,balanced,1.788373311360677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,1,balanced,2.0299253463745117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,1,balanced,3.3130880991617837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,1,balanced,3.80622927347819
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,1,power_law_1.2,1.7783231735229492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,1,balanced,4.422117233276367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,1,balanced,7.041877110799153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,1,power_law_1.2,2.511961555480957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,1,balanced,13.075108846028646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,1,power_law_1.2,3.248953628540039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,16,balanced,0.054229333996772766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,16,balanced,0.05602133274078369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,16,balanced,0.054005334774653115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,16,balanced,0.06154133379459381
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,16,balanced,0.08447466293970744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,16,balanced,0.11576533317565918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,16,balanced,0.14686933159828186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,16,balanced,0.14761066436767578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,16,balanced,0.14748266339302063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,16,balanced,0.14818132917086282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,16,balanced,0.14893333117167154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,16,balanced,0.15280532836914062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,16,balanced,0.1532853345076243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,16,balanced,0.15402133266131082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,16,balanced,0.15997866789499918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,16,balanced,0.16030399998029074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,16,balanced,0.1662666698296865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,16,balanced,0.17594132820765176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,16,balanced,0.18201599518458048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,16,balanced,0.2002026637395223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,16,balanced,0.21733866135279337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,16,balanced,0.25406932830810547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,64,power_law_1.2,0.06349440217018128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,16,balanced,0.28809066613515216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,16,balanced,0.357749342918396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,16,balanced,0.4400800069173177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,1,power_law_1.2,3.983084869384766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,16,balanced,0.5748266776402792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,16,balanced,0.7405760288238525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,16,balanced,0.915066639582316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,16,balanced,1.3253653049468994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,64,power_law_1.2,0.05602560043334961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,16,balanced,2.54528538386027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,64,power_law_1.2,0.05315200090408325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,64,power_law_1.2,0.05191680192947388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,1,power_law_1.2,6.1847679138183596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,64,power_law_1.2,0.05263360142707825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,64,power_law_1.2,0.05294079780578613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,64,power_law_1.2,0.05321599841117859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,64,power_law_1.2,0.0542527973651886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,16,power_law_1.01,0.08584319949150085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,1,power_law_1.2,12.00719985961914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,64,power_law_1.2,0.05418879985809326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,16,power_law_1.01,0.11571840047836304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,64,power_law_1.2,0.05491840243339539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,64,power_law_1.2,0.05537279844284058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,16,power_law_1.01,0.05694720149040222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,1,power_law_1.01,0.07464960217475891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,64,power_law_1.2,0.05790719985961914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,1,power_law_1.01,0.10037120580673217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,16,power_law_1.01,0.06575999855995178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,64,power_law_1.2,0.05808640122413635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,1,power_law_1.01,0.1487552046775818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,16,power_law_1.01,0.07861120104789734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,64,power_law_1.2,0.05959039926528931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,1,power_law_1.01,0.18627840280532837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,64,power_law_1.2,0.06481279730796814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,16,power_law_1.01,0.09221760034561158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,1,power_law_1.01,0.3334719896316528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,64,power_law_1.2,0.06650239825248719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,16,power_law_1.01,0.09664639830589294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,1,power_law_1.01,0.4794816017150879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,64,power_law_1.2,0.07170559763908387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,16,power_law_1.01,0.09850239753723145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,1,power_law_1.01,0.5995007991790772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,64,power_law_1.2,0.08076159954071045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,32,8,power_law_1.01,0.0537280023097992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,1,power_law_1.01,0.7615295886993408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,16,power_law_1.01,0.10163840055465698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,64,power_law_1.2,0.09113600254058837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,1,power_law_1.01,0.8251328468322754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,64,power_law_1.2,0.11117440462112427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,16,power_law_1.01,0.10179200172424316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,32,8,power_law_1.01,0.05156480073928833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,1,power_law_1.01,0.8583871841430664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,64,power_law_1.2,0.1300480008125305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,16,power_law_1.01,0.10236159563064576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,32,8,power_law_1.01,0.04990079998970032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,1,power_law_1.01,0.9059776306152344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,64,power_law_1.2,0.17345279455184937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,16,power_law_1.01,0.10511360168457032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,32,8,power_law_1.01,0.04796159863471985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,1,power_law_1.01,0.9148287773132324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,64,power_law_1.2,0.19638400077819823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,16,power_law_1.01,0.1094655990600586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,32,8,power_law_1.01,0.05416319966316223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,1,power_law_1.01,0.9339520454406738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,64,power_law_1.2,0.2671871900558472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,16,power_law_1.01,0.1094208002090454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,1,power_law_1.01,1.004531192779541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,32,8,power_law_1.01,0.05725439786911011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,64,power_law_1.2,0.35016961097717286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,16,power_law_1.01,0.11920640468597413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,32,8,power_law_1.01,0.06103039979934692
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,1,power_law_1.01,1.030239963531494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,64,power_law_1.2,0.5280831813812256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,16,power_law_1.01,0.12109440565109253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,32,8,power_law_1.01,0.06104320287704468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,1,power_law_1.01,1.0443391799926758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,32,8,power_law_1.01,0.06280320286750793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,64,power_law_1.2,0.6433856010437011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,16,power_law_1.01,0.12753280401229858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,1,power_law_1.01,1.1048768043518067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,32,8,power_law_1.01,0.0644864022731781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,16,power_law_1.01,0.1460479974746704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,1,power_law_1.01,1.2160832405090332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,64,power_law_1.2,0.8634176254272461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,32,8,power_law_1.01,0.06650239825248719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,1,power_law_1.01,1.2752127647399902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,16,power_law_1.01,0.1579519987106323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,32,8,power_law_1.01,0.0668287992477417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,64,power_law_1.2,1.3629311561584472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,1,power_law_1.01,1.43951358795166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,16,power_law_1.01,0.19484800100326538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,32,8,power_law_1.01,0.06837120056152343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,1,power_law_1.01,1.511244773864746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,32,8,power_law_1.01,0.07291520237922669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,16,power_law_1.01,0.22771201133728028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,32,8,power_law_1.01,0.07807999849319458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,1,power_law_1.01,1.8736383438110351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,64,power_law_1.2,2.664441680908203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,16,power_law_1.01,0.29034240245819093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,32,8,power_law_1.01,0.0803264021873474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,1,power_law_1.01,1.9143360137939454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,32,8,power_law_1.01,0.09125120043754578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,16,power_law_1.01,0.3077440023422241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,1,power_law_1.01,2.3834367752075196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,32,8,power_law_1.01,0.10202239751815796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,32,8,power_law_1.01,0.1085055947303772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,16,power_law_1.01,0.4073023796081543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,1,power_law_1.01,2.9575679779052733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,32,8,power_law_1.01,0.13008639812469483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,16,power_law_1.01,0.534662389755249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,32,8,power_law_1.01,0.15383679866790773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,1,power_law_1.01,3.818521499633789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,32,8,power_law_1.01,0.2065984010696411
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,16,power_law_1.01,0.7200704097747803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,32,8,power_law_1.01,0.24644479751586915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,1,power_law_1.01,4.735756683349609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,32,8,power_law_1.01,0.3194943904876709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,16,power_law_1.01,0.8395199775695801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,64,power_law_1.2,0.15983999967575074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,32,8,power_law_1.01,0.3882944107055664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,1,power_law_1.01,5.722444915771485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,64,power_law_1.2,0.1679360032081604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,32,8,power_law_1.01,0.5617536067962646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,64,power_law_1.2,0.1541375994682312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,2,16,power_law_1.01,1.037452793121338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,32,8,power_law_1.01,0.7520832061767578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,1,power_law_1.01,8.592588806152344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,64,power_law_1.2,0.10927360057830811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,32,8,power_law_1.01,0.8931520462036133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,64,power_law_1.2,0.11603200435638428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,64,power_law_1.2,0.1378432035446167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,2,16,power_law_1.01,1.5150591850280761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,32,8,power_law_1.01,1.3469056129455566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,64,power_law_1.2,0.1261247992515564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,64,power_law_1.2,0.14142719507217408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,1,power_law_1.01,15.748985290527344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,64,power_law_1.2,0.15835520029067993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,32,8,power_law_1.01,2.8019712448120115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,64,power_law_1.2,0.16335359811782837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,64,power_law_1.2,0.15391360521316527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,2,16,power_law_1.01,3.2731136322021483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,64,power_law_1.2,0.16513919830322266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,64,power_law_1.2,0.16133760213851928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,64,power_law_1.2,0.15562880039215088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,64,power_law_1.2,0.16430720090866088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,64,power_law_1.2,0.15470720529556276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,64,power_law_1.2,0.16335359811782837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,16,power_law_1.01,0.07733759880065919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,64,power_law_1.2,0.18709759712219237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,16,power_law_1.2,0.09867519736289979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,16,power_law_1.01,0.13432320356369018
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,64,power_law_1.2,0.186080002784729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,16,power_law_1.01,0.06093440055847168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,16,power_law_1.2,0.13617279529571533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,64,power_law_1.2,0.23930881023406983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,16,power_law_1.01,0.058323198556900026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,32,balanced,0.035690667728583016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,32,balanced,0.03199466566244761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,16,power_law_1.2,0.06540160179138184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,32,balanced,0.03150933235883713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,32,balanced,0.03338133295377096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,32,balanced,0.05470400055249532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,32,balanced,0.05403733253479004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,32,balanced,0.053904001911481224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,32,balanced,0.056346664826075234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,32,balanced,0.05523733297983805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,32,balanced,0.05615466833114624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,32,balanced,0.05597866574923197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,16,power_law_1.01,0.06862720251083373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,32,balanced,0.055258666475613914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,32,balanced,0.05730666716893514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,32,balanced,0.05836800237496694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,32,balanced,0.060922667384147644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,32,balanced,0.062181333700815834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,32,balanced,0.06300800045331319
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,32,balanced,0.06577066580454509
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,32,balanced,0.06824000179767609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,32,balanced,0.07012266914049785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,64,power_law_1.2,0.2588160037994385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,32,balanced,0.07273066540559132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,32,balanced,0.08100800216197968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,32,balanced,0.08823999762535095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,32,balanced,0.10044800241788228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,32,balanced,0.12546133001645407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,32,balanced,0.1548533340295156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,32,balanced,0.18146665891011557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,16,power_law_1.01,0.08018559813499451
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,16,power_law_1.2,0.07624959945678711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,32,balanced,0.21649599075317383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,32,balanced,0.29815999666849774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,32,balanced,0.524943987528483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,64,power_law_1.2,0.3380863904953003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,16,power_law_1.01,0.09032319784164429
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,16,power_law_1.01,0.0918079972267151
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,64,power_law_1.2,0.3997632026672363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,16,power_law_1.2,0.09408640265464782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,16,power_law_1.01,0.0925055980682373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,16,power_law_1.01,0.0970687985420227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,64,power_law_1.2,0.5185152053833008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,16,power_law_1.01,0.098854398727417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,16,power_law_1.2,0.09964159727096558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,16,power_law_1.01,0.0998080015182495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,64,power_law_1.2,0.6940288066864013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,16,power_law_1.01,0.10175360441207885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,16,power_law_1.2,0.13123199939727784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,16,power_law_1.01,0.10311039686203002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,64,power_law_1.2,0.9048704147338867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,16,power_law_1.01,0.10977280139923096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,16,power_law_1.2,0.13313920497894288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,16,power_law_1.01,0.11276799440383911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,64,power_law_1.2,1.2080896377563477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,16,power_law_1.01,0.12136960029602051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,16,power_law_1.2,0.13567999601364136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,16,power_law_1.01,0.1318400025367737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,16,power_law_1.2,0.12577919960021972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,64,power_law_1.2,1.5611200332641602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,16,power_law_1.01,0.1449728012084961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,8,power_law_1.2,0.07184640169143677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,16,power_law_1.01,0.18053760528564453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,16,power_law_1.2,0.13155839443206788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,8,power_law_1.2,0.09488000273704529
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,16,power_law_1.01,0.20383999347686768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,16,power_law_1.2,0.1475648045539856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,64,power_law_1.2,2.8303295135498048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,16,power_law_1.01,0.25220479965209963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,8,power_law_1.2,0.0948032021522522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,16,power_law_1.01,0.3121920108795166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,16,power_law_1.2,0.14454400539398193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,8,power_law_1.2,0.06726400256156921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,16,power_law_1.01,0.4067071914672852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,16,power_law_1.2,0.1471935987472534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,8,power_law_1.2,0.08438400030136109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,16,power_law_1.01,0.5231872081756592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,16,power_law_1.2,0.16161919832229615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,64,power_law_1.2,5.812416076660156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,8,power_law_1.2,0.10320639610290527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,16,power_law_1.01,0.7475647926330566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,16,power_law_1.2,0.16195839643478394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,8,power_law_1.2,0.12109440565109253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,16,power_law_1.01,0.9355775833129882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,16,power_law_1.2,0.1782464027404785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,8,power_law_1.2,0.13868160247802735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,8,power_law_1.2,0.14675840139389038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,16,power_law_1.01,1.1187968254089355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,16,power_law_1.2,0.21157760620117189
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,8,power_law_1.2,0.1550528049468994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,16,power_law_1.2,0.21772160530090331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,16,power_law_1.01,1.6907392501831056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,8,power_law_1.2,0.16508159637451172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,16,power_law_1.2,0.26591360569000244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,8,power_law_1.2,0.16122879981994628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,16,power_law_1.2,0.28954880237579345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,8,power_law_1.2,0.1722432017326355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,16,power_law_1.01,3.304972839355469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,16,power_law_1.2,0.360319995880127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,8,power_law_1.2,0.17827199697494506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,16,power_law_1.2,0.4247039794921875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,2,balanced,0.036501333117485046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,2,balanced,0.06870933373769124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,8,power_law_1.2,0.18227839469909668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,2,balanced,0.07577600081761678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,2,balanced,0.11819199721018474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,2,balanced,0.20490666230519614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,2,balanced,0.20614933967590332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,2,balanced,0.20706133047739664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,2,balanced,0.20642133553822836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,2,balanced,0.20616000890731812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,2,balanced,0.20733332633972168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,2,balanced,0.20946667591730753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,2,balanced,0.2111253341039022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,2,balanced,0.21406932671864828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,2,balanced,0.2161440054575602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,2,balanced,0.22209600607554117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,16,power_law_1.2,0.5450047969818115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,2,balanced,0.22622400522232056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,2,balanced,0.23280000686645508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,2,balanced,0.24397865931193033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,8,power_law_1.2,0.18780159950256348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,2,balanced,0.25679999589920044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,2,balanced,0.2963520089785258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,2,balanced,0.3121760090192159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,2,balanced,0.38681066036224365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,2,balanced,0.4265973170598348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,2,balanced,0.5618453423182169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,2,balanced,0.6603999932607015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,2,balanced,0.919871966044108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,2,balanced,1.1788000265757244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,8,power_law_1.2,0.2016319990158081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,16,power_law_1.2,0.6559487819671631
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,2,balanced,1.441813309987386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,2,balanced,2.2384427388509116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,8,power_law_1.2,0.22760319709777832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,2,balanced,4.307365417480469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,16,power_law_1.2,0.9888383865356445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,8,power_law_1.2,0.24887681007385254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,16,power_law_1.2,1.2177984237670898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,8,power_law_1.2,0.2943871974945068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,8,power_law_1.2,0.3044095993041992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,16,power_law_1.2,1.5158464431762695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,8,power_law_1.2,0.3827199935913086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,1,balanced,0.09842133522033691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,1,balanced,0.19197332859039307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,1,balanced,0.242959996064504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,1,balanced,0.4403359889984131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,1,balanced,0.8282559712727865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,8,power_law_1.2,0.4301504135131836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,1,balanced,1.6007413864135742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,16,power_law_1.2,2.598150444030762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,1,balanced,2.3951573371887207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,1,balanced,2.3805012702941895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,1,balanced,2.3846774101257324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,8,power_law_1.2,0.526854419708252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,1,balanced,2.3846613566080728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,1,balanced,2.3953493436177573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,1,balanced,2.4099733034769693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,1,balanced,2.4146080017089844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,8,power_law_1.2,0.6700863838195801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,1,balanced,2.429743925730387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,1,balanced,2.4408532778422036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,16,power_law_1.2,5.116543960571289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,1,balanced,2.452415943145752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,1,balanced,2.474736054738363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,8,power_law_1.2,0.9035967826843262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,1,power_law_1.01,0.04733439981937408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,1,balanced,2.5278560320536294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,1,balanced,2.5674826304117837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,1,power_law_1.01,0.05479679703712463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,1,balanced,2.6628479957580566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,1,power_law_1.01,0.06893439888954163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,1,balanced,2.7348480224609375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,8,power_law_1.2,1.1807552337646485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,1,balanced,2.9025227228800454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,1,power_law_1.01,0.09151999950408936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,1,balanced,3.0714613596598306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,1,power_law_1.01,0.1168063998222351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,1,balanced,3.4390665690104165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,1,power_law_1.01,0.1608896017074585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,8,power_law_1.2,1.3987135887145996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,1,balanced,3.828495979309082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,1,power_law_1.01,0.17384320497512817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,1,power_law_1.01,0.1819584012031555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,1,balanced,4.387114524841309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,1,power_law_1.01,0.19080959558486937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,1,balanced,5.839152018229167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,1,power_law_1.01,0.1935807943344116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,8,power_law_1.2,1.9627328872680665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,1,power_law_1.01,0.20113279819488525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,1,balanced,6.75434684753418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,1,power_law_1.01,0.20995841026306153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,1,power_law_1.01,0.22017281055450438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,1,balanced,9.652085622151693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,1,power_law_1.01,0.22057600021362306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,1,power_law_1.01,0.23583359718322755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,1,power_law_1.01,0.24595839977264405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,8,power_law_1.2,3.9405502319335937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,1,power_law_1.01,0.2657919883728027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,1,power_law_1.01,0.30461440086364744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,8,power_law_1.01,0.07820159792900086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,1,balanced,18.16687520345052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,1,power_law_1.01,0.34143359661102296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,8,power_law_1.01,0.044019201397895814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,1,power_law_1.01,0.4284800052642822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,1,power_law_1.01,0.49558401107788086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,8,power_law_1.01,0.06204800009727478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,1,power_law_1.01,0.6470335960388184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,8,power_law_1.01,0.07050880193710327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,1,power_law_1.01,0.8310527801513672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,8,power_law_1.01,0.08164479732513427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,8,power_law_1.01,0.1039423942565918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,1,power_law_1.01,1.1124287605285645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,8,power_law_1.01,0.11304320096969604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,1,power_law_1.01,1.4440704345703126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,8,power_law_1.01,0.11291520595550537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,8,power_law_1.01,0.11199359893798828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,1,power_law_1.01,1.9616512298583983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,8,power_law_1.01,0.11749119758605957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,8,power_law_1.01,0.11679359674453735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,1,power_law_1.01,2.4999679565429687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,8,power_law_1.01,0.11516799926757812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,8,power_law_1.01,0.1193727970123291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,32,balanced,0.033386667569478355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,32,balanced,0.032842665910720825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,32,balanced,0.029232000311215717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,32,balanced,0.029487999776999157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,32,balanced,0.03342933456103007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,32,balanced,0.03166399896144867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,32,balanced,0.033557333052158356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,1,power_law_1.01,3.0128000259399412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,32,balanced,0.03348266581694285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,32,balanced,0.033376000821590424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,32,balanced,0.03364799916744232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,32,balanced,0.03345066557327906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,32,balanced,0.03346133232116699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,32,balanced,0.036890665690104164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,32,balanced,0.035429333647092186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,32,balanced,0.035877334574858345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,8,power_law_1.01,0.1272447943687439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,32,balanced,0.03755199909210205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,32,balanced,0.0395413339138031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,32,balanced,0.03992533435424169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,32,balanced,0.04162133236726125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,32,balanced,0.04372266431649526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,32,balanced,0.04807466765244802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,32,balanced,0.0543093333641688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,32,balanced,0.05898133416970571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,32,balanced,0.07409599920113881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,32,balanced,0.09276266892751057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,32,balanced,0.039674667020638786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,32,balanced,0.10968533158302307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,32,balanced,0.038032000263532005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,32,balanced,0.13582400480906168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,32,balanced,0.035349334279696144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,32,balanced,0.1565546691417694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,32,balanced,0.035904000202814736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,8,power_law_1.01,0.13105920553207398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,32,balanced,0.035973332822322845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,32,balanced,0.21266667048136392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,32,balanced,0.03748266647259394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,32,balanced,0.03749866783618927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,32,balanced,0.039605334401130676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,32,balanced,0.3681439956029256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,32,balanced,0.03752533346414566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,32,balanced,0.03752533346414566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,32,balanced,0.03961066653331121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,32,balanced,0.03961066653331121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,32,balanced,0.03952533255020777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,32,balanced,0.03967999915281931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,32,balanced,0.04377066592375437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,32,balanced,0.0448586642742157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,32,balanced,0.04394666850566864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,32,balanced,0.047770669062932335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,32,balanced,0.048058668772379555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,8,power_law_1.01,0.13964799642562867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,32,balanced,0.0521066685517629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,32,balanced,0.053786665201187134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,32,balanced,0.06611733138561249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,32,balanced,0.07443200051784515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,32,balanced,0.09664533535639445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,32,balanced,0.11090667049090068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,32,balanced,0.14517866571744284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,1,power_law_1.01,4.641939163208008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,32,balanced,0.17826133966445923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,32,balanced,0.2125813364982605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,8,power_law_1.01,0.13898240327835082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,32,balanced,0.3123573263486226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,32,balanced,0.5686826705932617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,8,power_law_1.01,0.15603840351104736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,8,power_law_1.01,0.15831040143966674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,8,power_law_1.01,0.17872639894485473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,8,power_law_1.01,0.19444479942321777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,8,power_law_1.01,0.25950720310211184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,1,power_law_1.01,8.895935821533204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,8,power_law_1.01,0.28114559650421145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,8,power_law_1.01,0.36784000396728517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,8,power_law_1.01,0.4072127819061279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,8,power_law_1.01,0.6109951972961426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,8,power_law_1.01,0.7959424018859863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,8,power_law_1.01,0.9746303558349609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,2,balanced,0.075914666056633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,2,balanced,0.1195199986298879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,2,balanced,0.15358933806419373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,2,balanced,0.25593600670496625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,2,balanced,0.4724106788635254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,2,balanced,0.5731626749038696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,2,balanced,0.575488011042277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,2,balanced,0.5826773246129354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,8,power_law_1.01,1.6237695693969727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,2,balanced,0.5888533194859823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,2,balanced,0.5895413160324097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,8,2,balanced,0.037733333806196846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,2,balanced,0.5861866474151611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,8,2,balanced,0.03957333415746689
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,8,2,balanced,0.041706666350364685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,8,2,balanced,0.041264000038305916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,2,balanced,0.5941439867019653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,8,2,balanced,0.043866669138272606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,8,2,balanced,0.0451200008392334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,8,2,balanced,0.04456000030040741
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,2,balanced,0.5922826528549194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,8,2,balanced,0.04363733530044556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,8,2,balanced,0.04453866680463155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,8,2,balanced,0.04566933214664459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,2,balanced,0.6028800010681152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,8,2,balanced,0.04587199787298838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,8,2,balanced,0.046122665206591286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,8,2,balanced,0.045824001232783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,2,balanced,0.6130559841791788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,8,2,balanced,0.04994133114814758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,8,2,balanced,0.062080000837643944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,2,balanced,0.6188906828562418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,8,2,balanced,0.06169599791367849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,8,2,balanced,0.0702400008837382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,8,2,balanced,0.07971733311812083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,2,balanced,0.6270346641540527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,8,2,balanced,0.08873599767684937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,8,2,balanced,0.09962667028109233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,2,balanced,0.7251413663228353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,8,2,balanced,0.11314133803049724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,8,2,balanced,0.14387200276056925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,8,2,balanced,0.17068799336751303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,2,balanced,0.6747146447499593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,8,2,balanced,0.22780267397562662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,2,balanced,0.8026506900787354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,8,2,balanced,0.2911626696586609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,8,2,balanced,0.4084959824879964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,2,balanced,0.7734399636586508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,8,2,balanced,0.5232799847920736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,8,2,balanced,0.6388266483942667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,2,balanced,1.2162826855977376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,2,balanced,0.9891253312428793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,8,2,balanced,0.9750293095906576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,2,balanced,1.5084800720214844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,8,2,balanced,1.882912000020345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,2,balanced,1.5296586354573567
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,8,power_law_1.01,3.1391103744506834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,2,balanced,2.150928020477295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,2,balanced,2.7834345499674478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,2,balanced,3.0736265182495117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,32,power_law_1.01,0.10170880556106568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,2,balanced,5.04747200012207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,2,balanced,9.508570353190104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,32,power_law_1.01,0.12855679988861085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,32,power_law_1.01,0.06375679969787598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,32,power_law_1.01,0.07527679800987244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,32,power_law_1.01,0.08101760149002075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,32,power_law_1.01,0.09574400186538697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,32,power_law_1.2,0.03591679930686951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,128,power_law_1.2,0.08748160004615783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,32,power_law_1.01,0.0944703996181488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,32,power_law_1.2,0.035180801153182985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,128,power_law_1.2,0.09004160165786743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,32,power_law_1.01,0.09441919922828675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,32,power_law_1.2,0.030214399099349976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,128,power_law_1.2,0.08088319897651672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,32,power_law_1.2,0.029580798745155335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,32,power_law_1.01,0.0980288028717041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,2,power_law_1.2,0.046009600162506104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,128,power_law_1.2,0.07155839800834655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,32,power_law_1.2,0.030067199468612672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,2,power_law_1.2,0.05066879987716675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,128,power_law_1.2,0.07613440155982971
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,32,power_law_1.01,0.09743360280990601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,2,power_law_1.2,0.053299200534820554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,32,power_law_1.2,0.030675199627876282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,2,power_law_1.2,0.06410239934921265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,32,power_law_1.2,0.031737598776817325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,128,power_law_1.2,0.07354879975318909
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,2,power_law_1.2,0.07581440210342408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,32,power_law_1.01,0.10361599922180176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,32,power_law_1.2,0.03150720000267029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,2,power_law_1.2,0.10725760459899902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,128,power_law_1.2,0.08053759932518005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,32,power_law_1.2,0.03089280128479004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,2,power_law_1.2,0.11257599592208863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,32,power_law_1.01,0.1027008056640625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,2,power_law_1.2,0.11771520376205444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,32,power_law_1.2,0.03127039968967438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,128,power_law_1.2,0.08593279719352723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,2,power_law_1.2,0.12255359888076782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,32,power_law_1.2,0.03195520043373108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,2,power_law_1.2,0.1220736026763916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,32,power_law_1.01,0.10631040334701539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,128,power_law_1.2,0.08421120047569275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,32,power_law_1.2,0.03294720053672791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,2,power_law_1.2,0.1277184009552002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,2,power_law_1.2,0.13304320573806763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,32,power_law_1.2,0.03322240114212036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,128,power_law_1.2,0.08398720026016235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,32,power_law_1.01,0.10952960252761841
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,2,power_law_1.2,0.13745919466018677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,32,power_law_1.2,0.03336319923400879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,2,power_law_1.2,0.14302719831466676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,128,power_law_1.2,0.07914239764213563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,32,power_law_1.2,0.036800000071525577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,2,power_law_1.2,0.14951679706573487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,32,power_law_1.01,0.11534080505371094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,2,power_law_1.2,0.15510400533676147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,32,power_law_1.2,0.037145599722862244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,128,power_law_1.2,0.08482559919357299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,2,power_law_1.2,0.16388479471206666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,32,power_law_1.2,0.03765760064125061
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,32,power_law_1.01,0.1171455979347229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,2,power_law_1.2,0.1877568006515503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,128,power_law_1.2,0.08355839848518372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,32,power_law_1.2,0.04176000058650971
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,2,power_law_1.2,0.21812479496002196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,32,power_law_1.2,0.045075199007987975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,2,power_law_1.2,0.26828799247741697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,32,power_law_1.01,0.11979520320892334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,128,power_law_1.2,0.08954880237579346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,2,power_law_1.2,0.3176512002944946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,32,power_law_1.2,0.05063040256500244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,2,power_law_1.2,0.41006078720092776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,128,power_law_1.2,0.09214079976081849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,32,power_law_1.2,0.05538560152053833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,32,power_law_1.01,0.13877760171890258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,2,power_law_1.2,0.5063807964324951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,32,power_law_1.2,0.06362879872322083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,128,power_law_1.2,0.09872639775276185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,2,power_law_1.2,0.6970304012298584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,32,power_law_1.2,0.08047360181808472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,32,power_law_1.01,0.1513216018676758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,128,power_law_1.2,0.10371199846267701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,2,power_law_1.2,0.9599040031433106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,32,power_law_1.2,0.10185600519180298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,32,power_law_1.01,0.1807039976119995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,32,power_law_1.2,0.10514559745788574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,128,power_law_1.2,0.11587200164794922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,2,power_law_1.2,1.244883155822754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,32,power_law_1.2,0.2064512014389038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,128,power_law_1.2,0.13386240005493164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,2,power_law_1.2,1.693984031677246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,32,power_law_1.01,0.2099008083343506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,32,power_law_1.2,0.19949439764022828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,128,power_law_1.2,0.15790079832077025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,2,power_law_1.2,1.8800064086914063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,4,32,power_law_1.2,0.24666240215301513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,32,power_law_1.01,0.2744256019592285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,128,power_law_1.2,0.187283194065094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,4,32,power_law_1.2,0.32272000312805177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,128,power_law_1.2,0.24762239456176757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,2,power_law_1.2,2.9981184005737305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,32,power_law_1.01,0.26857600212097166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,128,power_law_1.2,0.3143743991851807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,4,32,power_law_1.2,0.8739583969116211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,32,power_law_1.01,0.358572793006897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,128,power_law_1.2,0.42145280838012694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,2,power_law_1.2,5.634572982788086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,128,power_law_1.2,0.5448448181152343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,32,power_law_1.01,0.43184638023376465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,128,power_law_1.2,0.800607967376709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,32,power_law_1.01,0.5727615833282471
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,128,power_law_1.2,1.0566975593566894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,128,power_law_1.2,1.3001279830932617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,32,power_law_1.01,0.6326272010803222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,128,power_law_1.2,2.0871999740600584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,2,32,power_law_1.01,0.81976318359375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,128,balanced,0.047877331574757896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,128,balanced,0.04781866570313772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,128,balanced,0.04795733094215393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,128,balanced,0.04778666794300079
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,128,balanced,0.04626133541266123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,128,balanced,0.0481279989083608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,128,balanced,0.0499893327554067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,128,balanced,0.05172266562779745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,128,balanced,0.0498986691236496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,128,balanced,0.049829334020614624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,128,balanced,0.05002133548259735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,128,balanced,0.04785066843032837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,128,balanced,0.05160533388455709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,128,balanced,0.04981866478919983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,128,balanced,0.05197333296140035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,128,balanced,0.050106664498647056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,128,balanced,0.04994133114814758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,128,balanced,0.05227733155091604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,128,balanced,0.05195199946562449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,128,balanced,0.05190933247407278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,128,balanced,0.05604266623655955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,128,balanced,0.06256533165772755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,128,balanced,0.06832533578077953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,128,balanced,0.08275199929873149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,128,balanced,0.09459199508031209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,128,balanced,0.11969600121180217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,128,balanced,0.14871999621391296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,128,balanced,0.17552000284194946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,128,balanced,0.24827200174331665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,128,balanced,0.45362667242685956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,128,power_law_1.2,4.094758224487305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,1,power_law_1.2,0.17015039920806885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,2,32,power_law_1.01,1.2597951889038086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,1,power_law_1.2,0.16570240259170532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,1,power_law_1.2,0.21811840534210206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,1,power_law_1.2,0.31254398822784424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,2,32,power_law_1.01,2.7203647613525392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,1,power_law_1.2,0.42993922233581544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,8,1,power_law_1.2,0.04761599898338318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,8,1,power_law_1.2,0.05191680192947388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,8,1,power_law_1.2,0.056992000341415404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,1,power_law_1.2,0.8041472434997559
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,8,1,power_law_1.2,0.06750720143318176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,64,power_law_1.2,0.07701759934425353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,8,1,power_law_1.2,0.074726402759552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,8,1,power_law_1.2,0.10940799713134766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,8,1,power_law_1.2,0.115065598487854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,64,power_law_1.2,0.09500160217285156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,8,1,power_law_1.2,0.11938560009002686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,1,power_law_1.2,0.8686847686767578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,8,1,power_law_1.2,0.12104959487915039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,8,1,power_law_1.2,0.12280319929122925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,64,power_law_1.2,0.056934398412704465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,8,1,power_law_1.2,0.12811520099639892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,8,1,power_law_1.2,0.1309440016746521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,1,power_law_1.2,0.9356096267700196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,8,1,power_law_1.2,0.1335935950279236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,64,power_law_1.2,0.060864001512527466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,8,1,power_law_1.2,0.13697279691696168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,8,1,power_law_1.2,0.14824960231781006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,8,1,power_law_1.2,0.1534719944000244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,64,power_law_1.2,0.06408960223197938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,1,power_law_1.2,0.9773119926452637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,8,1,power_law_1.2,0.16661759614944457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,16,balanced,0.043968002001444496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,16,balanced,0.044922664761543274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,16,balanced,0.043951998154322304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,16,balanced,0.04757866760094961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,16,balanced,0.0476746658484141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,16,balanced,0.0516480008761088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,16,balanced,0.05186666548252106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,8,1,power_law_1.2,0.19705599546432495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,16,balanced,0.051551997661590576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,16,balanced,0.053472002347310386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,16,balanced,0.05219733218352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,16,balanced,0.05380799869696299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,16,balanced,0.053914666175842285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,16,balanced,0.05584000051021576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,16,balanced,0.05592533449331919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,16,balanced,0.06087466577688853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,16,balanced,0.06234666705131531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,16,balanced,0.06198933223883311
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,16,balanced,0.07101333141326904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,8,1,power_law_1.2,0.22652800083160402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,64,power_law_1.2,0.07729920148849487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,16,balanced,0.07271466652552287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,16,balanced,0.09712533156077068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,16,balanced,0.11180800199508667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,16,balanced,0.15639467040697733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,16,balanced,0.18319465716679892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,16,balanced,0.25145600239435834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,16,balanced,0.30209600925445557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,8,1,power_law_1.2,0.29160959720611573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,16,balanced,0.42322667439778644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,16,balanced,0.5325760046641032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,16,balanced,0.6476853291193644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,1,power_law_1.2,0.9925439834594727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,16,balanced,0.9969813028971354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,8,1,power_law_1.2,0.34823040962219237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,64,power_law_1.2,0.07633919715881347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,16,balanced,1.9484480222066243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,8,1,power_law_1.2,0.4755136013031006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,8,1,power_law_1.2,0.5981247901916504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,64,power_law_1.2,0.07648000121116638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,1,power_law_1.2,1.0444479942321778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,8,1,power_law_1.2,0.847878360748291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,8,1,power_law_1.2,1.0862015724182128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,64,power_law_1.2,0.07761279940605163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,1,power_law_1.2,1.0992128372192382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,8,1,power_law_1.2,1.5651328086853027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,64,power_law_1.2,0.07979519963264466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,8,1,power_law_1.2,2.0793664932250975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,64,power_law_1.2,0.08024960160255432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,1,power_law_1.2,1.1070783615112305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,8,1,power_law_1.2,2.553273582458496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,64,power_law_1.2,0.08258559703826904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,1,power_law_1.2,1.1628095626831054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,8,1,power_law_1.2,3.9903358459472655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,64,power_law_1.2,0.08577280044555664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,64,power_law_1.2,0.08914560079574585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,1,power_law_1.2,1.193280029296875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,128,balanced,0.04995200037956238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,128,balanced,0.04978133241335551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,128,balanced,0.0481279989083608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,128,balanced,0.0539680023988088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,128,balanced,0.0544106662273407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,64,power_law_1.2,0.09638400077819824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,128,balanced,0.05412800113360087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,128,balanced,0.052602668603261314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,128,balanced,0.056218668818473816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,128,balanced,0.05418133238951365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,128,balanced,0.05589333176612854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,128,balanced,0.056464001536369324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,128,balanced,0.0569706658522288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,128,balanced,0.06044800082842509
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,8,1,power_law_1.2,7.809843444824219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,128,balanced,0.058287998040517174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,128,balanced,0.062090665102005005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,128,balanced,0.06683733562628429
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,128,balanced,0.06637866795063019
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,128,balanced,0.08249066770076752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,128,balanced,0.0867146650950114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,128,balanced,0.1095199982325236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,128,balanced,0.12050666411717732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,128,balanced,0.1725813349088033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,128,balanced,0.17157334089279175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,1,power_law_1.2,1.2278528213500977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,128,balanced,0.21898667017618814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,128,balanced,0.25828800598780316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,128,balanced,0.34652801354726154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,128,balanced,0.4225173393885295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,64,power_law_1.2,0.09820160269737244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,2,128,balanced,0.5150719881057739
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,2,128,balanced,0.7575146357218424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,2,128,balanced,1.605471928914388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,64,power_law_1.2,0.10587519407272339
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,1,power_law_1.2,1.2845184326171875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,64,power_law_1.2,0.12640639543533325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,1,power_law_1.2,1.4315967559814453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,64,power_law_1.2,0.13477760553359985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,1,power_law_1.2,1.575699234008789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,64,power_law_1.2,0.1728767991065979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,64,power_law_1.2,0.18636800050735475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,1,power_law_1.2,1.873151969909668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,64,power_law_1.2,0.23856000900268554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,128,power_law_1.01,0.060102397203445436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,1,power_law_1.2,2.024959945678711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,128,power_law_1.01,0.058950400352478026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,64,power_law_1.2,0.26225919723510743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,128,power_law_1.01,0.05851519703865051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,128,power_law_1.01,0.056576001644134524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,64,power_law_1.2,0.3696320056915283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,16,power_law_1.2,0.06914560198783874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,1,power_law_1.2,2.5366783142089844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,128,power_law_1.01,0.05825279951095581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,128,power_law_1.01,0.05902720093727112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,64,power_law_1.2,0.42987518310546874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,16,power_law_1.2,0.06302719712257385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,128,power_law_1.01,0.059545600414276124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,128,power_law_1.01,0.060575997829437254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,1,power_law_1.2,3.034579277038574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,16,power_law_1.2,0.05141760110855102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,128,power_law_1.01,0.060211199522018435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,64,power_law_1.2,0.6470016002655029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,128,power_law_1.01,0.06165120005607605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,16,power_law_1.2,0.05796480178833008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,128,power_law_1.01,0.06270719766616821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,1,power_law_1.2,4.004915237426758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,128,power_law_1.01,0.06000639796257019
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,16,power_law_1.2,0.06115840077400207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,128,power_law_1.01,0.0634112000465393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,64,power_law_1.2,0.8365632057189941
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,128,power_law_1.01,0.0639743983745575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,16,power_law_1.2,0.07711359858512878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,128,power_law_1.01,0.06917120218276977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,128,power_law_1.01,0.06915199756622314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,1,power_law_1.2,5.268025588989258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,16,power_law_1.2,0.07848960161209106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,128,power_law_1.01,0.07167999744415283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,1,64,power_law_1.2,0.9489791870117188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,128,power_law_1.01,0.08581119775772095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,16,power_law_1.2,0.07938560247421264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,128,power_law_1.01,0.09296000003814697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,128,power_law_1.01,0.11879040002822876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,16,power_law_1.2,0.08250240087509156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,128,power_law_1.01,0.13855359554290772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,1,power_law_1.2,7.211872100830078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,16,power_law_1.2,0.08335360288619995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,1,64,power_law_1.2,1.3834624290466309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,128,power_law_1.01,0.1824447989463806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,128,power_law_1.01,0.23098878860473632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,16,power_law_1.2,0.08412160277366638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,128,power_law_1.01,0.30180480480194094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,16,power_law_1.2,0.08739200234413147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,128,power_law_1.01,0.38380160331726076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,1,power_law_1.2,9.030271911621094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,16,power_law_1.2,0.08958079814910888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,128,power_law_1.01,0.5556096076965332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,128,power_law_1.01,0.7091775894165039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,16,power_law_1.2,0.09241600036621093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,1,64,power_law_1.2,3.4554367065429688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,128,power_law_1.01,0.9312831878662109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,16,power_law_1.2,0.09916160106658936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,16,power_law_1.2,0.10448640584945679
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,128,power_law_1.01,1.4066880226135254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,1,1,power_law_1.2,10.95203857421875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,16,power_law_1.2,0.10851839780807496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,128,power_law_1.01,2.9055999755859374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,16,power_law_1.2,0.12543359994888306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,16,power_law_1.2,0.13953919410705568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,16,power_law_1.2,0.17118079662323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,1,1,power_law_1.2,17.050758361816406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,4,4,balanced,0.03392533212900162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,16,power_law_1.2,0.2010495901107788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,4,4,balanced,0.03549866626660029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,4,4,balanced,0.03562666724125544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,4,4,balanced,0.03759466608365377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,4,4,balanced,0.036746665835380554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,4,4,balanced,0.036144000788529716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,4,4,balanced,0.03755733370780945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,4,4,balanced,0.03933866570393244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,4,4,balanced,0.03786666691303253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,4,4,balanced,0.039520000418027244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,4,4,balanced,0.03945599993069967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,4,4,balanced,0.043791999419530235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,4,4,balanced,0.044735997915267944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,4,4,balanced,0.04664533336957296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,4,4,balanced,0.051856001218159996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,4,4,balanced,0.05186666548252106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,4,4,balanced,0.06353066861629486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,4,4,balanced,0.07261866827805837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,4,4,balanced,0.07857066889603932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,4,4,balanced,0.09494933485984802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,4,4,balanced,0.10060800115267436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,4,4,balanced,0.1255466639995575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,4,4,balanced,0.14642666776974997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,4,4,balanced,0.1990293264389038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,4,4,balanced,0.24864532550175986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,4,4,balanced,0.34643201033274335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,4,4,balanced,0.4355413516362508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,4,4,balanced,0.5355519851048788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,4,4,balanced,0.8131466706593832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,16,power_law_1.2,0.26472959518432615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,4,4,balanced,1.5524373054504395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,16,power_law_1.2,0.3016767978668213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,1,8,balanced,0.08508800466855367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,1,8,balanced,0.10244266192118327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,1,8,balanced,0.08767466743787129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,1,8,balanced,0.08474666873613994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,1,8,balanced,0.08500267068545024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,1,8,balanced,0.08457066615422566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,1,8,balanced,0.08553066849708557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,1,8,balanced,0.0846453309059143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,1,8,balanced,0.08824533224105835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,1,8,balanced,0.08764266967773438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,1,8,balanced,0.09033067027727763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,1,8,balanced,0.09589333335558574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,1,8,balanced,0.09699199597040813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,16,power_law_1.2,0.4139391899108887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,1,8,balanced,0.09550399581591289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,1,8,balanced,0.10777599612871806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,1,8,balanced,0.10345600048700969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,1,8,balanced,0.10539733370145161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,1,8,balanced,0.1234933336575826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,1,8,balanced,0.12110933661460876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,1,8,balanced,0.16310933232307434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,1,8,balanced,0.15125333269437155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,1,8,balanced,0.2286293307940165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,1,8,balanced,0.25058666865030926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,1,8,balanced,0.34913599491119385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,1,8,balanced,0.42155198256174725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,1,8,balanced,0.6101013422012329
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,1,8,balanced,0.8001386324564616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,1,8,balanced,0.9849173227945963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,1,8,balanced,1.4954454104105632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,1,8,balanced,3.0098241170247397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,16,power_law_1.2,0.5068096160888672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,16,power_law_1.2,0.6843200206756592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,1,1,power_law_1.2,32.23570556640625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,32,power_law_1.01,0.044172799587249754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,16,power_law_1.2,0.8347007751464843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,32,power_law_1.01,0.04883840084075928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,32,power_law_1.01,0.037529599666595456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,32,power_law_1.01,0.03480960130691528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,32,power_law_1.01,0.03492479920387268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,32,power_law_1.01,0.03607040047645569
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,32,power_law_1.01,0.036211198568344115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,4,16,power_law_1.2,1.069267177581787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,32,power_law_1.01,0.036422398686408994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,32,power_law_1.01,0.036371201276779175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,32,power_law_1.01,0.03688960075378418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,8,balanced,0.043712000052134194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,8,balanced,0.04372799893220266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,32,power_law_1.01,0.03680639863014221
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,8,balanced,0.045797333121299744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,8,balanced,0.047824000318845115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,8,balanced,0.05635733405749003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,8,balanced,0.07946133116881053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,8,balanced,0.09910933176676433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,8,balanced,0.09707199533780415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,8,balanced,0.0986293355623881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,8,balanced,0.09690133730570476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,32,power_law_1.01,0.03718400001525879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,8,balanced,0.10242133339246114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,8,balanced,0.10322133700052898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,8,balanced,0.10339732964833577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,8,balanced,0.10608533024787903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,8,balanced,0.11531733473141988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,8,balanced,0.11617599924405415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,8,balanced,0.12864533066749573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,8,balanced,0.14059199889500937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,32,power_law_1.01,0.03787519931793213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,8,balanced,0.15170133113861084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,8,balanced,0.1779306729634603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,8,balanced,0.1998186707496643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,8,balanced,0.24778133630752563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,8,balanced,0.31194667021433514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,8,balanced,0.4099200169245402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,32,power_law_1.01,0.03804160058498383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,8,balanced,0.5227253437042236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,8,balanced,0.729861338933309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,8,balanced,0.9350240230560303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,32,power_law_1.01,0.04349440038204193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,8,balanced,1.1581919987996419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,32,power_law_1.01,0.043161600828170776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,8,balanced,1.7783573468526204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,4,16,power_law_1.2,1.7213504791259766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,32,power_law_1.01,0.0458624005317688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,8,balanced,3.4684534072875977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,32,power_law_1.01,0.05127679705619812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,32,power_law_1.01,0.05414400100708008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,32,power_law_1.01,0.06170240044593811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,32,power_law_1.01,0.0662015974521637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,4,64,power_law_1.2,0.023027199506759643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,32,power_law_1.01,0.07905920147895813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,32,power_law_1.01,0.10383360385894776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,4,64,power_law_1.2,0.023827199637889863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,32,power_law_1.01,0.13000320196151732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,4,64,power_law_1.2,0.023455999791622162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,32,power_law_1.01,0.13879040479660035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,4,64,power_law_1.2,0.02353920042514801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,32,power_law_1.01,0.19898879528045654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,4,64,power_law_1.2,0.02622720003128052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,32,power_law_1.01,0.2476288080215454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,4,16,power_law_1.2,3.6698047637939455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,4,64,power_law_1.2,0.02584959864616394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,2688,1856,6,128,1,32,power_law_1.01,0.3001728057861328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,4,64,power_law_1.2,0.026259198784828186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,2688,1856,6,128,1,32,power_law_1.01,0.43837437629699705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,4,64,power_law_1.2,0.026713600754737853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,4,64,power_law_1.2,0.02810879945755005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,2688,1856,6,128,1,32,power_law_1.01,0.9736384391784668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,4,64,power_law_1.2,0.0320576012134552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,2,power_law_1.2,0.04865919947624207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,4,64,power_law_1.2,0.030067199468612672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,2,power_law_1.2,0.0837119996547699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,4,64,power_law_1.2,0.03020800054073334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,2,power_law_1.2,0.06254079937934875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,4,64,power_law_1.2,0.030508801341056824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,2,power_law_1.2,0.08701440095901489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,4,64,power_law_1.2,0.030112001299858093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,2,power_law_1.2,0.11137280464172364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,4,64,power_law_1.2,0.032627201080322264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,2,power_law_1.2,0.13639039993286134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,4,64,power_law_1.2,0.03374719917774201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,2,power_law_1.2,0.1827455997467041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,4,64,power_law_1.2,0.03335680067539215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,4,64,power_law_1.2,0.03484799861907959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,2,power_law_1.2,0.19448319673538209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,32,balanced,0.04884799818197886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,32,balanced,0.05825600028038025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,32,balanced,0.06293866535027821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,32,balanced,0.08833066622416179
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,32,balanced,0.10961066683133443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,32,balanced,0.10709866881370544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,32,balanced,0.10630933443705241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,32,balanced,0.1076746682325999
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,64,power_law_1.2,0.04206080138683319
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,32,balanced,0.10678399602572124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,2,power_law_1.2,0.20184319019317626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,32,balanced,0.10762666662534077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,32,balanced,0.10961066683133443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,32,balanced,0.10850666960080464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,32,balanced,0.11135466893513997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,32,balanced,0.11062933007876079
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,32,balanced,0.1149120032787323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,32,balanced,0.11712533235549927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,32,balanced,0.11874666810035706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,32,balanced,0.13474133610725403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,32,balanced,0.13674666484196982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,64,power_law_1.2,0.04288640022277832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,32,balanced,0.16866666078567505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,32,balanced,0.17110933860143027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,2,power_law_1.2,0.21426560878753662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,32,balanced,0.23692800601323447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,32,balanced,0.21192532777786255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,32,balanced,0.2703253428141276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,32,balanced,0.2966880003611247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,32,balanced,0.41093866030375165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,32,balanced,0.46366933981577557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,64,power_law_1.2,0.04937599897384644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,2,32,balanced,0.5786933501561483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,2,power_law_1.2,0.21998720169067382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,2,32,balanced,0.8085813522338867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,64,power_law_1.2,0.05756160020828247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,2,32,balanced,1.6766986846923828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,2,power_law_1.2,0.23998079299926758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,64,power_law_1.2,0.06878719925880432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,1,power_law_1.01,0.047577598690986635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,2,power_law_1.2,0.24116480350494385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,1,power_law_1.01,0.057196801900863646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,64,power_law_1.2,0.09416319727897644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,2,power_law_1.2,0.25217280387878416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,1,power_law_1.01,0.0697920024394989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,64,power_law_1.2,0.10611200332641602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,1,power_law_1.01,0.09141119718551635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,2,power_law_1.2,0.2651007890701294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,1,power_law_1.01,0.11646080017089844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,64,power_law_1.2,0.15319679975509642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,2,power_law_1.2,0.27971839904785156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,1,power_law_1.01,0.1712000012397766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,1,power_law_1.01,0.17671680450439453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,2,power_law_1.2,0.29932799339294436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,64,power_law_1.2,0.22796800136566162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,1,power_law_1.01,0.18853119611740113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,2,power_law_1.2,0.3409535884857178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,1,power_law_1.01,0.19572479724884034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,64,power_law_1.2,0.2949376106262207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,1,power_law_1.01,0.19780479669570922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,2,power_law_1.2,0.3683583974838257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,1,power_law_1.01,0.20300159454345704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,2,power_law_1.2,0.4441984176635742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,1,power_law_1.01,0.2130431890487671
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,64,power_law_1.2,0.46467838287353513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,1,power_law_1.01,0.21639039516448974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,2,power_law_1.2,0.48700799942016604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,1,power_law_1.01,0.22497920989990233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,1,power_law_1.01,0.2422976016998291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,2,power_law_1.2,0.6171135902404785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,1,power_law_1.01,0.25461759567260744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,64,power_law_1.2,1.0476160049438477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,1,power_law_1.01,0.27370240688323977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,2,power_law_1.2,0.7480959892272949
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,1,power_law_1.01,0.3207808017730713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,2,power_law_1.2,0.9967743873596191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,1,power_law_1.01,0.3547391891479492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,1,power_law_1.2,0.05443840026855469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,1,power_law_1.01,0.4461503982543945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,2,power_law_1.2,1.2230400085449218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,1,power_law_1.01,0.5196608066558838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,1,power_law_1.2,0.08915839791297912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,1,power_law_1.01,0.6843135833740235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,2,power_law_1.2,1.6852735519409179
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,1,power_law_1.01,0.9086272239685058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,1,power_law_1.2,0.08693119883537292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,1,power_law_1.01,1.2173312187194825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,2,power_law_1.2,2.109427261352539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,1,power_law_1.2,0.12348159551620483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,1,power_law_1.01,1.5385024070739746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,2,power_law_1.2,2.60118408203125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,1,power_law_1.2,0.17951359748840331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,1,power_law_1.01,2.1432512283325194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,1,power_law_1.2,0.2617408037185669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,1,power_law_1.01,2.7610111236572266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,2,power_law_1.2,3.939936065673828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,1,power_law_1.2,0.29684479236602784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,1,power_law_1.01,3.371206283569336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,1,power_law_1.2,0.40183677673339846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,1,power_law_1.2,0.4295231819152832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,2,power_law_1.01,0.0989184021949768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,1,power_law_1.01,5.135961532592773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,2,power_law_1.2,7.83834228515625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,1,power_law_1.2,0.44033279418945315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,1,power_law_1.2,0.4796544075012207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,2,power_law_1.01,0.14586880207061767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,1,power_law_1.2,0.49993600845336916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,1,power_law_1.01,9.940831756591797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,2,power_law_1.01,0.242195200920105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,1,power_law_1.2,0.5123136043548584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,1,power_law_1.2,0.5441152095794678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,2,power_law_1.01,0.17537920475006102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,1,power_law_1.2,0.5594304084777832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,2,power_law_1.01,0.29322240352630613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,1,power_law_1.2,0.5854720115661621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,1,power_law_1.2,0.6242047786712647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,2,power_law_1.01,0.4930880069732666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,1,power_law_1.2,0.696288013458252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,16,balanced,0.029461334149042766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,16,balanced,0.02940266579389572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,16,balanced,0.030752000709374745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,16,balanced,0.03148266673088074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,16,balanced,0.033439998825391136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,16,balanced,0.03385066737731298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,16,balanced,0.033615998923778534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,16,balanced,0.03561066587766012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,16,balanced,0.03568533311287562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,16,balanced,0.03705599904060364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,16,balanced,0.03543466577927271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,16,balanced,0.035786665976047516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,16,balanced,0.03769599894682566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,16,balanced,0.03764266769091288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,16,balanced,0.04144533226887385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,16,balanced,0.041536000867684685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,16,balanced,0.04223999877770742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,16,balanced,0.046069333950678505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,16,balanced,0.04794133206208547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,16,balanced,0.05601066847642263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,16,balanced,0.06484266618887584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,2,power_law_1.01,0.5956736087799073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,16,balanced,0.07677866518497467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,16,balanced,0.08273066580295563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,16,balanced,0.1009333332379659
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,1,power_law_1.2,0.7472447872161865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,16,balanced,0.12527466813723245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,16,balanced,0.15426133076349893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,16,balanced,0.1930346687634786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,16,balanced,0.23006399472554526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,16,balanced,0.3227786620457967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,16,balanced,0.5900160074234009
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,1,power_law_1.2,0.8875264167785645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,2,power_law_1.01,0.7589632034301758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,2,balanced,0.05418133238951365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,2,balanced,0.0543093333641688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,2,balanced,0.05807999769846598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,2,balanced,0.07918400069077809
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,2,balanced,0.11052266756693523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,2,balanced,0.17101333538691202
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,2,balanced,0.23003733158111572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,2,balanced,0.2344320019086202
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,1,power_law_1.2,0.9996607780456543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,2,balanced,0.2344213326772054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,2,balanced,0.237936000029246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,2,balanced,0.23609066009521484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,2,balanced,0.23813867568969727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,2,balanced,0.24060799678166708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,2,balanced,0.24805333216985068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,2,balanced,0.2516319950421651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,2,balanced,0.25430933634440106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,2,balanced,0.26850666602452594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,2,balanced,0.29293866952260333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,2,balanced,0.3102133274078369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,2,balanced,0.35622934500376385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,2,power_law_1.01,0.8222463607788086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,2,balanced,0.4033973217010498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,2,balanced,0.48286934693654376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,2,balanced,0.5800000031789144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,2,balanced,0.740768035252889
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,1,power_law_1.2,1.2470848083496093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,2,balanced,0.9634772936503092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,2,balanced,1.290239969889323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,2,balanced,1.7051733334859211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,2,balanced,2.1459733645121255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,2,power_law_1.01,0.8246720314025879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,2,balanced,3.167695999145508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,1,power_law_1.2,1.5585151672363282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,2,balanced,6.059722900390625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,2,power_law_1.01,0.8917119979858399
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,1,power_law_1.2,1.9875648498535157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,2,power_law_1.01,0.9053183555603027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,1,power_law_1.2,2.284480094909668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,2,power_law_1.01,0.9554944038391113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,64,power_law_1.01,0.1379520058631897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,1,power_law_1.2,3.0279232025146485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,64,power_law_1.01,0.06741120219230652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,2,power_law_1.01,0.9685824394226075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,64,power_law_1.01,0.06301440000534057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,64,power_law_1.01,0.054816001653671266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,1,power_law_1.2,3.941996765136719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,2,power_law_1.01,0.9766847610473632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,64,power_law_1.01,0.059539198875427246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,64,power_law_1.01,0.0627135992050171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,2,power_law_1.01,1.0214783668518066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,64,power_law_1.01,0.06906880140304565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,1,power_law_1.2,4.7807167053222654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,64,power_law_1.01,0.06985599994659424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,2,power_law_1.01,1.0548095703125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,64,power_law_1.01,0.06910079717636108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,1,power_law_1.2,0.08271359801292419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,64,power_law_1.01,0.0695039987564087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,2,power_law_1.01,1.1270079612731934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,64,power_law_1.01,0.07116159796714783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,1,power_law_1.2,6.967481231689453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,1,power_law_1.2,0.07953919768333435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,64,power_law_1.01,0.0692736029624939
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,64,power_law_1.01,0.07260159850120544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,2,power_law_1.01,1.163212776184082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,1,power_law_1.2,0.10709760189056397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,64,power_law_1.01,0.07212799787521362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,64,power_law_1.01,0.08021119832992554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,2,power_law_1.01,1.2931839942932128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,1,power_law_1.2,0.1466048002243042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,64,power_law_1.01,0.08206719756126404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,64,power_law_1.01,0.08513919711112976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,1,power_law_1.2,0.1925055980682373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,2,power_law_1.01,1.3508095741271973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,1,power_law_1.2,12.934873962402344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,64,power_law_1.01,0.09456639885902404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,64,power_law_1.01,0.10212479829788208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,1,power_law_1.2,0.33812479972839354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,2,power_law_1.01,1.5589311599731446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,64,power_law_1.01,0.12074240446090698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,1,power_law_1.2,0.3728895902633667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,64,power_law_1.01,0.13555840253829957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,2,power_law_1.01,1.8029951095581054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,64,power_law_1.01,0.1743872046470642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,1,power_law_1.2,0.3897727966308594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,64,power_law_1.01,0.1841279983520508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,2,power_law_1.01,2.1745536804199217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,64,power_law_1.01,0.2500864028930664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,1,power_law_1.2,0.4072512149810791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,64,power_law_1.01,0.34882559776306155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,1,power_law_1.2,0.40784640312194825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,2,power_law_1.01,2.304377555847168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,64,power_law_1.01,0.5040319919586181
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,1,power_law_1.2,0.43067522048950196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,64,power_law_1.01,0.5636352062225342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,1,power_law_1.2,0.44599041938781736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,2,power_law_1.01,2.9000576019287108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,16,balanced,0.04193066557248434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,16,balanced,0.03750933210055033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,16,balanced,0.03595199932654699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,16,balanced,0.0377813329299291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,16,balanced,0.03945599993069967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,16,balanced,0.04197333256403605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,64,power_law_1.01,0.6267903804779053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,16,balanced,0.041477332512537636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,16,balanced,0.04179200033346812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,16,balanced,0.042021334171295166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,16,balanced,0.04162666698296865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,16,balanced,0.04181333382924398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,16,balanced,0.04160533348719279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,16,balanced,0.04221333563327789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,16,balanced,0.04368533194065094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,16,balanced,0.04775999983151754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,16,balanced,0.04531733194986979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,16,balanced,0.04775466521581014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,16,balanced,0.04970666766166687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,16,balanced,0.051967998345692955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,16,balanced,0.062224000692367554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,16,balanced,0.06037333110968272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,16,balanced,0.07881066699822743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,16,balanced,0.08545066912968953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,16,balanced,0.10940800110499065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,1,power_law_1.2,0.4612864017486572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,16,balanced,0.12524799505869547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,16,balanced,0.16547200083732605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,16,balanced,0.19164800643920898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,2688,1856,6,128,1,16,balanced,0.23179733753204346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,2688,1856,6,128,1,16,balanced,0.3358986775080363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,2688,1856,6,128,1,16,balanced,0.6192959944407145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,64,power_law_1.01,1.024614429473877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,2,power_law_1.01,3.584492874145508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,1,power_law_1.2,0.4857471942901611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,1,power_law_1.2,0.49939842224121095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,64,power_law_1.01,1.985593605041504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,2,power_law_1.01,4.348767852783203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,1,power_law_1.2,0.514847993850708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,1,power_law_1.2,0.5531199932098388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,2,power_law_1.01,6.150271987915039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,1,power_law_1.2,0.647814416885376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,1,power_law_1.2,0.6879424095153809
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,128,balanced,0.0336053321758906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,128,balanced,0.035631999373435974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,128,balanced,0.03572800010442734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,128,balanced,0.03465600063403448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,128,balanced,0.03566933423280716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,128,balanced,0.03566399961709976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,128,balanced,0.03758399933576584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,128,balanced,0.03566933423280716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,128,balanced,0.03781333317359289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,128,balanced,0.037471999724706016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,128,balanced,0.037818667789300285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,128,balanced,0.03777066618204117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,128,balanced,0.03762666632731756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,128,balanced,0.03987200061480204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,128,balanced,0.041573333243529
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,1,power_law_1.2,0.837279987335205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,128,balanced,0.041759997606277466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,128,balanced,0.043893332282702126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,128,balanced,0.04572266836961111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,128,balanced,0.04585599899291992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,128,balanced,0.04996266464392344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,128,balanced,0.053530668218930565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,128,balanced,0.05753066639105479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,128,balanced,0.06297066807746887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,128,balanced,0.08660266796747844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,128,balanced,0.1141919990380605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,128,balanced,0.14800533652305603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,128,balanced,0.18532266219456991
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,2,128,balanced,0.2206826607386271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,2,128,balanced,0.32493333021799725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,2,128,balanced,0.608784000078837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,2,power_law_1.01,11.160050964355468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,1,power_law_1.2,1.0278079986572266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,1,power_law_1.2,0.054841601848602296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,1,power_law_1.2,1.2900416374206543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,1,power_law_1.2,0.07014399766921997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,1,power_law_1.2,0.0921280026435852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,1,power_law_1.2,0.12347520589828491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,1,power_law_1.2,0.16510720252990724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,1,power_law_1.2,1.4610560417175293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,1,power_law_1.2,0.3157952070236206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,1,power_law_1.2,0.3334719896316528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,1,power_law_1.2,0.3488384008407593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,1,power_law_1.2,0.36600959300994873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,1,power_law_1.2,1.9822271347045899
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,1,power_law_1.2,0.355020809173584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,1,power_law_1.2,0.3677311897277832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,1,power_law_1.2,0.3787008047103882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,1,power_law_1.2,0.3959552049636841
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,1,power_law_1.2,0.3999871969223022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,1,power_law_1.2,2.395635223388672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,1,power_law_1.2,0.42115201950073244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,1,power_law_1.2,0.43479042053222655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,8,power_law_1.01,0.03761920034885406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,1,power_law_1.2,0.44087681770324705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,8,power_law_1.01,0.03639039993286133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,1,power_law_1.2,0.49381117820739745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,8,power_law_1.01,0.03454079926013946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,1,power_law_1.2,0.5219327926635742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,8,power_law_1.01,0.037049600481987
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,32,power_law_1.2,0.1204416036605835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,1,power_law_1.2,3.3492225646972655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,8,power_law_1.01,0.03993600010871887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,1,power_law_1.2,0.6123519897460937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,8,power_law_1.01,0.05487359762191772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,1,power_law_1.2,0.6631872177124023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,8,power_law_1.01,0.055846399068832396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,32,power_law_1.2,0.08370559811592101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,1,power_law_1.2,0.8373567581176757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,8,power_law_1.01,0.0580735981464386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,8,power_law_1.01,0.0576960027217865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,1,power_law_1.2,0.9943360328674317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,32,power_law_1.2,0.08241919875144958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,8,power_law_1.01,0.05793280005455017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,1,power_law_1.2,1.307551956176758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,8,power_law_1.01,0.05667200088500977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,1,power_law_1.2,4.425395202636719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,32,power_law_1.2,0.08475520014762879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,8,power_law_1.01,0.060115200281143186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,8,power_law_1.01,0.062041598558425906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,1,power_law_1.2,1.6518720626831054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,32,power_law_1.2,0.10017919540405273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,8,power_law_1.01,0.06580479741096497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,8,power_law_1.01,0.0718720018863678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,1,power_law_1.2,2.2957952499389647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,32,power_law_1.2,0.1071936011314392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,8,power_law_1.01,0.07778559923171997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,8,power_law_1.01,0.07780479788780212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,32,power_law_1.2,0.11619839668273926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,8,power_law_1.01,0.08878719806671143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,1,power_law_1.2,2.9372352600097655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,8,power_law_1.01,0.10537600517272949
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,4,1,power_law_1.2,5.380384063720703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,32,power_law_1.2,0.11271040439605713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,8,power_law_1.01,0.11952639818191528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,8,power_law_1.01,0.12758400440216064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,1,power_law_1.2,3.597260665893555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,32,power_law_1.2,0.11033600568771362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,8,power_law_1.01,0.1733888030052185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,32,power_law_1.2,0.12120959758758545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,8,power_law_1.01,0.19761919975280762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,8,power_law_1.01,0.3101952075958252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,32,power_law_1.2,0.13304320573806763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,1,power_law_1.2,5.451712036132813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,8,power_law_1.01,0.36856319904327395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,32,power_law_1.2,0.13251199722290039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,8,power_law_1.01,0.5022079944610596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,32,power_law_1.2,0.13571840524673462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,4,1,power_law_1.2,8.238240051269532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,8,power_law_1.01,0.6737279891967773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,32,power_law_1.2,0.13456640243530274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,32,power_law_1.2,0.14257279634475709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,8,power_law_1.01,0.7902400016784668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,1,power_law_1.2,10.405548858642579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,32,power_law_1.2,0.15639679431915282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,8,power_law_1.01,1.2375935554504394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,32,power_law_1.2,0.1510848045349121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,32,power_law_1.2,0.1750656008720398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,32,power_law_1.2,0.2089087963104248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,8,power_law_1.01,2.5947776794433595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,32,power_law_1.2,0.27832961082458496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,32,power_law_1.2,0.290720009803772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,32,power_law_1.2,0.36957440376281736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,32,power_law_1.2,0.37944960594177246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,4,1,power_law_1.2,15.644288635253906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,32,power_law_1.2,0.5066304206848145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,32,power_law_1.2,0.6398975849151611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,32,power_law_1.2,0.9907456398010254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,32,power_law_1.2,1.459059238433838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,64,power_law_1.01,0.04051840007305145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,64,power_law_1.01,0.03909119963645935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,32,power_law_1.2,1.8685504913330078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,64,power_law_1.01,0.033190399408340454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,64,power_law_1.01,0.03374719917774201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,64,power_law_1.01,0.03402239978313446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,64,power_law_1.01,0.033452799916267394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,64,power_law_1.01,0.03408640027046204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,32,power_law_1.2,2.8455615997314454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,16,power_law_1.01,0.044870400428771974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,64,power_law_1.01,0.03424000144004822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,16,power_law_1.01,0.04438399970531463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,64,power_law_1.01,0.034227201342582704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,16,power_law_1.01,0.029529601335525513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,64,power_law_1.01,0.035046398639678955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,16,power_law_1.01,0.030444800853729248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,64,power_law_1.01,0.03535360097885132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,16,power_law_1.01,0.030630400776863097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,64,power_law_1.01,0.03549439907073974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,16,power_law_1.01,0.03245440125465393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,64,power_law_1.01,0.036006399989128114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,16,power_law_1.01,0.032313600182533264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,64,power_law_1.01,0.036934399604797365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,1,balanced,0.10100799798965454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,1,balanced,0.10233066479365031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,1,balanced,0.11568533380826314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,1,balanced,0.1400320033232371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,1,balanced,0.19737066825230917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,1,balanced,0.2866400082906087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,32,power_law_1.2,6.880300903320313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,16,power_law_1.01,0.03304960131645203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,64,power_law_1.01,0.040633600950241086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,1,balanced,0.3768213192621867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,1,balanced,0.3811306556065877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,1,balanced,0.3848693370819092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,1,balanced,0.38049066066741943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,1,balanced,0.38790400822957355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,1,balanced,0.3939359982808431
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,16,power_law_1.01,0.03336319923400879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,64,power_law_1.01,0.04079999923706055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,1,balanced,0.3973173300425212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,1,balanced,0.4065920114517212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,1,balanced,0.41327468554178876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,1,balanced,0.42075733343760174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,1,balanced,0.4334506591161092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,16,power_law_1.01,0.03386879861354828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,64,power_law_1.01,0.04257279932498932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,1,balanced,0.48205868403116864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,1,balanced,0.5161813497543335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,1,balanced,0.5983893473943075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,16,power_law_1.01,0.03514240086078644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,1,balanced,0.6701707045237223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,64,power_law_1.01,0.04738560020923614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,1,balanced,0.8246133327484131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,1,balanced,0.995253324508667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,16,power_law_1.01,0.03609600067138672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,64,power_law_1.01,0.04938879907131195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,1,balanced,1.3381120363871257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,1,balanced,1.8137386639912922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,16,power_law_1.01,0.03591679930686951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,64,power_law_1.01,0.053491199016571046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,1,balanced,2.498576005299886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,16,power_law_1.01,0.03770880103111267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,64,power_law_1.01,0.06087039709091187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,1,balanced,3.3359734217325845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,16,power_law_1.01,0.042316800355911253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,64,power_law_1.01,0.07131519913673401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,16,power_law_1.01,0.04311679899692535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,1,balanced,4.170810699462891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,64,power_law_1.01,0.08099200129508972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,16,power_law_1.01,0.0440447986125946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,64,power_law_1.01,0.10593279600143432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,16,power_law_1.01,0.0486847996711731
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,1,balanced,6.422058741251628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,64,power_law_1.01,0.12412799596786499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,16,power_law_1.01,0.0493120014667511
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,16,power_law_1.01,0.056358402967453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,64,power_law_1.01,0.17599999904632568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,16,power_law_1.01,0.06601600050926208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,64,power_law_1.01,0.20997118949890137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,1,balanced,12.193946838378906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,16,power_law_1.01,0.08214399814605713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,16,power_law_1.01,0.10359040498733521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,2688,1856,6,128,1,64,power_law_1.01,0.28906879425048826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,16,power_law_1.01,0.13500800132751464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,16,power_law_1.01,0.15978879928588868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,2688,1856,6,128,1,64,power_law_1.01,0.3895103931427002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,16,power_law_1.01,0.22113919258117676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,16,balanced,0.05003733436266581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,16,balanced,0.05055999755859375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,16,balanced,0.05276800195376078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,16,balanced,0.05909866591294607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,16,balanced,0.07799466451009114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,16,balanced,0.08376000324885051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,16,balanced,0.08306133250395457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,16,balanced,0.08429867029190063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,16,balanced,0.08293866614500682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,16,power_law_1.01,0.27370240688323977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,16,balanced,0.08338133494059245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,16,balanced,0.0860693355401357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,16,balanced,0.08506666620572408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,16,balanced,0.08899733424186707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,2688,1856,6,128,1,64,power_law_1.01,0.8281791687011719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,16,balanced,0.08972799777984619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,16,balanced,0.09294933080673218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,16,balanced,0.0953386624654134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,16,balanced,0.09934399525324504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,16,balanced,0.1151626706123352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,16,balanced,0.11823466420173645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,16,balanced,0.14446933070818582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,16,balanced,0.16830933094024658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,2,16,power_law_1.01,0.338592004776001
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,16,balanced,0.22285866737365723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,16,balanced,0.20892800887425741
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,16,balanced,0.28569600979487103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,16,balanced,0.3266666730244954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,16,balanced,0.4503146807352702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,16,balanced,0.5397013425827026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,8,16,balanced,0.6696693102518717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,2,16,power_law_1.01,0.47482881546020506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,8,16,balanced,0.9660159746805826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,8,16,balanced,2.0084479649861655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,2,16,power_law_1.01,0.8662528038024903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,8,balanced,0.05217599868774414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,8,balanced,0.054197331269582115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,8,balanced,0.06033066908518473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,8,balanced,0.08358933528264363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,8,balanced,0.11556800206502278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,8,balanced,0.17529600858688354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,8,balanced,0.1737066706021627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,8,balanced,0.17828265825907388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,8,balanced,0.17714667320251465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,8,balanced,0.1781546672185262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,8,balanced,0.17828265825907388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,8,balanced,0.18017067511876425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,8,balanced,0.18027732769648233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,8,balanced,0.18487467368443808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,8,balanced,0.1914880077044169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,8,balanced,0.19049066305160522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,8,balanced,0.19749333461125693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,8,balanced,0.20505066712697348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,8,balanced,0.21388266483942667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,8,balanced,0.23212265968322754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,8,balanced,0.2560639977455139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,8,balanced,0.29425599177678424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,8,balanced,0.33505598704020184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,8,balanced,0.44276265303293866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,8,balanced,0.5087626775105795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,8,balanced,0.7144532998402914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,8,balanced,0.874842643737793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,8,balanced,1.071727991104126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,8,balanced,1.592410723368327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,8,balanced,2.9992106755574546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,64,power_law_1.01,0.030649599432945252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,64,power_law_1.01,0.02993279993534088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,64,power_law_1.01,0.030451199412345885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,64,power_law_1.01,0.03076480031013489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,64,power_law_1.01,0.030623999238014222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,64,power_law_1.01,0.031033599376678468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,64,power_law_1.01,0.031814399361610415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,64,power_law_1.01,0.03190400004386902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,16,power_law_1.01,0.15949440002441406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,1,power_law_1.2,0.09702399969100953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,64,power_law_1.01,0.03238399922847748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,16,power_law_1.01,0.15663360357284545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,64,power_law_1.01,0.03247359991073608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,16,power_law_1.01,0.09494400024414062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,64,power_law_1.01,0.033318400382995605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,64,power_law_1.01,0.036447998881340024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,16,power_law_1.01,0.11627520322799682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,1,power_law_1.2,0.14763519763946534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,64,power_law_1.01,0.03529599905014038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,16,power_law_1.01,0.1297279953956604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,64,power_law_1.01,0.03684479892253876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,16,power_law_1.01,0.1547711968421936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,64,power_law_1.01,0.04094719886779785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,1,power_law_1.2,0.1599295973777771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,16,power_law_1.01,0.1766144037246704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,64,power_law_1.01,0.04208639860153198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,16,power_law_1.01,0.17230080366134642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,64,power_law_1.01,0.044588801264762876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,64,power_law_1.01,0.04989440143108368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,16,power_law_1.01,0.18099839687347413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,1,power_law_1.2,0.2963648080825806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,16,balanced,0.05693333347638448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,64,power_law_1.01,0.05416960120201111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,16,balanced,0.06261866788069408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,16,power_law_1.01,0.17607680559158326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,16,balanced,0.06870933373769124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,16,balanced,0.09401599566141765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,16,balanced,0.11715733011563619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,16,balanced,0.1253493328889211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,16,balanced,0.12924266854921976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,16,balanced,0.1281599998474121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,16,balanced,0.1276693344116211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,64,power_law_1.01,0.06311039924621582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,16,balanced,0.13090667128562927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,16,balanced,0.1275040010611216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,16,balanced,0.13177067041397095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,16,power_law_1.01,0.17975679636001587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,16,balanced,0.1357439955075582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,16,balanced,0.1329759955406189
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,16,balanced,0.141375998655955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,16,balanced,0.14166399836540222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,16,balanced,0.14666133125623068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,64,power_law_1.01,0.07242879867553711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,16,balanced,0.16127467155456543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,16,balanced,0.17033600807189941
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,1,power_law_1.2,0.4855679988861084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,16,balanced,0.21473600467046103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,16,power_law_1.01,0.18531839847564696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,16,balanced,0.22365866104761759
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,16,balanced,0.30264000097910565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,16,balanced,0.2990986704826355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,16,balanced,0.43889065583546955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,64,power_law_1.01,0.10482560396194458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,16,balanced,0.4688906669616699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,16,power_law_1.01,0.1843135952949524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,16,balanced,0.6613813241322836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,16,balanced,0.7852266629536947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,64,power_law_1.01,0.13309439420700073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,1,16,balanced,0.966053326924642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,16,power_law_1.01,0.18774399757385254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,1,power_law_1.2,0.6448448181152344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,1,16,balanced,1.4366240501403809
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,64,power_law_1.01,0.1712000012397766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,16,power_law_1.01,0.19692800045013428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,1,16,balanced,2.921205202738444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,64,power_law_1.01,0.20238080024719238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,16,power_law_1.01,0.20490241050720215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,1,power_law_1.2,1.0365504264831542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,16,power_law_1.01,0.22106239795684815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,64,power_law_1.01,0.31002240180969237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,16,power_law_1.01,0.25024640560150146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,64,power_law_1.01,0.41208319664001464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,1,power_law_1.2,1.1310848236083983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,16,power_law_1.01,0.2650432109832764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,16,power_law_1.01,0.3309376001358032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,64,power_law_1.01,0.5038911819458007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,16,power_law_1.01,0.31559679508209226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,1,power_law_1.2,1.1971263885498047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,16,power_law_1.01,0.3737472057342529
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,64,power_law_1.01,0.7449920177459717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,1,power_law_1.2,1.2149888038635255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,16,power_law_1.01,0.44785280227661134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,16,power_law_1.01,0.5443840026855469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,1,power_law_1.2,1.2835455894470216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,64,power_law_1.01,1.4569791793823241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,16,power_law_1.01,0.7879231929779053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,1,power_law_1.2,1.3592063903808593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,16,power_law_1.01,1.192255973815918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,1,power_law_1.2,1.3566656112670898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,16,power_law_1.01,1.265120029449463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,1,8,balanced,0.07331199944019318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,1,8,balanced,0.09869866569836934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,1,8,balanced,0.0796800007422765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,1,8,balanced,0.07361066838105519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,1,8,balanced,0.073594664533933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,1,8,balanced,0.0735093355178833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,1,8,balanced,0.0741066684325536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,1,8,balanced,0.07300266623497009
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,1,8,balanced,0.07588799794514973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,1,8,balanced,0.07505600154399872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,1,8,balanced,0.0747519979874293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,1,8,balanced,0.07679466903209686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,1,8,balanced,0.07652799785137177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,1,power_law_1.2,1.414572811126709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,1,8,balanced,0.08140799899895985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,1,8,balanced,0.08496532837549846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,16,power_law_1.01,1.6456895828247071
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,1,8,balanced,0.08507200082143147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,1,8,balanced,0.08888000249862671
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,1,8,balanced,0.09691733121871948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,1,8,balanced,0.10687999924023946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,1,8,balanced,0.12045866250991821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,1,8,balanced,0.13935466607411703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,1,8,balanced,0.1689280072848002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,1,8,balanced,0.20801599820454916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,1,8,balanced,0.26708799600601196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,1,8,balanced,0.33824535210927326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,1,8,balanced,0.47460798422495526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,1,8,balanced,0.6094346841176351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,1,8,balanced,0.7749120394388834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,1,power_law_1.2,1.4515647888183594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,1,8,balanced,1.176415999730428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,1,8,balanced,2.255930741628011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,16,power_law_1.01,2.6460287094116213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,1,power_law_1.2,1.5107839584350586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,1,power_law_1.2,1.593132781982422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,16,power_law_1.01,4.710822296142578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,1,power_law_1.2,1.6968767166137695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,4,balanced,0.046122665206591286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,4,balanced,0.047839999198913574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,4,balanced,0.050111999114354454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,4,balanced,0.05599466462930044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,4,balanced,0.07710400223731995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,4,balanced,0.11038933197657268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,4,balanced,0.14447466532389322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,4,balanced,0.14081066846847534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,4,balanced,0.14101866881052652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,4,balanced,0.14229333400726318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,4,balanced,0.1458399991194407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,1,power_law_1.2,1.8044672012329102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,4,balanced,0.14781333009401956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,4,balanced,0.1502240002155304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,4,balanced,0.15165332953135172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,4,balanced,0.16012266278266907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,4,balanced,0.1625653306643168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,4,balanced,0.1742666761080424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,4,balanced,0.1857973337173462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,4,balanced,0.2004106640815735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,4,balanced,0.223855992158254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,4,balanced,0.24796799818674722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,4,balanced,0.30005866289138794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,4,balanced,0.36813334623972577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,4,balanced,0.47339733441670734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,4,balanced,0.5979093313217163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,4,balanced,0.8099253177642822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,1,power_law_1.2,2.011916732788086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,4,balanced,1.0461493333180745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,4,balanced,1.307487964630127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,4,balanced,1.9443732897440593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,4,balanced,3.7461067835489907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,1,power_law_1.2,2.0123071670532227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,1,power_law_1.2,2.2693759918212892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,32,power_law_1.01,0.04206719994544983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,1,power_law_1.2,2.614899253845215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,32,power_law_1.01,0.03656319975852966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,32,power_law_1.01,0.031769600510597226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,32,power_law_1.01,0.033344000577926636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,1,power_law_1.2,3.0553024291992186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,32,power_law_1.01,0.033030399680137636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,32,power_law_1.01,0.03728640079498291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,32,power_law_1.01,0.034969601035118106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,8,balanced,0.03133333226044973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,8,balanced,0.02956799914439519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,8,balanced,0.0320266659061114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,8,balanced,0.03136533250411352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,8,balanced,0.037231999138991036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,8,balanced,0.03533866753180822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,8,balanced,0.03571200122435888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,8,balanced,0.03565866748491923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,8,balanced,0.03568000098069509
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,8,balanced,0.037658666570981346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,32,power_law_1.01,0.036671999096870425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,8,balanced,0.035818666219711304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,8,balanced,0.037802666425704956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,8,balanced,0.03956266740957896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,8,balanced,0.03953066716591517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,8,balanced,0.04393066465854645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,8,balanced,0.043568000197410583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,1,power_law_1.2,3.5236927032470704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,8,balanced,0.04562133550643921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,8,balanced,0.04641066491603851
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,8,balanced,0.05211733281612396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,8,balanced,0.05648000041643778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,32,power_law_1.01,0.03951359987258911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,8,balanced,0.06307200094064076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,8,balanced,0.08285333216190338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,8,balanced,0.09924800197283427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,8,balanced,0.1341653366883596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,8,balanced,0.16910399993260702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,8,balanced,0.228277325630188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,8,balanced,0.2849973241488139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,32,power_law_1.01,0.039743998646736146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,8,balanced,0.351365327835083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,8,balanced,0.5380160013834635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,8,balanced,1.0469653606414795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,32,power_law_1.01,0.04399999976158142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,32,power_law_1.01,0.04531840085983276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,32,power_law_1.01,0.04683519899845123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,1,power_law_1.2,4.51781120300293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,32,power_law_1.01,0.048793599009513855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,32,power_law_1.01,0.053625601530075076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,4,32,power_law_1.01,0.036364799737930296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,32,power_law_1.01,0.05787519812583923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,32,power_law_1.01,0.06169599890708923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,4,32,power_law_1.01,0.032102400064468385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,4,32,power_law_1.01,0.032313600182533264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,32,power_law_1.01,0.06809599995613098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,1,power_law_1.2,5.5724609375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,4,32,power_law_1.01,0.03121280074119568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,32,power_law_1.01,0.07329279780387879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,4,32,power_law_1.01,0.03394559919834137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,32,power_law_1.01,0.0942143976688385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,4,32,power_law_1.01,0.03365119993686676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,32,power_law_1.01,0.10776959657669068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,4,32,power_law_1.01,0.03304319977760315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,32,power_law_1.01,0.12993919849395752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,4,32,power_law_1.01,0.033881598711013795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,32,power_law_1.01,0.1602944016456604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,4,32,power_law_1.01,0.03423359990119934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,1,power_law_1.2,6.5785667419433596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,4,32,power_law_1.01,0.03454079926013946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,32,power_law_1.01,0.19530880451202393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,4,32,power_law_1.01,0.03485440015792847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,32,power_law_1.01,0.27913599014282225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,4,32,power_law_1.01,0.03555839955806732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,4,32,power_law_1.01,0.036025598645210266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,32,power_law_1.01,0.34420480728149416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,4,32,power_law_1.01,0.03694080114364624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,4,32,power_law_1.01,0.03945600092411041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,32,power_law_1.01,0.48241281509399414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,4,32,power_law_1.01,0.03994239866733551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,8,balanced,0.05359466870625814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,8,balanced,0.05215999980767568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,8,balanced,0.053871999184290566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,8,balanced,0.06695466736952464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,8,balanced,0.08683199683825175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,8,balanced,0.0976639986038208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,1,power_law_1.2,9.499334716796875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,8,balanced,0.09937066833178203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,8,balanced,0.0983679989973704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,4,32,power_law_1.01,0.042124798893928526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,8,balanced,0.09852799773216248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,8,balanced,0.09923199812571208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,8,balanced,0.09894399841626485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,8,balanced,0.10043733318646748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,8,balanced,0.10346666971842448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,8,balanced,0.10371733705202739
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,8,balanced,0.10902399818102519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,8,balanced,0.11153067151705424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,32,power_law_1.01,0.6413375854492187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,8,balanced,0.11479999621709187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,4,32,power_law_1.01,0.04414080083370209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,8,balanced,0.12281066179275513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,8,balanced,0.13184533516565958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,8,balanced,0.16472533345222473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,8,balanced,0.16299733519554138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,8,balanced,0.20985599358876547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,8,balanced,0.23463465770085654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,8,balanced,0.3206933339436849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,4,32,power_law_1.01,0.04738560020923614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,8,balanced,0.37626131375630695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,8,balanced,0.5120906829833984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,8,balanced,0.6609813372294108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,4,32,power_law_1.01,0.051769602298736575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,8,balanced,0.7803786595662435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,8,balanced,1.2079253196716309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,32,power_law_1.01,0.9289600372314453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,4,32,power_law_1.01,0.056383997201919556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,8,balanced,2.2968799273173013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,4,32,power_law_1.01,0.06480000019073487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,4,32,power_law_1.01,0.07434880137443542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,4,32,power_law_1.01,0.10493439435958862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,32,power_law_1.01,1.8327680587768556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,4,32,power_law_1.01,0.1307711958885193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,4,32,power_law_1.01,0.173580801486969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,1,power_law_1.2,17.4503173828125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,4,32,power_law_1.01,0.23002240657806397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,4,32,power_law_1.01,0.24871039390563965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,4,32,power_law_1.01,0.43793277740478515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,4,32,power_law_1.01,0.7628928184509277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,2,power_law_1.2,0.05093119740486145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,8,balanced,0.044250667095184326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,8,balanced,0.04619200030962626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,8,balanced,0.046069333950678505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,2,power_law_1.2,0.05550720095634461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,8,balanced,0.04994666576385498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,8,balanced,0.058176000912984215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,8,balanced,0.06849599877993266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,8,balanced,0.06734933455785115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,8,balanced,0.07062399884064992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,8,balanced,0.06947733461856842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,8,balanced,0.07126933336257935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,8,balanced,0.07247999807198842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,8,balanced,0.07308266560236613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,8,balanced,0.07523199915885925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,8,balanced,0.08056533336639404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,8,balanced,0.08409066994984944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,8,balanced,0.08770133058230083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,8,balanced,0.09034132957458496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,8,balanced,0.10261866450309753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,8,balanced,0.1104906698067983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,8,balanced,0.13411200046539307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,8,balanced,0.14642666776974997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,8,balanced,0.1981226603190104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,2,power_law_1.2,0.06133120059967041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,8,balanced,0.2277173399925232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,8,balanced,0.3284800052642822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,8,balanced,0.39031465848286945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,8,balanced,0.5473066568374634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,8,balanced,0.7050720055898031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,8,balanced,0.8638453483581543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,8,balanced,1.3230613072713215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,2,power_law_1.2,0.06696320176124573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,8,balanced,2.565194606781006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,1,power_law_1.01,0.07127040028572082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,2,power_law_1.2,0.0917631983757019
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,2,power_law_1.2,0.11201920509338378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,1,power_law_1.01,0.08799359798431397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,2,power_law_1.2,0.1287935972213745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,2,balanced,0.040005333721637726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,2,balanced,0.04257066547870636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,2,balanced,0.04794133206208547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,2,balanced,0.06517333288987477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,2,balanced,0.08957866827646892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,2,balanced,0.11004799604415894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,2,balanced,0.1097866694132487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,2,balanced,0.10974933703740437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,2,balanced,0.11193066835403442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,2,balanced,0.11141866445541382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,2,balanced,0.11262933413187663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,2,balanced,0.11348799864451091
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,2,balanced,0.1137600044409434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,2,balanced,0.11844799915949504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,2,balanced,0.1218826671441396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,2,balanced,0.1225279966990153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,2,balanced,0.12844266494115195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,2,balanced,0.13849066694577536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,2,balanced,0.1495680014292399
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,2,power_law_1.2,0.16188160181045533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,2,balanced,0.19692800442377725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,1,power_law_1.01,0.10776959657669068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,2,balanced,0.19302932421366373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,2,balanced,0.26077866554260254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,2,balanced,0.29316266377766925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,2,balanced,0.41390399138132733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,2,balanced,0.48559999465942383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,2,balanced,0.6971840063730875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,2,balanced,0.8681759834289551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,2688,1856,6,128,1,2,balanced,1.069706678390503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,2,power_law_1.2,0.17058559656143188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,2688,1856,6,128,1,2,balanced,1.653183937072754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,2688,1856,6,128,1,2,balanced,3.2251628239949546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,2,power_law_1.2,0.17212159633636476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,2,power_law_1.2,0.18844799995422362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,1,power_law_1.01,0.15510400533676147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,2,power_law_1.2,0.19221760034561158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,128,power_law_1.01,0.039392000436782836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,2,power_law_1.2,0.20451200008392334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,128,power_law_1.01,0.03956480026245117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,1,power_law_1.01,0.21415040493011475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,128,power_law_1.01,0.04019840061664581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,2,power_law_1.2,0.21070079803466796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,128,power_law_1.01,0.03893119990825653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,128,power_law_1.01,0.04049279987812042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,128,power_law_1.01,0.04103679955005646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,2,power_law_1.2,0.2307904005050659
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,128,power_law_1.01,0.04132480025291443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,1,power_law_1.01,0.34238719940185547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,128,power_law_1.01,0.04041599929332733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,2,power_law_1.2,0.2348479986190796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,128,power_law_1.01,0.041145598888397215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,128,power_law_1.01,0.04126079976558685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,128,power_law_1.01,0.04202240109443665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,2,power_law_1.2,0.2586368083953857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,128,power_law_1.01,0.04248960018157959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,128,power_law_1.01,0.0429504007101059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,1,power_law_1.01,0.3688256025314331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,2,power_law_1.2,0.2931328058242798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,128,power_law_1.01,0.042694398760795595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,16,8,balanced,0.03957866628964742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,16,8,balanced,0.03956799954175949
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,16,8,balanced,0.03330666571855545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,16,8,balanced,0.031498665610949196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,16,8,balanced,0.03262399882078171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,16,8,balanced,0.03141333411137263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,16,8,balanced,0.033488000432650246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,128,power_law_1.01,0.046649599075317384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,16,8,balanced,0.03339199970165888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,16,8,balanced,0.03270400067170461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,16,8,balanced,0.03365333378314972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,16,8,balanced,0.03379733363787333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,16,8,balanced,0.033802665770053864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,16,8,balanced,0.035418666899204254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,16,8,balanced,0.0354666660229365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,16,8,balanced,0.040005333721637726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,16,8,balanced,0.04171200096607208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,16,8,balanced,0.03959999978542328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,16,8,balanced,0.044981335600217186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,16,8,balanced,0.043696001172065735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,128,power_law_1.01,0.04755840003490448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,16,8,balanced,0.04985066751639048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,2,power_law_1.2,0.30127999782562254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,16,8,balanced,0.05587733288606008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,16,8,balanced,0.07487466434637706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,16,8,balanced,0.08049599826335907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,16,8,balanced,0.10748266180356343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,16,8,balanced,0.11745599905649821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,16,8,balanced,0.15647466977437338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,16,8,balanced,0.19150932629903158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,128,power_law_1.01,0.048876801133155824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,16,8,balanced,0.23818665742874146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,16,8,balanced,0.3283413251241048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,16,8,balanced,0.5798559983571371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,1,power_law_1.01,0.3842560052871704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,128,power_law_1.01,0.05244799852371216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,2,power_law_1.2,0.36640639305114747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,128,power_law_1.01,0.05667200088500977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,128,power_law_1.01,0.06446719765663148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,2,power_law_1.2,0.4378687858581543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,128,power_law_1.01,0.07232000231742859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,1,power_law_1.01,0.39772160053253175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,128,power_law_1.01,0.08970239758491516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,2,power_law_1.2,0.532755184173584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,128,power_law_1.01,0.10702719688415527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,128,power_law_1.01,0.15516159534454346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,2,power_law_1.2,0.6220608234405518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,128,power_law_1.01,0.18049919605255127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,1,power_law_1.01,0.40381441116333006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,128,power_law_1.01,0.25031681060791017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,2,power_law_1.2,0.8392959594726562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,128,power_law_1.01,0.34689280986785886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,1,power_law_1.01,0.41703038215637206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,128,power_law_1.01,0.4493567943572998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,2,power_law_1.2,1.065113639831543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,128,power_law_1.01,0.6574848175048829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,1,power_law_1.01,0.4371967792510986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,2,power_law_1.2,1.447488021850586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,16,power_law_1.2,0.07240319848060608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,16,power_law_1.2,0.03727999925613403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,16,power_law_1.2,0.04377599954605103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,2,power_law_1.2,1.9647808074951172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,1,power_law_1.01,0.45194239616394044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,128,power_law_1.01,1.2527680397033691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,16,power_law_1.2,0.05364480018615723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,16,power_law_1.2,0.05619199872016907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,16,power_law_1.2,0.0730239987373352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,2,power_law_1.2,2.2611967086791993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,1,power_law_1.01,0.457260799407959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,16,power_law_1.2,0.07279999852180481
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,16,power_law_1.2,0.07454079985618592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,16,power_law_1.2,0.07559040188789368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,1,power_law_1.01,0.48271999359130857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,16,power_law_1.2,0.07545599937438965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,2,power_law_1.2,3.473286437988281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,16,power_law_1.2,0.07864320278167725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,16,power_law_1.2,0.08364160060882568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,16,power_law_1.2,0.08680319786071777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,1,power_law_1.01,0.5046207904815674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,16,power_law_1.2,0.08441600203514099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,16,power_law_1.2,0.08887680172920227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,16,power_law_1.2,0.0930624008178711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,16,power_law_1.2,0.09607679843902588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,16,power_law_1.2,0.11204479932785034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,1,power_law_1.01,0.5345215797424316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,16,power_law_1.2,0.1229248046875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,2,power_law_1.2,6.765068817138672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,16,power_law_1.2,0.15599360466003417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,16,power_law_1.2,0.14727040529251098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,1,power_law_1.01,0.6193727970123291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,16,power_law_1.2,0.19343359470367433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,16,power_law_1.2,0.23614718914031982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,16,power_law_1.2,0.325926399230957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,2,power_law_1.01,0.24391040802001954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,1,power_law_1.01,0.6703872203826904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,16,power_law_1.2,0.4295360088348389
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,2,power_law_1.01,0.31738240718841554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,16,power_law_1.2,0.6341695785522461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,16,power_law_1.2,0.8606143951416015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,1,power_law_1.01,0.8282367706298828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,2,power_law_1.01,0.23386878967285157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,16,power_law_1.2,1.1025279998779296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,2,power_law_1.01,0.4311103820800781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,1,power_law_1.01,1.024396800994873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,16,power_law_1.2,1.8441984176635742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,2,power_law_1.01,0.6322368144989013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,1,power_law_1.01,1.296895980834961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,1,power_law_1.01,0.0773248016834259
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,2,power_law_1.01,1.1637120246887207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,16,power_law_1.2,3.230963134765625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,1,power_law_1.01,0.0827072024345398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,2,power_law_1.01,1.2546879768371582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,1,power_law_1.01,0.09109119772911071
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,1,power_law_1.01,1.6040000915527344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,1,power_law_1.01,0.11266560554504394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,2,power_law_1.01,1.29169921875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,1,power_law_1.01,0.14480639696121217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,1,power_law_1.01,0.16789120435714722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,2,power_law_1.01,1.3538687705993653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,1,power_law_1.01,2.1652544021606444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,1,power_law_1.01,0.21361920833587647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,1,power_law_1.01,0.22702720165252685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,2,power_law_1.01,1.3518272399902345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,1,power_law_1.01,0.22810239791870118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,1,power_law_1.01,0.23546879291534423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,2,power_law_1.01,1.41210880279541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,1,power_law_1.01,2.584351921081543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,1,power_law_1.01,0.24170238971710206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,1,power_law_1.01,0.25372159481048584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,2,power_law_1.01,1.4254400253295898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,1,power_law_1.01,0.25791358947753906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,32,power_law_1.2,0.047865599393844604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,1,power_law_1.01,0.2678719997406006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,2,power_law_1.01,1.470809555053711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,1,power_law_1.01,3.685830307006836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,32,power_law_1.2,0.05275520086288452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,1,power_law_1.01,0.27945599555969236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,32,power_law_1.2,0.04310399889945984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,1,power_law_1.01,0.28482561111450194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,2,power_law_1.01,1.5351231575012207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,32,power_law_1.2,0.042208001017570496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,1,power_law_1.01,0.3079871892929077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,32,power_law_1.2,0.04428800046443939
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,1,power_law_1.01,0.3549312114715576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,2,power_law_1.01,1.5703743934631347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,32,power_law_1.2,0.04609920084476471
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,1,power_law_1.01,0.39366400241851807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,1,power_law_1.01,4.817606353759766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,32,power_law_1.2,0.04710400104522705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,2,power_law_1.01,1.6161279678344727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,1,power_law_1.01,0.4835072040557861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,32,power_law_1.2,0.04935680031776428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,1,power_law_1.01,0.5616064071655273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,32,power_law_1.2,0.04906879961490631
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,2,power_law_1.01,1.654649543762207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,1,power_law_1.01,0.7433087825775146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,32,power_law_1.2,0.049439999461174014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,32,power_law_1.2,0.051020801067352295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,1,power_law_1.01,0.9100607872009278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,2,power_law_1.01,1.8353279113769532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,32,power_law_1.2,0.0515392005443573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,8,1,power_law_1.01,5.8893695831298825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,1,power_law_1.01,1.2607040405273438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,32,power_law_1.2,0.05419520139694214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,32,power_law_1.2,0.056550401449203494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,2,power_law_1.01,1.900716781616211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,1,power_law_1.01,1.603321647644043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,32,power_law_1.2,0.06264320015907288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,32,power_law_1.2,0.06551039814949036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,1,power_law_1.01,2.2981695175170898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,2,power_law_1.01,2.190553665161133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,32,power_law_1.2,0.07073280215263367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,32,power_law_1.2,0.08168960213661194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,1,power_law_1.01,3.00710391998291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,2,power_law_1.01,2.245996856689453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,32,power_law_1.2,0.08894079923629761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,8,1,power_law_1.01,8.981740570068359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,32,power_law_1.2,0.11923840045928955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,1,power_law_1.01,3.688691329956055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,32,power_law_1.2,0.14591360092163086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,2,power_law_1.01,2.6544000625610353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,32,power_law_1.2,0.19889919757843016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,32,power_law_1.2,0.2321471929550171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,2,power_law_1.01,2.7528831481933596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,1,power_law_1.01,5.716966247558593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,32,power_law_1.2,0.32857599258422854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,32,power_law_1.2,0.42058238983154295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,2,power_law_1.01,3.440428924560547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,32,power_law_1.2,0.6112512111663818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,32,power_law_1.2,0.7785600185394287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,1,power_law_1.01,11.136934661865235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,2,power_law_1.01,4.481913757324219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,32,power_law_1.2,1.0138303756713867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,8,1,power_law_1.01,16.93953857421875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,32,power_law_1.2,1.5290687561035157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,2,power_law_1.01,5.802956771850586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,32,power_law_1.2,3.3129150390625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,2,power_law_1.01,6.4870147705078125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,8,1,power_law_1.01,0.042822399735450746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,8,1,power_law_1.01,0.04959999918937683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,1,2,power_law_1.01,7.841382598876953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,8,1,power_law_1.01,0.05045120120048523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,8,1,power_law_1.01,0.06421120166778564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,8,1,power_law_1.01,0.06488959789276123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,8,1,power_law_1.01,0.06525440216064453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,8,1,power_law_1.01,0.06629120111465454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,2,power_law_1.2,0.07084800004959106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,8,1,power_law_1.01,0.06702719926834107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,8,1,power_law_1.01,0.06677759885787964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,8,1,power_law_1.01,0.06677759885787964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,8,1,power_law_1.01,0.06924800276756286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,2,power_law_1.2,0.074508798122406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,8,1,power_law_1.01,0.07301759719848633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,8,1,power_law_1.01,0.07522559762001038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,8,1,power_law_1.01,0.07893760204315185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,16,balanced,0.05201066533724467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,16,balanced,0.04342933495839437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,16,balanced,0.0436106671889623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,16,balanced,0.04563733438650767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,8,1,power_law_1.01,0.08780159950256347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,16,balanced,0.047983999053637184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,16,balanced,0.05388266841570536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,16,balanced,0.05813866853713989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,16,balanced,0.05397866666316986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,16,balanced,0.056048000852266945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,8,1,power_law_1.01,0.08893439769744874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,16,balanced,0.056101332108179726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,16,balanced,0.056176001826922096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,2,power_law_1.2,0.07785599827766418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,16,balanced,0.058058664202690125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,16,balanced,0.0601440022389094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,16,balanced,0.06163733204205831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,16,balanced,0.06444799900054932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,16,balanced,0.06837333242098491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,8,1,power_law_1.01,0.096697598695755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,16,balanced,0.07057066758473714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,16,balanced,0.0798773318529129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,16,balanced,0.08474133412043254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,16,balanced,0.09546666344006856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,16,balanced,0.11508267124493916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,8,1,power_law_1.01,0.10800000429153442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,16,balanced,0.1357493301232656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,16,balanced,0.15405333042144775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,16,balanced,0.19765865802764893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,16,balanced,0.24714134136835733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,8,1,power_law_1.01,0.12295680046081543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,16,balanced,0.34440000851949054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,16,balanced,0.4222346544265747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,8,1,power_law_1.01,0.15215359926223754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,2,power_law_1.2,0.08904320001602173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,16,balanced,0.5237919886906942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,16,balanced,0.7703839937845866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,8,1,power_law_1.01,0.17311999797821045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,16,balanced,1.4972693125406902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,8,1,power_law_1.01,0.22836480140686036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,1,2,power_law_1.01,11.753298950195312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,8,1,power_law_1.01,0.2749119997024536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,2,power_law_1.2,0.1048192024230957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,8,1,power_law_1.01,0.3810816049575806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,8,1,power_law_1.01,0.4806528091430664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,8,1,power_law_1.01,0.6834112167358398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,2,power_law_1.2,0.12179839611053467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,8,1,power_law_1.01,0.8839551925659179
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,8,1,power_law_1.01,1.0781567573547364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,2,power_law_1.2,0.1343999981880188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,8,1,power_law_1.01,1.6814464569091796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,2,power_law_1.2,0.15712000131607057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,8,1,power_law_1.01,3.2701953887939452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,2,power_law_1.2,0.16681599617004395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,2,power_law_1.2,0.1714560031890869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,2,power_law_1.2,0.17740800380706787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,2,power_law_1.2,0.18435200452804565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,1,2,power_law_1.01,20.513343811035156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,2,1,power_law_1.01,0.06593919992446899
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,2,power_law_1.2,0.18775039911270142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,2,1,power_law_1.01,0.07241600155830383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,2,1,power_law_1.01,0.08934400081634522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,2,1,power_law_1.01,0.13747199773788452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,2,1,power_law_1.01,0.13647359609603882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,2,1,power_law_1.01,0.14142080545425414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,2,1,power_law_1.01,0.14295040369033812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,2,power_law_1.2,0.19895039796829223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,2,1,power_law_1.01,0.14426239728927612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,2,1,power_law_1.01,0.1422719955444336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,2,1,power_law_1.01,0.14507520198822021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,2,1,power_law_1.01,0.15125759840011596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,2,1,power_law_1.01,0.1552191972732544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,2,1,power_law_1.01,0.15523840188980104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,2,power_law_1.2,0.20698239803314208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,2,1,power_law_1.01,0.15935360193252562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,2,1,power_law_1.01,0.17831679582595825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,2,1,power_law_1.01,0.18336000442504882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,2,1,power_law_1.01,0.19676799774169923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,2,1,power_law_1.01,0.22922880649566652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,2,1,power_law_1.01,0.2557440042495728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,2,power_law_1.2,0.21383039951324462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,2,1,power_law_1.01,0.32354559898376467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,2,1,power_law_1.01,0.39285759925842284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,8,balanced,0.033039999504884086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,8,balanced,0.032831999162832894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,2,1,power_law_1.01,0.5140416145324707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,8,balanced,0.0339626669883728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,8,balanced,0.03772266705830892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,8,balanced,0.04976533353328705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,2,1,power_law_1.01,0.6152448177337646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,8,balanced,0.059088001648585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,8,balanced,0.05946666498978933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,8,balanced,0.060496002435684204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,8,balanced,0.05961066484451294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,2,power_law_1.2,0.2270143985748291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,2,1,power_law_1.01,0.8501631736755371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,8,balanced,0.06001600126425425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,8,balanced,0.0595413347085317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,8,balanced,0.05909866591294607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,8,balanced,0.06180266539255778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,8,balanced,0.06198399762312571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,8,balanced,0.06634133557478587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,2,1,power_law_1.01,1.0644991874694825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,8,balanced,0.07055466870466869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,8,balanced,0.06950933237870534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,8,balanced,0.07564799984296162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,8,balanced,0.07526400188604991
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,8,balanced,0.08353066444396973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,8,balanced,0.0899679958820343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,2,1,power_law_1.01,1.5226688385009766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,8,balanced,0.10626666744550069
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,2,power_law_1.2,0.2560447931289673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,8,balanced,0.1157866617043813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,8,balanced,0.16368533174196878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,8,balanced,0.18002132574717203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,8,balanced,0.2480319937070211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,8,balanced,0.277893324693044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,2,1,power_law_1.01,1.9713920593261718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,1,8,balanced,0.3379733165105184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,1,8,balanced,0.4989173412322998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,1,8,balanced,0.9192480246225992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,2,power_law_1.2,0.28286080360412597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,2,1,power_law_1.01,2.437977600097656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,2,1,power_law_1.01,3.783366394042969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,2,power_law_1.2,0.3440448045730591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,2,power_law_1.2,0.40266880989074705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,2,1,power_law_1.01,7.608608245849609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,4,power_law_1.01,0.04808320105075836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,4,power_law_1.01,0.04962559938430786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,2,power_law_1.2,0.5090303897857666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,4,power_law_1.01,0.048844799399375916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,4,power_law_1.01,0.06307200193405152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,4,power_law_1.01,0.06825600266456604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,2,power_law_1.2,0.6340672016143799
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,4,power_law_1.01,0.08533120155334473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,4,power_law_1.01,0.08817920088768005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,2,power_law_1.2,0.8265791893005371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,4,power_law_1.01,0.09047039747238159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,4,power_law_1.01,0.09267200231552124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,4,power_law_1.01,0.09395840167999267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,4,power_law_1.01,0.09591040015220642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,2,power_law_1.2,1.1043456077575684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,4,power_law_1.01,0.09685760140419006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,4,power_law_1.01,0.10089600086212158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,4,power_law_1.01,0.10597759485244751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,4,4,power_law_1.2,0.06251519918441772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,2,power_law_1.2,1.5006848335266114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,4,power_law_1.01,0.11037440299987793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,4,4,power_law_1.2,0.058297598361968996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,4,power_law_1.01,0.1126911997795105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,4,4,power_law_1.2,0.05628799796104431
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,4,4,power_law_1.2,0.05811840295791626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,4,power_law_1.01,0.12363519668579101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,4,4,power_law_1.2,0.06042879819869995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,4,power_law_1.01,0.14168959856033325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,4,4,power_law_1.2,0.0607807993888855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,2,power_law_1.2,1.8918975830078124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,4,4,power_law_1.2,0.06470400094985962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,4,power_law_1.01,0.14922239780426025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,4,4,power_law_1.2,0.06259840130805969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,4,4,power_law_1.2,0.06408320069313049
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,4,power_law_1.01,0.186572802066803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,32,2,power_law_1.01,0.03208959996700287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,4,4,power_law_1.2,0.0659775972366333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,32,2,power_law_1.01,0.03155199885368347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,4,power_law_1.01,0.21646080017089844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,32,2,power_law_1.01,0.03318400084972382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,4,4,power_law_1.2,0.06765440106391907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,32,2,power_law_1.01,0.0336896002292633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,32,2,power_law_1.01,0.03377279937267304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,4,4,power_law_1.2,0.06755840182304382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,32,2,power_law_1.01,0.03359360098838806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,4,power_law_1.01,0.2559936046600342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,2,power_law_1.2,2.3646015167236327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,32,2,power_law_1.01,0.03463039994239807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,4,4,power_law_1.2,0.06832640171051026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,32,2,power_law_1.01,0.034720000624656674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,4,4,power_law_1.2,0.07631999850273133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,32,2,power_law_1.01,0.03821440041065216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,32,2,power_law_1.01,0.038073599338531494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,4,power_law_1.01,0.3164736032485962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,4,4,power_law_1.2,0.08469759821891784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,32,2,power_law_1.01,0.03971199989318848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,32,2,power_law_1.01,0.04245119988918304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,4,4,power_law_1.2,0.09777920246124268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,32,2,power_law_1.01,0.04175359904766083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,32,2,power_law_1.01,0.04398080110549927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,4,4,power_law_1.2,0.10104320049285889
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,32,2,power_law_1.01,0.0532800018787384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,4,power_law_1.01,0.43263998031616213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,32,2,power_law_1.01,0.057171201705932616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,4,4,power_law_1.2,0.11569279432296753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,32,2,power_law_1.01,0.05872640013694763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,32,2,power_law_1.01,0.07800319790840149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,4,4,power_law_1.2,0.13326079845428468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,32,2,power_law_1.01,0.07859839797019959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,4,power_law_1.01,0.5448832035064697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,32,2,power_law_1.01,0.10163840055465698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,4,4,power_law_1.2,0.16474239826202391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,32,2,power_law_1.01,0.1158784031867981
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,4,4,power_law_1.2,0.16536320447921754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,32,2,power_law_1.01,0.153331196308136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,2,power_law_1.2,4.170508956909179
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,32,2,power_law_1.01,0.17815040349960326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,4,4,power_law_1.2,0.24074239730834962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,32,2,power_law_1.01,0.25744640827178955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,4,power_law_1.01,0.773311996459961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,32,2,power_law_1.01,0.29903359413146974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,4,4,power_law_1.2,0.344870400428772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,32,2,power_law_1.01,0.4103104114532471
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,4,4,power_law_1.2,0.398470401763916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,32,2,power_law_1.01,0.612499189376831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,4,power_law_1.01,0.9615936279296875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,4,4,power_law_1.2,0.5713471889495849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,32,2,power_law_1.01,0.7641088008880615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,4,4,power_law_1.2,0.9084992408752441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,32,2,power_law_1.01,1.1148799896240233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,4,4,power_law_1.2,0.9918144226074219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,4,power_law_1.01,1.1842559814453124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,4,4,power_law_1.2,1.255123233795166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,32,2,power_law_1.01,2.0525888442993163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,4,4,power_law_1.2,1.9196672439575195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,4,power_law_1.01,1.9228096008300781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,2,power_law_1.2,7.478880310058594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,4,4,power_law_1.2,5.161324691772461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,4,power_law_1.01,3.641632080078125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,4,balanced,0.04381866753101349
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,4,balanced,0.04380266865094503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,4,balanced,0.04765866696834564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,4,balanced,0.056501333912213646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,4,balanced,0.08011733492215474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,4,balanced,0.09035733342170715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,4,balanced,0.09293333689371745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,4,balanced,0.09403733412424724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,4,balanced,0.09078400333722432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,4,balanced,0.09490133325258891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,4,balanced,0.09357333183288574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,4,balanced,0.0965173343817393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,4,balanced,0.09865599870681763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,4,balanced,0.10132799545923869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,4,balanced,0.10756267110506694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,4,balanced,0.11043733358383179
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,4,balanced,0.11772800485293071
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,4,balanced,0.13467199603716531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,4,balanced,0.14155733585357666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,4,balanced,0.18036266167958578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,16,balanced,0.05620799958705902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,4,balanced,0.20334933201471964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,16,balanced,0.04649066428343455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,4,balanced,0.2759840091069539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,16,balanced,0.04632000128428141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,16,balanced,0.04975466430187225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,4,balanced,0.3123573263486226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,16,balanced,0.049914668003718056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,4,balanced,0.44809067249298096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,16,balanced,0.05561066667238871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,16,balanced,0.07214933137098949
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,4,balanced,0.5562826792399088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,16,balanced,0.07364266614119212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,16,balanced,0.07439466814200084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,4,balanced,0.7780746618906657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,16,balanced,0.07522133489449818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,16,balanced,0.07601599891980489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,16,balanced,0.07648000121116638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,4,balanced,1.0148800214131672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,16,balanced,0.07737599809964497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,16,balanced,0.07940266529719035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,128,power_law_1.01,0.058246397972106935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,16,balanced,0.08458667000134786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,4,balanced,1.222101370493571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,16,balanced,0.08447466293970744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,16,balanced,0.0905013382434845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,16,balanced,0.09470933675765991
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,16,balanced,0.10238400101661682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,16,balanced,0.11690666278203328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,4,balanced,1.9114559491475422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,16,balanced,0.12917866309483847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,16,balanced,0.15447466572125754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,16,balanced,0.19176000356674194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,16,balanced,0.24664000670115152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,16,balanced,0.30236266056696576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,4,balanced,3.7211198806762695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,16,balanced,0.42556265989939374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,16,balanced,0.5275573333104452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,16,balanced,0.6546239852905273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,128,power_law_1.01,0.05514879822731018
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,16,balanced,0.9801706473032633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,16,balanced,1.9018826484680176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,4,power_law_1.01,0.11184639930725097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,128,power_law_1.01,0.05065600275993347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,4,power_law_1.01,0.14907519817352294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,128,power_law_1.01,0.054655998945236206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,4,power_law_1.01,0.09800320267677307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,128,power_law_1.01,0.05532159805297852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,4,power_law_1.01,0.1299008011817932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,128,power_law_1.01,0.057120001316070555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,4,power_law_1.01,0.17461760044097902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,128,power_law_1.01,0.055852800607681274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,4,power_law_1.01,0.24907519817352294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,128,power_law_1.01,0.05560960173606873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,32,balanced,0.03140799949566523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,32,balanced,0.029535998900731403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,32,balanced,0.03141333411137263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,32,balanced,0.03130666663249334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,32,balanced,0.03324266771475474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,32,balanced,0.03483733286460241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,32,balanced,0.033386667569478355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,32,balanced,0.03545066714286804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,32,balanced,0.03365866591533025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,32,balanced,0.03369066615899404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,32,balanced,0.03571200122435888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,32,balanced,0.03559466699759165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,4,power_law_1.01,0.26702079772949217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,32,balanced,0.03568533311287562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,32,balanced,0.03774933268626531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,32,balanced,0.040207999447981514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,32,balanced,0.041536000867684685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,32,balanced,0.0436160018046697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,32,balanced,0.04373333354791006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,32,balanced,0.0462666650613149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,32,balanced,0.05392000079154968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,32,balanced,0.06412266691525777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,32,balanced,0.0823466678460439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,128,power_law_1.01,0.05606399774551392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,32,balanced,0.10106666882832845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,32,balanced,0.1328213314215342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,32,balanced,0.1643786629041036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,32,balanced,0.22196799516677856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,32,balanced,0.27450132369995117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,32,balanced,0.33179734150568646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,32,balanced,0.48606932163238525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,32,balanced,0.9238666693369547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,32,2,power_law_1.2,0.04684799909591675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,4,power_law_1.01,0.28545920848846434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,128,power_law_1.01,0.06325759887695312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,32,2,power_law_1.2,0.05128960013389587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,4,power_law_1.01,0.2888511896133423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,128,power_law_1.01,0.06616960167884826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,32,2,power_law_1.2,0.050425601005554196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,4,power_law_1.01,0.2987967967987061
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,32,2,power_law_1.2,0.058432000875473025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,128,power_law_1.01,0.06855679750442505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,4,power_law_1.01,0.3010495901107788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,32,2,power_law_1.2,0.06647040247917176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,128,power_law_1.01,0.07162240147590637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,32,2,power_law_1.2,0.07833600044250488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,4,power_law_1.01,0.31498239040374754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,128,power_law_1.01,0.07322239875793457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,32,2,power_law_1.2,0.09882879853248597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,4,power_law_1.01,0.3178816080093384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,128,power_law_1.01,0.08149120211601257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,32,2,power_law_1.2,0.10212479829788208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,4,power_law_1.01,0.3268480062484741
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,32,2,power_law_1.2,0.10862720012664795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,128,power_law_1.01,0.08468480110168457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,32,1,balanced,0.05592533449331919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,32,1,balanced,0.059792002042134605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,32,1,balanced,0.06411733229955037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,32,1,balanced,0.08310933411121368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,32,1,balanced,0.12421866257985432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,32,1,balanced,0.1772586703300476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,32,1,balanced,0.23598400751749674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,32,1,balanced,0.236842672030131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,32,2,power_law_1.2,0.10915199518203736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,4,power_law_1.01,0.33369600772857666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,32,1,balanced,0.23925334215164185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,32,1,balanced,0.24244266748428345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,32,1,balanced,0.24637333552042642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,32,1,balanced,0.2529386679331462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,32,1,balanced,0.2563146750132243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,32,1,balanced,0.26359466711680096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,32,1,balanced,0.2733866572380066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,32,1,balanced,0.28066132465998334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,128,power_law_1.01,0.08902400135993957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,32,1,balanced,0.29921066761016846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,32,1,balanced,0.3497813145319621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,32,1,balanced,0.3832159837086995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,32,1,balanced,0.4586293299992879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,32,1,balanced,0.5265546639760336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,32,2,power_law_1.2,0.1152959942817688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,32,1,balanced,0.6686826546986898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,32,1,balanced,0.8592320283253988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,4,power_law_1.01,0.34796159267425536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,32,1,balanced,1.1486132939656575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,32,1,balanced,1.5812746683756511
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,128,power_law_1.01,0.10756479501724243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,32,2,power_law_1.2,0.118777596950531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,32,1,balanced,2.2594186464945474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,32,1,balanced,2.9342826207478843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,4,power_law_1.01,0.3641088008880615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,32,2,power_law_1.2,0.12521599531173705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,32,1,balanced,3.685296058654785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,128,power_law_1.01,0.11119999885559081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,32,1,balanced,5.66978136698405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,32,2,power_law_1.2,0.13590400218963622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,4,power_law_1.01,0.4095295906066895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,128,power_law_1.01,0.13684480190277098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,32,2,power_law_1.2,0.1466752052307129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,32,1,balanced,10.981082916259766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,4,power_law_1.01,0.45487360954284667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,32,2,power_law_1.2,0.15578240156173706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,128,power_law_1.01,0.15485440492630004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,4,power_law_1.01,0.542464017868042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,32,2,power_law_1.2,0.15873279571533203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,128,power_law_1.01,0.1976639986038208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,32,2,power_law_1.2,0.1821951985359192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,4,power_law_1.01,0.6357952117919922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,32,2,power_law_1.2,0.21378560066223146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,128,power_law_1.01,0.21989119052886963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,4,power_law_1.01,0.7826752185821533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,32,2,power_law_1.2,0.26718080043792725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,128,power_law_1.01,0.29005439281463624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,32,2,power_law_1.2,0.3146239995956421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,4,power_law_1.01,0.884665584564209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,32,2,power_law_1.2,0.4289087772369385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,128,power_law_1.01,0.3417344093322754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,32,2,power_law_1.2,0.5257535934448242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,4,power_law_1.01,1.1758336067199706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,2,2,balanced,0.04774933556715647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,2,2,balanced,0.06819200019041698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,2,2,balanced,0.09444800019264221
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,2,2,balanced,0.09462933739026387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,2,2,balanced,0.09457066655158997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,2,2,balanced,0.0962506632010142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,2,2,balanced,0.09482666850090027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,2,2,balanced,0.09749866525332133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,2,2,balanced,0.09831466277440389
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,2,2,balanced,0.09750399986902873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,2,2,balanced,0.09787733356157939
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,2,2,balanced,0.09946133693059285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,2,2,balanced,0.10073066751162212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,32,2,power_law_1.2,0.7395264148712158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,2,2,balanced,0.10429333647092183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,2,2,balanced,0.11133333047231038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,2,2,balanced,0.11345600088437398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,2,2,balanced,0.11897599697113037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,2,2,balanced,0.13573867082595825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,2,2,balanced,0.14843199650446573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,128,power_law_1.01,0.49147520065307615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,2,2,balanced,0.1751520037651062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,2,2,balanced,0.208021342754364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,2,2,balanced,0.2804853320121765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,2,2,balanced,0.3250986735026042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,2,2,balanced,0.45054932435353595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,4,power_law_1.01,1.4534720420837401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,2,2,balanced,0.5737653175989786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,2,2,balanced,0.8168906370798746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,2,2,balanced,1.0590559641520183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,2,2,balanced,1.2972479661305745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,32,2,power_law_1.2,0.9363519668579101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,2,2,balanced,2.0195466677347818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,2,2,balanced,3.9872213999430337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,128,power_law_1.01,0.5733376026153565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,32,2,power_law_1.2,1.3214719772338868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,4,power_law_1.01,1.9869184494018555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,32,2,power_law_1.2,1.7706880569458008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,1,128,power_law_1.01,0.7325439929962159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,4,power_law_1.01,2.5809791564941404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,32,2,power_law_1.2,2.1905855178833007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,32,2,power_law_1.2,3.5386688232421877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,1,4,power_law_1.01,3.33359375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,1,128,power_law_1.01,1.1141695976257324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,1,2,balanced,0.05794133245944977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,1,2,balanced,0.09058666229248047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,1,2,balanced,0.10409067074457805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,1,2,balanced,0.10152000188827515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,1,2,balanced,0.1032533347606659
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,1,2,balanced,0.1027733286221822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,1,2,balanced,0.10419733325640361
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,1,2,balanced,0.10529067118962605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,1,2,balanced,0.10641066233317058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,1,2,balanced,0.10591466228167216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,1,2,balanced,0.10735467076301575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,1,2,balanced,0.10789866248766582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,1,2,balanced,0.10985599954922994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,1,2,balanced,0.11738666892051697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,1,2,balanced,0.12020267049471538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,1,2,balanced,0.12500799695650736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,1,2,balanced,0.1279306709766388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,1,2,balanced,0.15065600474675497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,1,2,balanced,0.16454399625460306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,1,2,balanced,0.20174400011698404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,1,2,balanced,0.24675732851028442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,32,2,power_law_1.2,6.516671752929687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,1,2,balanced,0.32662399609883624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,1,2,balanced,0.4086079994837443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,1,2,balanced,0.5612533489863077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,1,2,balanced,0.7286026477813721
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,1,2,balanced,1.0539360046386719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,1,2,balanced,1.3698827425638835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,1,4,power_law_1.01,4.918259048461914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,1,2,balanced,1.7546826998392742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,1,2,balanced,2.583834648132324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,1,2,balanced,5.0554507573445635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,1,128,power_law_1.01,2.3010623931884764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,1,4,power_law_1.01,9.197151947021485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,8,balanced,0.04784533381462097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,8,balanced,0.04885333279768626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,8,balanced,0.05145066479841868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,8,balanced,0.0738613357146581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,8,balanced,0.1025279959042867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,8,balanced,0.14962133765220642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,8,balanced,0.2020053267478943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,8,balanced,0.20397865772247314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,8,balanced,0.204693337281545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,8,balanced,0.20403732856114706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,8,balanced,0.20812799533208212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,8,balanced,0.20768000682195029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,8,balanced,0.2099519968032837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,8,balanced,0.21224000056584677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,8,balanced,0.2193440000216166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,8,balanced,0.22206399838129678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,8,balanced,0.22965333859125772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,8,balanced,0.2421813408533732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,8,balanced,0.254314661026001
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,8,balanced,0.2822293241818746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,8,balanced,0.30404265721638996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,8,balanced,0.35929067929585773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,8,balanced,0.42021334171295166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,8,balanced,0.5223733186721802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,8,balanced,0.6447626749674479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,8,balanced,0.8588266372680664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,8,balanced,1.0887893040974934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,8,balanced,1.3445547421773274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,8,balanced,1.9962293306986492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,8,power_law_1.2,0.06883839964866638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,8,balanced,3.8709119160970054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,2,power_law_1.2,0.07166079878807068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,8,power_law_1.2,0.10875519514083862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,2,power_law_1.2,0.08335360288619995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,2,power_law_1.2,0.0970304012298584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,8,power_law_1.2,0.06042240262031555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,2,power_law_1.2,0.12285439968109131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,8,power_law_1.2,0.06670079827308655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,2,power_law_1.2,0.16483839750289916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,2,power_law_1.2,0.3140928030014038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,8,power_law_1.2,0.08235520124435425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,2,power_law_1.2,0.3253312110900879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,2,power_law_1.2,0.3502975940704346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,8,power_law_1.2,0.09765759706497193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,2,power_law_1.2,0.34891519546508787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,8,power_law_1.2,0.12113280296325683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,2,power_law_1.2,0.3566783905029297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,32,power_law_1.01,0.08522239923477173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,2,power_law_1.2,0.3600064039230347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,8,power_law_1.2,0.1256127953529358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,32,power_law_1.01,0.08265600204467774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,2,power_law_1.2,0.36803839206695554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,32,power_law_1.01,0.05687040090560913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,8,power_law_1.2,0.12984319925308227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,2,power_law_1.2,0.3798975944519043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,32,power_law_1.01,0.06158080101013184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,2,power_law_1.2,0.39626240730285645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,8,power_law_1.2,0.13251839876174926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,32,power_law_1.01,0.06405119895935059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,2,power_law_1.2,0.4093696117401123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,32,power_law_1.01,0.07020800113677979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,8,power_law_1.2,0.13148159980773927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,2,power_law_1.2,0.4276480197906494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,32,power_law_1.01,0.07390080094337463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,2,power_law_1.2,0.4227263927459717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,8,power_law_1.2,0.14216959476470947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,32,power_law_1.01,0.07570559978485107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,32,8,balanced,0.044437333941459656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,2,power_law_1.2,0.4639743804931641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,32,8,balanced,0.046207999189694725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,32,8,balanced,0.047397335370381675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,32,8,balanced,0.048309331138928734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,32,8,balanced,0.05264000097910563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,8,power_law_1.2,0.14090240001678467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,32,8,balanced,0.05764266848564148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,32,power_law_1.01,0.07212160229682922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,32,8,balanced,0.06389333307743073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,32,8,balanced,0.06227200229962667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,2,power_law_1.2,0.469708776473999
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,32,8,balanced,0.06257600088914235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,32,8,balanced,0.06263466676076253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,32,8,balanced,0.06465599934260051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,32,8,balanced,0.0653546651204427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,32,8,balanced,0.06673599779605865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,32,power_law_1.01,0.07292799949645996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,32,8,balanced,0.06879466772079468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,32,8,balanced,0.0745119998852412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,8,power_law_1.2,0.14606080055236817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,2,power_law_1.2,0.5005055904388428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,32,8,balanced,0.07867733140786488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,32,8,balanced,0.08065600196520488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,32,8,balanced,0.08925333619117737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,32,8,balanced,0.09964799880981445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,32,power_law_1.01,0.07448959946632386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,32,8,balanced,0.12619733810424805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,2,power_law_1.2,0.5844992160797119
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,32,8,balanced,0.16421866416931152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,32,8,balanced,0.2032159964243571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,8,power_law_1.2,0.15648000240325927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,32,8,balanced,0.24122132857640585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,32,power_law_1.01,0.07701119780540466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,32,8,balanced,0.3335893154144287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,32,8,balanced,0.42220266660054523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,2,power_law_1.2,0.7234816074371337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,32,8,balanced,0.6121280193328857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,32,power_law_1.01,0.07628160119056701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,32,8,balanced,0.7813546657562256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,8,power_law_1.2,0.16568319797515868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,32,8,balanced,0.9702933629353842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,2,power_law_1.2,0.8292799949645996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,32,power_law_1.01,0.0796288013458252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,32,8,balanced,1.4900906880696614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,8,power_law_1.2,0.1743232011795044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,32,power_law_1.01,0.08528640270233154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,32,8,balanced,2.9536479314168296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,2,power_law_1.2,1.0309056282043456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,8,power_law_1.2,0.20312960147857667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,32,power_law_1.01,0.08618239760398864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,2,power_law_1.2,1.3103039741516114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,32,power_law_1.01,0.09353600144386291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,8,power_law_1.2,0.21386239528656006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,32,power_law_1.01,0.10276479721069336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,2,power_law_1.2,1.746086311340332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,8,power_law_1.2,0.2663167953491211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,32,power_law_1.01,0.11812479496002197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,32,power_law_1.01,0.12666239738464355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,8,power_law_1.2,0.3041215896606445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,2,power_law_1.2,2.237772750854492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,32,power_law_1.01,0.1473536014556885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,8,power_law_1.2,0.3847552061080933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,32,power_law_1.01,0.20081920623779298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,2,power_law_1.2,2.6420671463012697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,32,power_law_1.01,0.20007040500640869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,8,power_law_1.2,0.45423359870910646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,32,power_law_1.01,0.30595200061798095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,8,power_law_1.2,0.583948802947998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,32,power_law_1.01,0.3743743896484375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,2,power_law_1.2,4.800320053100586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,8,power_law_1.2,0.7625664234161377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,32,power_law_1.01,0.5087168216705322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,32,power_law_1.01,0.6332992076873779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,8,power_law_1.2,1.0923392295837402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,32,power_law_1.01,0.7496064186096192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,8,power_law_1.2,1.2424256324768066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,2,power_law_1.2,6.966521453857422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,32,power_law_1.01,1.196396827697754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,8,power_law_1.2,1.7263168334960937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,64,power_law_1.2,0.08214399814605713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,8,power_law_1.2,2.490291213989258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,32,power_law_1.01,2.3271615982055662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,64,power_law_1.2,0.09534720182418824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,8,power_law_1.2,5.70624008178711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,64,power_law_1.2,0.05691519975662231
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,64,power_law_1.2,0.05880320072174072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,16,power_law_1.2,0.06394240260124207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,64,power_law_1.2,0.06380800008773804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,16,power_law_1.2,0.07810559868812561
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,64,power_law_1.2,0.07614719867706299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,64,power_law_1.2,0.03274880051612854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,16,power_law_1.2,0.05866879820823669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,64,power_law_1.2,0.034406399726867674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,64,power_law_1.2,0.029471999406814574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,16,power_law_1.2,0.05546240210533142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,64,power_law_1.2,0.03018240034580231
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,64,power_law_1.2,0.07452800273895263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,64,power_law_1.2,0.030604800581932066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,64,power_law_1.2,0.03041279911994934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,16,power_law_1.2,0.0590719997882843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,64,power_law_1.2,0.031097599864006044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,128,power_law_1.01,0.0634112000465393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,64,power_law_1.2,0.03160960078239441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,64,power_law_1.2,0.07533439993858337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,16,power_law_1.2,0.06245120167732239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,64,power_law_1.2,0.030508801341056824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,128,power_law_1.01,0.06100479960441589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,64,power_law_1.2,0.031040000915527343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,64,power_law_1.2,0.032390400767326355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,16,power_law_1.2,0.06430720090866089
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,64,power_law_1.2,0.03248000144958496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,128,power_law_1.01,0.06265599727630615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,64,power_law_1.2,0.07703040242195129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,64,power_law_1.2,0.033907198905944826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,16,power_law_1.2,0.07441279888153077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,64,power_law_1.2,0.03489919900894165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,128,power_law_1.01,0.05671679973602295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,64,power_law_1.2,0.03821440041065216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,64,power_law_1.2,0.03887999951839447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,16,power_law_1.2,0.07290239930152893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,64,power_law_1.2,0.07841920256614685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,64,power_law_1.2,0.04068480134010315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,128,power_law_1.01,0.06135680079460144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,64,power_law_1.2,0.04725759923458099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,16,power_law_1.2,0.07420799732208253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,64,power_law_1.2,0.054041600227355956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,128,power_law_1.01,0.06112639904022217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,64,power_law_1.2,0.057555198669433594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,64,power_law_1.2,0.08133119940757752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,64,power_law_1.2,0.06875519752502442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,16,power_law_1.2,0.07433599829673768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,128,power_law_1.01,0.06204800009727478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,64,power_law_1.2,0.09168000221252441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,64,power_law_1.2,0.11118079423904419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,16,power_law_1.2,0.08181759715080261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,128,power_law_1.01,0.0657472014427185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,64,power_law_1.2,0.1421183943748474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,64,power_law_1.2,0.08008959889411926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,16,power_law_1.2,0.081632000207901
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,64,power_law_1.2,0.18710399866104127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,128,power_law_1.01,0.06626560091972351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,64,power_law_1.2,0.2902400016784668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,16,power_law_1.2,0.08259199857711792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,128,power_law_1.01,0.06531839966773986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,64,power_law_1.2,0.08266879916191101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,64,power_law_1.2,0.3485055923461914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,16,power_law_1.2,0.0921343982219696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,64,power_law_1.2,0.47420802116394045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,128,power_law_1.01,0.06528000235557556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,16,power_law_1.2,0.09784319996833801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,64,power_law_1.2,0.08524799942970276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,128,power_law_1.01,0.06584320068359376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,64,power_law_1.2,1.0618368148803712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,16,power_law_1.2,0.10497280359268188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,128,power_law_1.01,0.06735360026359558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,16,power_law_1.2,0.11850880384445191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,64,power_law_1.2,0.09074559807777405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,128,power_law_1.01,0.070796799659729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,64,power_law_1.2,1.761356735229492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,16,power_law_1.2,0.1307584047317505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,128,power_law_1.01,0.07541120052337646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,16,power_law_1.2,0.15913599729537964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,64,power_law_1.2,0.0966592013835907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,128,power_law_1.01,0.07826560139656066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,16,power_law_1.2,0.18497920036315918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,128,power_law_1.01,0.08370559811592101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,64,power_law_1.2,0.10228480100631714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,16,power_law_1.2,0.24723200798034667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,128,power_law_1.01,0.09352959990501404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,16,power_law_1.2,0.283404803276062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,128,power_law_1.01,0.10606720447540283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,64,power_law_1.2,0.12680959701538086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,128,power_law_1.01,0.1323583960533142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,16,power_law_1.2,0.3538111925125122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,128,power_law_1.01,0.15512959957122802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,64,power_law_1.2,0.12538880109786987
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,16,power_law_1.2,0.45462398529052733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,128,power_law_1.01,0.20257279872894288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,16,power_law_1.2,0.6337152004241944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,64,power_law_1.2,0.16491520404815674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,128,power_law_1.01,0.2465536117553711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,4,power_law_1.01,0.15911680459976196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,16,power_law_1.2,0.8642623901367188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,4,power_law_1.01,0.17681920528411865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,128,power_law_1.01,0.3378432035446167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,64,power_law_1.2,0.16976640224456788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,4,power_law_1.01,0.13960959911346435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,128,power_law_1.01,0.4289408206939697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,4,power_law_1.01,0.2187135934829712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,16,power_law_1.2,1.01396484375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,4,power_law_1.01,0.288972806930542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,64,power_law_1.2,0.22946560382843018
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,128,power_law_1.01,0.613644790649414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,4,power_law_1.01,0.48691840171813966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,4,power_law_1.01,0.5085567951202392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,16,power_law_1.2,1.5385343551635742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,4,power_law_1.01,0.5246848106384278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,128,power_law_1.01,0.79169921875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,64,power_law_1.2,0.24742400646209717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,4,power_law_1.01,0.5247104167938232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,4,power_law_1.01,0.5033919811248779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,4,power_law_1.01,0.5198016166687012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,128,power_law_1.01,1.0025919914245605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,4,power_law_1.01,0.541542387008667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,64,power_law_1.2,0.30103681087493894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,4,power_law_1.01,0.53853440284729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,16,power_law_1.2,3.262860870361328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,4,power_law_1.01,0.5494527816772461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,128,power_law_1.01,1.5339903831481934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,4,power_law_1.01,0.5866687774658204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,4,power_law_1.01,0.5690432071685791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,64,power_law_1.2,0.4268671989440918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,4,power_law_1.01,0.6301695823669433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,4,power_law_1.01,0.7016255855560303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,4,power_law_1.01,0.7475391864776612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,128,power_law_1.01,3.1138111114501954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,4,power_law_1.01,0.9064640045166016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,64,power_law_1.2,0.6310080051422119
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,4,power_law_1.01,0.8322303771972657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,4,power_law_1.01,0.9765631675720214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,4,power_law_1.01,1.1454463958740235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,64,power_law_1.2,0.7041152000427247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,4,power_law_1.01,1.403872013092041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,4,power_law_1.01,1.5841471672058105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,4,power_law_1.01,2.1847871780395507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,8,power_law_1.01,0.03199360072612763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,2,64,power_law_1.2,0.8123264312744141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,8,power_law_1.01,0.03831680119037628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,4,power_law_1.01,3.162099266052246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,8,power_law_1.01,0.02953599989414215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,8,power_law_1.01,0.031206399202346802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,8,power_law_1.01,0.032691198587417605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,4,power_law_1.01,3.5246273040771485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,8,power_law_1.01,0.03625600039958954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,4,balanced,0.033546666304270424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,4,balanced,0.037621334195137024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,4,balanced,0.05528533458709717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,8,power_law_1.01,0.036268800497055054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,4,balanced,0.07860800127188365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,4,balanced,0.12165866295496623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,4,balanced,0.12316266695658366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,4,balanced,0.1232266624768575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,4,balanced,0.12173333764076233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,4,balanced,0.12387733658154805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,4,balanced,0.1227946678797404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,4,balanced,0.12456533312797546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,4,balanced,0.1266986628373464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,8,power_law_1.01,0.03565439879894257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,4,balanced,0.12758400042851767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,4,balanced,0.13108799854914346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,4,balanced,0.13824533422787985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,4,balanced,0.13820266723632812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,4,balanced,0.14274666706720987
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,4,balanced,0.15251200397809347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,4,balanced,0.16214932998021445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,4,balanced,0.18868800004323324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,8,power_law_1.01,0.0363072007894516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,4,balanced,0.1996799906094869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,4,balanced,0.25460267066955566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,4,balanced,0.2930720051129659
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,4,balanced,0.3985813458760579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,4,balanced,0.4755520025889079
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,8,power_law_1.01,0.03704319894313812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,4,balanced,0.6579306523005167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,2,64,power_law_1.2,1.3359487533569336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,4,balanced,0.8403519789377848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,4,power_law_1.01,5.313068771362305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,8,power_law_1.01,0.038822400569915774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,4,balanced,1.0286346276601155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,4,balanced,1.5930239359537761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,8,power_law_1.01,0.04069760143756866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,4,balanced,3.0686238606770835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,8,power_law_1.01,0.041433599591255185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,8,power_law_1.01,0.044531199336051944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,8,power_law_1.01,0.05125120282173157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,8,power_law_1.01,0.052288001775741576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,8,power_law_1.01,0.05452160239219665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,8,power_law_1.01,0.05885440111160278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,4,power_law_1.01,9.363289642333985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,8,power_law_1.01,0.06288639903068542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,8,power_law_1.01,0.07438079714775085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,8,power_law_1.01,0.07989760041236878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,8,power_law_1.01,0.10238720178604126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,8,power_law_1.01,0.11849600076675415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,8,power_law_1.01,0.14725760221481324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,2,64,power_law_1.2,3.204576110839844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,8,power_law_1.01,0.1957695960998535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,8,power_law_1.01,0.2555775880813599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,8,power_law_1.01,0.35757439136505126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,2,8,power_law_1.01,0.4009727954864502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,2,8,power_law_1.01,0.6241856098175049
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,1,power_law_1.01,0.0833791971206665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,2,8,power_law_1.01,1.1584063529968263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,1,power_law_1.01,0.08893439769744874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,1,power_law_1.01,0.10064640045166015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,4,power_law_1.2,0.05013120174407959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,1,power_law_1.01,0.12209279537200927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,4,power_law_1.2,0.07047039866447449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,1,power_law_1.01,0.17501440048217773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,4,power_law_1.2,0.0568448007106781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,2,balanced,0.05622399846712748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,2,balanced,0.08970133463541667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,4,power_law_1.2,0.06763520240783691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,2,balanced,0.10475732882817586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,2,balanced,0.15466666221618652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,1,power_law_1.01,0.22120959758758546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,2,balanced,0.26022932926813763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,2,balanced,0.3167733351389567
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,2,balanced,0.318725327650706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,2,balanced,0.3169493277867635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,2,balanced,0.3224800030390422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,2,balanced,0.32204266389211017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,4,power_law_1.2,0.07971199750900268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,2,balanced,0.3224800030390422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,2,balanced,0.32862399021784466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,2,balanced,0.33005332946777344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,1,power_law_1.01,0.24995839595794678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,2,balanced,0.33689598242441815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,2,balanced,0.34515198071797687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,2,balanced,0.35011200110117596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,2,balanced,0.3608906666437785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,4,power_law_1.2,0.09678720235824585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,2,balanced,0.3858666817347209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,2,balanced,0.4017546574274699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,2,balanced,0.4962559938430786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,2,balanced,0.4986666838328044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,2,balanced,0.6312106847763062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,1,power_law_1.01,0.3027008056640625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,4,power_law_1.2,0.12400000095367432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,2,balanced,0.6877760092417399
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,2,balanced,1.0511786937713623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,2,balanced,1.1640053590138753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,2,balanced,1.744858741760254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,4,power_law_1.2,0.12802560329437257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,1,power_law_1.01,0.31208961009979247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,2,balanced,2.1137332916259766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,2,balanced,2.459258715311686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,4,power_law_1.2,0.13563519716262817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,1,power_law_1.01,0.32503039836883546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,2,balanced,3.8508211771647134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,4,power_law_1.2,0.1362496018409729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,8,8,balanced,0.033701332906881966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,8,8,balanced,0.033488000432650246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,8,8,balanced,0.03384533276160558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,8,8,balanced,0.033189333975315094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,8,8,balanced,0.033615998923778534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,8,8,balanced,0.035386666655540466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,8,8,balanced,0.033674667278925575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,8,8,balanced,0.03482133398453394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,8,8,balanced,0.035391998787721
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,8,8,balanced,0.03570133447647095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,8,8,balanced,0.03741333385308584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,8,8,balanced,0.03581333408753077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,4,power_law_1.2,0.14172159433364867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,8,8,balanced,0.03745600084463755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,8,8,balanced,0.03944533318281174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,1,power_law_1.01,0.49580159187316897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,8,8,balanced,0.04364799956480662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,8,8,balanced,0.043509334325790405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,8,8,balanced,0.04374399781227112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,2,balanced,7.368570963541667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,8,8,balanced,0.05402666827042898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,8,8,balanced,0.05579199890295664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,8,8,balanced,0.06849066913127899
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,8,8,balanced,0.08832533160845439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,8,8,balanced,0.1137600044409434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,8,8,balanced,0.13594133655230203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,8,8,balanced,0.17985600233078003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,8,8,balanced,0.2260106603304545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,8,8,balanced,0.3118879993756612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,4,power_law_1.2,0.14727040529251098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,8,8,balanced,0.3986986478169759
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,8,8,balanced,0.4819360176722209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,8,8,balanced,0.7255893548329672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,1,power_law_1.01,0.5073728084564209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,8,8,balanced,1.3911253611246746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,4,power_law_1.2,0.15232640504837036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,4,power_law_1.2,0.16367360353469848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,1,power_law_1.01,0.5325695991516113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,4,power_law_1.2,0.17011200189590453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,1,power_law_1.01,0.5502783775329589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,4,power_law_1.2,0.1797376036643982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,1,power_law_1.01,0.5698624134063721
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,4,power_law_1.2,0.18588160276412963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,4,power_law_1.2,0.21250560283660888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,1,power_law_1.01,0.5697152137756347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,4,power_law_1.2,0.23304319381713867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,1,power_law_1.01,0.595366382598877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,4,power_law_1.2,0.2786495923995972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,1,power_law_1.01,0.6631231784820557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,4,power_law_1.2,0.3308799982070923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,4,power_law_1.2,0.4080639839172363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,1,power_law_1.01,0.7024896144866943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,4,power_law_1.2,0.47409281730651853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,1,power_law_1.01,0.8233535766601563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,4,power_law_1.2,0.033471998572349546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,4,power_law_1.2,0.655788803100586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,4,power_law_1.2,0.058880001306533813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,4,power_law_1.2,0.05919359922409058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,1,power_law_1.01,0.9078271865844727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,4,power_law_1.2,0.03277440071105957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,4,power_law_1.2,0.8284543991088867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,4,power_law_1.2,0.06033279895782471
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,4,power_law_1.2,0.06785280108451844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,4,power_law_1.2,0.03389439880847931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,1,power_law_1.01,1.1220864295959472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,4,power_law_1.2,1.1809663772583008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,4,power_law_1.2,0.07639039754867553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,4,power_law_1.2,0.0385343998670578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,4,power_law_1.2,0.0963648021221161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,1,power_law_1.01,1.3032832145690918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,4,power_law_1.2,0.10032000541687011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,4,power_law_1.2,0.03991680145263672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,4,power_law_1.2,1.521459197998047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,4,power_law_1.2,0.10321279764175414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,4,power_law_1.2,0.04981119930744171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,4,power_law_1.2,0.1040768027305603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,1,power_law_1.01,1.6988607406616212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,4,power_law_1.2,0.05415679812431336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,4,power_law_1.2,0.10422400236129761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,4,power_law_1.2,1.8391807556152344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,4,power_law_1.2,0.10627199411392212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,4,power_law_1.2,0.05625600218772888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,4,power_law_1.2,0.11022080183029175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,1,power_law_1.01,2.066265678405762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,4,power_law_1.2,0.05334399938583374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,4,power_law_1.2,0.11228159666061402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,4,power_law_1.2,2.7954944610595702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,4,power_law_1.2,0.05811200141906738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,4,power_law_1.2,0.11956479549407958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,4,power_law_1.2,0.058815997838973996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,4,power_law_1.2,0.13089920282363893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,1,power_law_1.01,2.840742492675781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,4,power_law_1.2,0.13099520206451415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,4,power_law_1.2,0.061536002159118655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,4,power_law_1.2,0.14044159650802612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,4,power_law_1.2,0.062003201246261595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,4,power_law_1.2,0.15857919454574584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,4,power_law_1.2,0.06332160234451294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,1,power_law_1.01,3.59090576171875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,4,power_law_1.2,0.17267199754714965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,4,power_law_1.2,5.332044982910157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,4,power_law_1.2,0.07029759883880615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,4,power_law_1.2,0.20436480045318603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,4,power_law_1.2,0.0711296021938324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,4,power_law_1.2,0.23447680473327637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,4,power_law_1.2,0.30577280521392824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,4,power_law_1.2,0.06972799897193908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,1,power_law_1.01,4.354982376098633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,4,power_law_1.2,0.3861952066421509
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,4,power_law_1.2,0.07824640274047852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,4,power_law_1.2,0.5198912143707275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,4,power_law_1.2,0.08402559757232667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,4,power_law_1.2,0.10121599435806275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,4,power_law_1.2,0.6689856052398682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,4,power_law_1.2,0.11172480583190918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,1,power_law_1.01,6.6575874328613285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,4,power_law_1.2,1.0420031547546387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,4,power_law_1.2,0.12641919851303102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,4,power_law_1.2,0.1720896005630493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,4,power_law_1.2,1.1809791564941405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,4,power_law_1.2,0.2148224115371704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,4,power_law_1.2,1.5382847785949707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,4,power_law_1.2,0.25532801151275636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,4,power_law_1.2,0.35752320289611816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,4,power_law_1.2,2.405612754821777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,4,power_law_1.2,0.48522238731384276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,1,power_law_1.01,12.933004760742188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,2,4,power_law_1.2,0.5507967948913575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,2,power_law_1.2,0.08483200073242188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,2,4,power_law_1.2,0.9147904396057129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,4,power_law_1.2,5.119609451293945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,2,power_law_1.2,0.10128639936447144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,2,power_law_1.2,0.09775360226631165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,2,4,power_law_1.2,1.7819135665893555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,2,power_law_1.2,0.14291199445724487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,2,power_law_1.2,0.1876479983329773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,2,power_law_1.2,0.32806398868560793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,2,power_law_1.2,0.3658112049102783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,16,power_law_1.2,0.042822399735450746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,2,power_law_1.2,0.3716928005218506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,16,power_law_1.2,0.03699199855327606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,16,power_law_1.2,0.033939200639724734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,16,power_law_1.2,0.03423359990119934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,2,power_law_1.2,0.38708479404449464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,16,power_law_1.2,0.03544960021972656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,1,power_law_1.2,0.049491199851036075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,16,power_law_1.2,0.038700801134109494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,1,power_law_1.2,0.05594879984855652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,16,power_law_1.2,0.03880960047245026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,2,power_law_1.2,0.3956736087799072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,1,power_law_1.2,0.062015998363494876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,16,power_law_1.2,0.03878400027751923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,1,power_law_1.2,0.07416960000991821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,16,power_law_1.2,0.03900800049304962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,1,power_law_1.2,0.09066879749298096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,16,power_law_1.2,0.03978880047798157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,2,power_law_1.2,0.40856318473815917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,1,power_law_1.2,0.1440127968788147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,16,power_law_1.2,0.040505599975585935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,1,power_law_1.2,0.15130239725112915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,16,power_law_1.2,0.04168959856033325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,1,power_law_1.2,0.16188160181045533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,4,power_law_1.2,0.09485440254211426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,16,power_law_1.2,0.043398401141166686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,2,power_law_1.2,0.42357120513916013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,1,power_law_1.2,0.16250879764556886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,16,power_law_1.2,0.044512000679969785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,1,power_law_1.2,0.16479359865188598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,16,power_law_1.2,0.05015680193901062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,1,power_law_1.2,0.1746880054473877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,16,power_law_1.2,0.0506816029548645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,2,power_law_1.2,0.4402495861053467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,1,power_law_1.2,0.1757375955581665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,16,power_law_1.2,0.05767679810523987
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,1,power_law_1.2,0.1820736050605774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,4,power_law_1.2,0.1211840033531189
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,16,power_law_1.2,0.06505600214004517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,1,power_law_1.2,0.188646399974823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,16,power_law_1.2,0.06823679804801941
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,2,power_law_1.2,0.4514495849609375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,1,power_law_1.2,0.19633920192718507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,16,power_law_1.2,0.08015999794006348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,1,power_law_1.2,0.20472319126129152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,16,power_law_1.2,0.09753599762916565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,1,power_law_1.2,0.21755518913269042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,16,power_law_1.2,0.1177664041519165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,4,power_law_1.2,0.12346240282058715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,2,power_law_1.2,0.4632575988769531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,1,power_law_1.2,0.24569599628448485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,16,power_law_1.2,0.17738879919052125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,1,power_law_1.2,0.2741760015487671
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,16,power_law_1.2,0.20894720554351806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,1,power_law_1.2,0.33279359340667725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,2,power_law_1.2,0.49459199905395507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,16,power_law_1.2,0.28275198936462403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,1,power_law_1.2,0.3734783887863159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,4,power_law_1.2,0.112716805934906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,1,power_law_1.2,0.4805439949035645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,16,power_law_1.2,0.46029438972473147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,1,power_law_1.2,0.5646592140197754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,2,power_law_1.2,0.5195648193359375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,16,power_law_1.2,0.5929088115692138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,1,power_law_1.2,0.758841609954834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,4,power_law_1.2,0.14581120014190674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,1,power_law_1.2,0.947116756439209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,16,power_law_1.2,0.6956160068511963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,2,power_law_1.2,0.588588809967041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,1,power_law_1.2,1.3401599884033204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,4,power_law_1.2,0.1811903953552246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,16,power_law_1.2,1.3249664306640625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,2,power_law_1.2,0.6060863971710205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,1,power_law_1.2,1.7315263748168945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,2688,1856,6,128,1,1,power_law_1.2,2.1192768096923826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,2,power_law_1.2,0.7160704135894775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,4,power_law_1.2,0.21923201084136962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,16,power_law_1.2,2.8833728790283204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,2688,1856,6,128,1,1,power_law_1.2,3.2794815063476563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,2,power_law_1.2,0.8020031929016114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,4,power_law_1.2,0.27942399978637694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,2,power_law_1.2,0.9908224105834961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,4,power_law_1.2,0.28526721000671384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,2688,1856,6,128,1,1,power_law_1.2,6.404927825927734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,2,power_law_1.2,1.0994688034057618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,4,power_law_1.2,0.30081279277801515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,2,power_law_1.2,1.4139583587646485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,4,power_law_1.2,0.3205312013626099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,4,power_law_1.2,0.32216320037841795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,2,power_law_1.2,1.8362239837646483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,4,power_law_1.2,0.33935999870300293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,32,2,balanced,0.031744000812371574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,32,2,balanced,0.03287466615438461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,32,2,balanced,0.03378133227427801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,32,2,balanced,0.034629332522551216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,32,2,balanced,0.03621333340803782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,32,2,balanced,0.035429333647092186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,32,2,balanced,0.035360001027584076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,32,2,balanced,0.03544000039498011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,32,2,balanced,0.03618666778008143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,32,2,balanced,0.03772266705830892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,32,2,balanced,0.037418665985266365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,32,2,balanced,0.03955733279387156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,32,2,balanced,0.03766933331886927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,32,2,balanced,0.039477333426475525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,32,2,balanced,0.0498879998922348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,32,2,balanced,0.05145066479841868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,32,2,balanced,0.05221866567929586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,32,2,balanced,0.06215466558933258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,32,2,balanced,0.07163199782371521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,32,2,balanced,0.08489066362380981
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,32,2,balanced,0.09915733337402344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,32,2,balanced,0.12735467155774435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,32,2,balanced,0.1306613286336263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,32,2,balanced,0.19080533583958945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,2,power_law_1.2,2.6328832626342775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,32,2,balanced,0.22909865776697794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,32,2,balanced,0.3166453242301941
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,4,power_law_1.2,0.34168961048126223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,32,2,balanced,0.40459732214609784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,32,2,balanced,0.4941706657409668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,32,2,balanced,0.7335466543833414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,32,2,balanced,1.41266663869222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,4,power_law_1.2,0.3605119943618774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,2,32,balanced,0.02590399980545044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,2,32,balanced,0.023455999791622162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,2,32,balanced,0.025455998877684276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,2,32,balanced,0.025557334224383037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,2,32,balanced,0.025477332373460133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,2,32,balanced,0.027136000494162243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,2,32,balanced,0.028362666567166645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,2,32,balanced,0.025418666501839954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,2,32,balanced,0.025514667232831318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,2,32,balanced,0.0271573339899381
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,2,32,balanced,0.029557332396507263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,2,32,balanced,0.029264000554879505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,2,32,balanced,0.029370665550231934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,2,32,balanced,0.03136000037193298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,2,32,balanced,0.033589333295822144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,2,32,balanced,0.035360001027584076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,2,32,balanced,0.03552533437808355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,2,32,balanced,0.03575466573238373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,32,balanced,0.03769599894682566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,32,balanced,0.039781334499518074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,32,balanced,0.03974399964014689
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,32,balanced,0.04267199834187826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,32,balanced,0.04611733555793762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,32,balanced,0.056186666091283165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,32,balanced,0.06414400041103363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,32,balanced,0.08488532900810242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,32,balanced,0.09640000263849895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,2,power_law_1.2,3.0816383361816406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,32,balanced,0.10893332958221436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,32,balanced,0.14565866192181906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,32,balanced,0.2400266726811727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,4,power_law_1.2,0.3777983903884888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,4,power_law_1.2,0.3928767919540405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,4,power_law_1.2,0.42357759475708007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,4,2,power_law_1.2,3.7781631469726564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,4,power_law_1.2,0.46512641906738283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,4,power_law_1.2,0.5156864166259766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,4,power_law_1.2,0.5763840198516845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,4,2,power_law_1.2,5.655859375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,8,power_law_1.2,0.11859840154647827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,4,power_law_1.2,0.7545407772064209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,8,power_law_1.2,0.10536960363388062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,1,1,power_law_1.2,0.13500159978866577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,1,1,power_law_1.2,0.18060799837112426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,1,1,power_law_1.2,0.26256000995635986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,1,1,power_law_1.2,0.43203201293945315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,1,1,power_law_1.2,0.41150717735290526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,4,power_law_1.2,0.784607982635498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,1,1,power_law_1.2,0.4341887950897217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,1,1,power_law_1.2,0.4484992027282715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,1,1,power_law_1.2,0.45395841598510744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,8,power_law_1.2,0.07709439992904663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,1,1,power_law_1.2,0.4641983985900879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,1,1,power_law_1.2,0.4566976070404053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,1,1,power_law_1.2,0.5059135913848877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,1,1,power_law_1.2,0.48718719482421874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,1,1,power_law_1.2,0.5025343894958496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,4,power_law_1.2,0.9926336288452149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,1,1,power_law_1.2,0.4971456050872803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,8,power_law_1.2,0.10251519680023194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,1,1,power_law_1.2,0.5423168182373047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,1,1,power_law_1.2,0.5523839950561523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,1,1,power_law_1.2,0.6241600036621093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,1,1,power_law_1.2,0.6855296134948731
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,1,1,power_law_1.2,0.7423103809356689
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,1,1,power_law_1.2,0.924953556060791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,8,power_law_1.2,0.12451200485229492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,4,power_law_1.2,1.239353561401367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,1,1,power_law_1.2,1.0669695854187011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,1,1,power_law_1.2,1.4287615776062013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,4,2,power_law_1.2,11.880339050292969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,1,1,power_law_1.2,1.7043968200683595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,8,power_law_1.2,0.1904960036277771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,1,1,power_law_1.2,2.429849624633789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,4,power_law_1.2,1.5952383995056152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,1,1,power_law_1.2,3.102412796020508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,8,power_law_1.2,0.20818560123443602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,1,1,power_law_1.2,4.516153717041016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,8,power_law_1.2,0.220467209815979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,4,power_law_1.2,2.115430450439453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,1,1,power_law_1.2,5.9587646484375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,8,power_law_1.2,0.22046079635620117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,1,1,power_law_1.2,7.450572967529297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,4,power_law_1.2,2.4892223358154295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,8,power_law_1.2,0.2237247943878174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,8,power_law_1.2,0.2283008098602295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,1,1,power_law_1.2,11.931890869140625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,8,power_law_1.2,0.23906559944152833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,4,power_law_1.2,3.7302272796630858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,2,balanced,0.0421066681543986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,2,balanced,0.043706665436426796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,2,balanced,0.050111999114354454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,2,balanced,0.07492266595363617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,2,balanced,0.10725333293279012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,8,power_law_1.2,0.24494080543518065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,2,balanced,0.10961066683133443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,2,balanced,0.10942400495211284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,2,balanced,0.10988266269365947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,2,balanced,0.11136533816655476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,2,balanced,0.116047998269399
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,2,balanced,0.11245333154996236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,2,balanced,0.11627200245857239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,2,balanced,0.1176639993985494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,2,balanced,0.1207413375377655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,2,balanced,0.1272586683432261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,2,balanced,0.12730133533477783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,2,balanced,0.13424533605575562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,2,balanced,0.14617600043614706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,2,balanced,0.1639786660671234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,2,balanced,0.18784532944361368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,2,balanced,0.21296000480651855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,2,balanced,0.2913813392321269
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,2,balanced,0.3396213452021281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,2,balanced,0.4846506516138713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,2,balanced,0.6030453443527222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,2,balanced,0.8740586439768473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,8,power_law_1.2,0.25012478828430174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,2,balanced,1.1333759625752766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,2,balanced,1.4047733942667644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,2,balanced,2.150858720143636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,1,1,power_law_1.2,23.505914306640626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,4,power_law_1.2,6.4953666687011715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,2,balanced,4.182991981506348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,8,power_law_1.2,0.26839680671691896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,8,power_law_1.2,0.2769407987594604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,8,power_law_1.2,0.28354558944702146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,8,power_law_1.2,0.3307391881942749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,8,power_law_1.2,0.35086081027984617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,8,power_law_1.2,0.45246081352233886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,8,power_law_1.2,0.4715007781982422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,32,1,balanced,0.05669333537419637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,32,1,balanced,0.05904000004132589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,32,1,balanced,0.06672533353169759
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,32,1,balanced,0.09266666571299235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,32,1,balanced,0.13738666971524557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,128,balanced,0.06610666712125142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,32,2,balanced,0.0317493329445521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,128,balanced,0.0662720004717509
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,32,2,balanced,0.029631999631722767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,32,1,balanced,0.21219199895858765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,32,2,balanced,0.03173866619666418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,128,balanced,0.06649066507816315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,32,1,balanced,0.21598933140436807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,32,2,balanced,0.033386667569478355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,128,balanced,0.06427200138568878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,32,2,balanced,0.03316266586383184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,128,balanced,0.06412800153096516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,32,2,balanced,0.03363200028737386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,32,1,balanced,0.2135466734568278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,32,2,balanced,0.03331200033426285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,128,balanced,0.07656533519426982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,32,2,balanced,0.03346666693687439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,128,balanced,0.08496000369389851
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,32,1,balanced,0.21739200750986734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,32,2,balanced,0.03421866645415624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,128,balanced,0.09080533186594646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,32,2,balanced,0.0337119996547699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,32,2,balanced,0.03578133384386698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,128,balanced,0.08684266606966655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,32,1,balanced,0.22116265694300333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,32,2,balanced,0.035391998787721
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,128,balanced,0.08498666683832805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,32,2,balanced,0.035349334279696144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,8,power_law_1.2,0.6300288200378418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,32,1,balanced,0.2210879921913147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,128,balanced,0.09066133697827657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,32,2,balanced,0.037290667494138084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,32,2,balanced,0.043621331453323364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,128,balanced,0.08679466446240743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,32,1,balanced,0.22463999191919962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,32,2,balanced,0.04172799984614054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,128,balanced,0.08771733442942302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,32,2,balanced,0.04391466577847799
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,128,balanced,0.09505599737167358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,32,1,balanced,0.22771199544270834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,32,2,balanced,0.05011733373006185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,128,balanced,0.09673600395520528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,32,2,balanced,0.05240533252557119
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,32,1,balanced,0.23338133096694946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,32,2,balanced,0.06603733201821645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,128,balanced,0.09893866380055745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,32,2,balanced,0.07800533374150594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,128,balanced,0.10513599713643391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,32,1,balanced,0.2364906668663025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,32,2,balanced,0.10431466499964397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,128,balanced,0.11773332953453064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,32,2,balanced,0.10339200496673584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,32,1,balanced,0.24343999226888022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,128,balanced,0.12357333302497864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,32,2,balanced,0.1435413360595703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,32,1,balanced,0.26234666506449383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,128,balanced,0.1437066694100698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,32,2,balanced,0.17495999733606973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,128,balanced,0.16222400466601053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,32,2,balanced,0.24219733476638794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,32,1,balanced,0.30170132716496784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,128,balanced,0.20274666945139566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,32,2,balanced,0.3083680073420207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,32,1,balanced,0.3195199966430664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,128,balanced,0.23755733172098795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,32,2,balanced,0.3715360164642334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,32,1,balanced,0.38412801424662274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,128,balanced,0.31856000423431396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,32,2,balanced,0.5453813473383585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,128,balanced,0.4076480070749919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,32,1,balanced,0.4586879809697469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,128,balanced,0.5573226610819498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,32,1,balanced,0.6356693506240845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,32,2,balanced,1.028277317682902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,128,balanced,0.7146933078765869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,32,1,balanced,0.7378880182902018
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,128,balanced,0.8889813423156738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,32,1,balanced,1.0920159816741943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,8,power_law_1.2,0.649951982498169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,128,balanced,1.3670506477355957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,32,1,balanced,1.3052106698354085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,32,1,balanced,1.9420533180236816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,128,balanced,2.6526986757914224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,32,1,balanced,2.5237654050191245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,32,1,balanced,3.1261812845865884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,8,power_law_1.2,0.8506943702697753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,32,1,balanced,4.865418752034505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,32,1,balanced,9.480805079142252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,8,power_law_1.2,1.017695999145508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,8,power_law_1.2,1.358131217956543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,8,power_law_1.2,1.7074239730834961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,1,power_law_1.01,0.07500799894332885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,1,power_law_1.01,0.09875839948654175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,16,power_law_1.2,0.05503360033035278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,1,power_law_1.01,0.12433919906616211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,2,8,power_law_1.2,1.8953983306884765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,1,power_law_1.01,0.19289599657058715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,16,power_law_1.2,0.05546879768371582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,1,power_law_1.01,0.26318719387054446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,16,8,power_law_1.2,0.03389439880847931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,16,8,power_law_1.2,0.03434880077838898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,16,power_law_1.2,0.04511359930038452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,1,power_law_1.01,0.44839038848876955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,16,8,power_law_1.2,0.03139840066432953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,1,power_law_1.01,0.47670397758483884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,16,8,power_law_1.2,0.030950400233268737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,16,power_law_1.2,0.046323201060295104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,16,8,power_law_1.2,0.031142398715019226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,1,power_law_1.01,0.4901951789855957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,16,8,power_law_1.2,0.031001600623130798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,16,8,power_law_1.2,0.032025599479675294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,16,power_law_1.2,0.046777600049972536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,1,power_law_1.01,0.5151360034942627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,16,8,power_law_1.2,0.03153280019760132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,16,8,power_law_1.2,0.03221760094165802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,1,power_law_1.01,0.5117055892944335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,16,8,power_law_1.2,0.033062401413917544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,2,8,power_law_1.2,3.1111936569213867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,16,power_law_1.2,0.04969600141048432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,16,8,power_law_1.2,0.03214080035686493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,1,power_law_1.01,0.5350399971008301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,16,8,power_law_1.2,0.034483200311660765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,16,8,power_law_1.2,0.035488000512123107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,16,power_law_1.2,0.051072001457214355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,16,8,power_law_1.2,0.0354559987783432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,1,power_law_1.01,0.5546304225921631
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,16,8,power_law_1.2,0.040294399857521056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,16,8,power_law_1.2,0.04413439929485321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,16,power_law_1.2,0.05308160185813904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,1,power_law_1.01,0.5540800094604492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,16,8,power_law_1.2,0.04696959853172302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,16,8,power_law_1.2,0.052960002422332765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,1,power_law_1.01,0.5725120067596435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,16,power_law_1.2,0.052300798892974856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,16,8,power_law_1.2,0.055936002731323244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,16,8,power_law_1.2,0.0691968023777008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,1,power_law_1.01,0.6002943992614747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,16,8,power_law_1.2,0.07827200293540955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,16,power_law_1.2,0.054067200422286986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,16,8,power_law_1.2,0.10928640365600586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,1,power_law_1.01,0.6226111888885498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,16,8,power_law_1.2,0.11857919692993164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,16,power_law_1.2,0.05532159805297852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,16,8,power_law_1.2,0.15733120441436768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,1,power_law_1.01,0.6477759838104248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,16,8,power_law_1.2,0.18960000276565553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,16,8,power_law_1.2,0.26450560092926023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,16,power_law_1.2,0.057222402095794676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,1,power_law_1.01,0.718284797668457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,16,8,power_law_1.2,0.3563391923904419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,16,power_law_1.2,0.0588096022605896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,16,8,power_law_1.2,0.4342463970184326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,1,power_law_1.01,0.7507711887359619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,16,8,power_law_1.2,0.6443264007568359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,16,power_law_1.2,0.061926400661468504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,1,power_law_1.01,0.8978943824768066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,2,8,power_law_1.2,6.2259521484375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,16,power_law_1.2,0.06787840127944947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,16,8,power_law_1.2,1.2137344360351563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,1,power_law_1.01,0.9947456359863281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,16,power_law_1.2,0.07237759828567505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,1,power_law_1.01,1.242905616760254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,16,power_law_1.2,0.07160320281982421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,1,power_law_1.01,1.3515263557434083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,16,power_law_1.2,0.0863103985786438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,1,power_law_1.01,1.7720256805419923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,16,power_law_1.2,0.09779840111732482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,1,power_law_1.01,2.4532096862792967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,16,power_law_1.2,0.1307968020439148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,16,power_law_1.2,0.16872960329055786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,1,power_law_1.01,3.2192447662353514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,16,power_law_1.2,0.21906559467315673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,16,power_law_1.2,0.248524808883667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,1,power_law_1.01,3.9752960205078125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,16,power_law_1.2,0.33999359607696533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,16,power_law_1.2,0.46090240478515626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,1,power_law_1.01,4.809439849853516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,16,1,power_law_1.01,0.0351936012506485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,16,power_law_1.2,0.7478015899658204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,16,1,power_law_1.01,0.037401598691940305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,16,1,power_law_1.01,0.03950720131397247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,4,power_law_1.2,0.041580799221992495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,16,1,power_law_1.01,0.05532159805297852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,16,1,power_law_1.01,0.058316802978515624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,4,power_law_1.2,0.03717760145664215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,16,1,power_law_1.01,0.056428802013397214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,16,power_law_1.2,0.8630271911621094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,16,1,power_law_1.01,0.05725439786911011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,4,power_law_1.2,0.042559999227523806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,16,1,power_law_1.01,0.057708799839019775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,1,power_law_1.01,6.9107521057128904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,16,1,power_law_1.01,0.058361601829528806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,4,power_law_1.2,0.04567680060863495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,16,1,power_law_1.01,0.06244480013847351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,16,1,power_law_1.01,0.06122879981994629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,4,power_law_1.2,0.05259519815444946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,16,power_law_1.2,1.0874624252319336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,16,1,power_law_1.01,0.06336640119552613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,4,power_law_1.2,0.0662015974521637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,16,1,power_law_1.01,0.06869760155677795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,4,power_law_1.2,0.070169597864151
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,16,1,power_law_1.01,0.06782079935073852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,4,power_law_1.2,0.07057920098304749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,16,1,power_law_1.01,0.07378559708595275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,16,1,power_law_1.01,0.07845119833946228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,4,power_law_1.2,0.07144960165023803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,16,1,power_law_1.01,0.08039680123329163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,16,power_law_1.2,1.7020736694335938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,4,power_law_1.2,0.07125120162963867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,16,1,power_law_1.01,0.09267200231552124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,16,1,power_law_1.01,0.1018496036529541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,4,power_law_1.2,0.07568640112876893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,16,1,power_law_1.01,0.11779839992523193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,4,power_law_1.2,0.07844480276107788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,16,1,power_law_1.01,0.13659520149230958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,16,1,power_law_1.01,0.1770815968513489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,1,power_law_1.01,13.24352569580078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,4,power_law_1.2,0.07857279777526856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,16,1,power_law_1.01,0.21649279594421386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,4,power_law_1.2,0.08199040293693542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,16,1,power_law_1.01,0.29616000652313235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,4,power_law_1.2,0.08931840062141419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,16,1,power_law_1.01,0.3542207956314087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,4,power_law_1.2,0.09171199798583984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,16,power_law_1.2,3.8275520324707033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,16,1,power_law_1.01,0.5005440235137939
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,4,power_law_1.2,0.093612802028656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,16,1,power_law_1.01,0.6397247791290284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,4,power_law_1.2,0.10327039957046509
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,16,1,power_law_1.01,0.7775424003601075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,4,power_law_1.2,0.11802239418029785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,16,1,power_law_1.01,1.1831744194030762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,4,power_law_1.2,0.14042880535125732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,4,power_law_1.2,0.1578559994697571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,16,1,power_law_1.01,2.3056320190429687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,4,power_law_1.2,0.21261439323425294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,4,power_law_1.2,0.27688961029052733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,4,power_law_1.2,0.36062719821929934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,4,power_law_1.2,0.4543039798736572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,4,power_law_1.2,0.7356544017791748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,4,power_law_1.2,0.9813887596130371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,4,power_law_1.2,1.168940830230713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,32,2,balanced,0.03183466692765554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,32,2,balanced,0.033674667278925575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,32,2,balanced,0.03743999948104223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,32,2,balanced,0.04008533308903376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,32,2,balanced,0.04798933366934458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,32,2,balanced,0.047930667797724404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,32,2,balanced,0.04849066833655039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,32,2,balanced,0.05177066723505656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,32,2,balanced,0.050928001602490745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,32,2,balanced,0.05031466484069824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,32,2,balanced,0.05299733579158783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,32,2,balanced,0.0537013312180837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,32,2,balanced,0.05839466551939646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,32,2,balanced,0.05829333265622457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,32,2,balanced,0.06621333460013072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,4,power_law_1.2,1.805766487121582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,32,2,balanced,0.06901866694291432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,32,2,balanced,0.07488533357779185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,32,2,balanced,0.085125337044398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,32,2,balanced,0.09763200084368388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,32,2,balanced,0.11179733276367188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,32,2,balanced,0.13776533802350363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,32,2,balanced,0.20914665857950845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,32,2,balanced,0.23267734050750732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,32,2,balanced,0.33506667613983154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,32,2,balanced,0.4292373259862264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,32,2,balanced,0.6215466658274332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,32,2,balanced,0.7835840384165446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,32,2,balanced,0.9741653601328532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,2,power_law_1.01,0.033318400382995605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,32,2,balanced,1.5101280212402344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,32,2,balanced,3.008277257283529
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,2,power_law_1.01,0.03697279989719391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,2,power_law_1.01,0.04041599929332733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,4,power_law_1.2,3.7386112213134766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,32,balanced,0.047584002216657005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,32,balanced,0.05076266825199127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,2,power_law_1.01,0.04873600006103516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,32,balanced,0.047194664676984154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,32,balanced,0.05329599976539612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,32,balanced,0.07448000212510426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,32,balanced,0.1014453371365865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,32,balanced,0.10122666756312053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,2,power_law_1.01,0.05988479852676391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,32,balanced,0.10247466961542766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,32,balanced,0.10390399893124898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,32,balanced,0.10417600472768147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,32,balanced,0.10295466581980388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,32,balanced,0.10575466354688008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,32,balanced,0.10698666175206502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,2,power_law_1.01,0.07716479897499084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,32,balanced,0.10841066638628642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,32,balanced,0.11417067050933838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,32,balanced,0.11660266915957133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,32,balanced,0.12015466888745625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,32,balanced,0.12974933783213297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,2,power_law_1.01,0.08172799944877625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,32,balanced,0.13606933752695718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,32,balanced,0.1539466679096222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,32,balanced,0.17092265685399374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,32,balanced,0.2032639980316162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,32,balanced,0.23542400201161703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,2,power_law_1.01,0.08449919819831848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,32,balanced,0.3112373352050781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,32,balanced,0.3699893156687419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,32,balanced,0.5113493204116821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,2,power_law_1.01,0.085343998670578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,32,balanced,0.6459146738052368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,32,balanced,0.7999626795450846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,2,power_law_1.01,0.08710399866104127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,32,balanced,1.1820800304412842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,2,power_law_1.01,0.0927295982837677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,32,balanced,2.3038719495137534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,2,power_law_1.01,0.095769602060318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,2,power_law_1.01,0.09730560183525086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,2,power_law_1.01,0.10234880447387695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,2,power_law_1.01,0.10840320587158203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,2,power_law_1.01,0.11315200328826905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,2,power_law_1.01,0.12138880491256714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,2,power_law_1.01,0.1394368052482605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,2,power_law_1.01,0.15835520029067993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,2,power_law_1.01,0.19852800369262696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,2,power_law_1.01,0.22789759635925294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,2,power_law_1.01,0.2928895950317383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,2,power_law_1.01,0.36786561012268065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,2,balanced,0.03143466760714849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,2,balanced,0.03125333289305369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,2,balanced,0.035445332527160645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,2,balanced,0.04853333532810211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,2,balanced,0.07238933444023132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,2,balanced,0.08538132905960083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,2,balanced,0.08479467034339905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,2,power_law_1.01,0.49292798042297364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,2,balanced,0.08474133412043254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,2,balanced,0.08493333061536153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,32,8,power_law_1.01,0.0535103976726532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,2,balanced,0.08452799916267395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,2,balanced,0.08521599570910136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,2,balanced,0.08524266878763835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,2,balanced,0.0865119993686676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,2,balanced,0.08703999718030293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,2,balanced,0.09174399574597676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,2,balanced,0.09409600496292114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,2,balanced,0.096778670946757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,2,balanced,0.10789866248766582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,2,balanced,0.1079360047976176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,2,balanced,0.13385599851608276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,2,balanced,0.13731732964515686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,2,balanced,0.18636266390482584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,2,power_law_1.01,0.6817855834960938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,2,balanced,0.1965226729710897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,2,balanced,0.2738826672236125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,2,balanced,0.3155999978383382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,2,balanced,0.4559839963912964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,2,balanced,0.5547200043996176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,32,8,power_law_1.01,0.051551997661590576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,2,2,balanced,0.6764640013376871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,2,2,balanced,1.030080000559489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,2,power_law_1.01,0.915436840057373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,2,2,balanced,1.9846879641215007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,32,8,power_law_1.01,0.059628802537918094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,2,power_law_1.01,1.2477439880371093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,32,8,power_law_1.01,0.04976640045642853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,4,power_law_1.01,0.03768959939479828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,2,power_law_1.01,1.4773311614990234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,32,8,power_law_1.01,0.054502397775650024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,4,power_law_1.01,0.03358719944953918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,2,power_law_1.01,2.4350208282470702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,32,8,power_law_1.01,0.061247998476028444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,4,power_law_1.01,0.03185920119285583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,4,power_law_1.01,0.03491199910640717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,32,8,power_law_1.01,0.06386560201644897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,4,power_law_1.01,0.03729279935359955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,32,8,power_law_1.01,0.06700159907341004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,4,power_law_1.01,0.04227199852466583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,2,power_law_1.01,4.578579330444336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,4,power_law_1.01,0.04221439957618713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,32,8,power_law_1.01,0.07091839909553528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,4,power_law_1.01,0.04382719993591309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,32,8,power_law_1.01,0.07226240038871765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,4,power_law_1.01,0.04366079866886139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,4,power_law_1.01,0.04455040097236633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,32,8,power_law_1.01,0.07326080203056336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,32,2,balanced,0.06856533388296764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,32,2,balanced,0.0745600014925003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,32,2,balanced,0.07865599791208903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,32,2,balanced,0.08703466256459554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,32,2,balanced,0.1120799978574117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,32,2,balanced,0.15779200196266174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,32,2,balanced,0.20758400360743204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,32,2,balanced,0.19707733392715454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,4,power_law_1.01,0.04652160108089447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,32,2,balanced,0.19766932725906372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,32,2,balanced,0.20001065731048584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,32,2,balanced,0.20415467023849487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,32,2,balanced,0.2059733271598816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,32,2,balanced,0.20517865816752115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,32,2,balanced,0.21036799748738608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,32,2,balanced,0.22174400091171265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,32,2,balanced,0.22228266795476279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,32,8,power_law_1.01,0.08083840012550354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,32,2,balanced,0.2324906587600708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,32,2,balanced,0.25362666447957355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,4,power_law_1.01,0.048492801189422605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,32,2,balanced,0.2743733326594035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,32,2,balanced,0.3200160066286723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,32,2,balanced,0.3589973449707031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,32,2,balanced,0.4498560031255086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,32,2,balanced,0.5389973322550455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,32,2,balanced,0.7305440107981364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,4,power_law_1.01,0.05109120011329651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,32,2,balanced,0.9948586622873942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,32,8,power_law_1.01,0.08009600043296813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,32,2,balanced,1.369823932647705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,32,2,balanced,1.8298667271931965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,4,power_law_1.01,0.05294719934463501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,32,2,balanced,2.278677304585775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,4,power_law_1.01,0.058771198987960814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,32,8,power_law_1.01,0.08298239707946778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,32,2,balanced,3.530949274698893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,4,power_law_1.01,0.06165760159492493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,32,8,power_law_1.01,0.08794879913330078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,32,2,balanced,6.800661087036133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,4,power_law_1.01,0.06368640065193176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,4,power_law_1.01,0.07214720249176025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,32,8,power_law_1.01,0.09174399971961975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,4,power_law_1.01,0.08612480163574218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,4,1,balanced,0.07726933558781941
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,32,8,power_law_1.01,0.09934719800949096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,4,1,balanced,0.17998933792114258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,4,power_law_1.01,0.10005120038986207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,4,1,balanced,0.2109973430633545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,4,1,balanced,0.20843199888865152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,4,1,balanced,0.208079993724823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,4,1,balanced,0.20570667584737143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,4,1,balanced,0.2063466707865397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,4,1,balanced,0.2065760095914205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,4,1,balanced,0.20815465847651163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,4,1,balanced,0.2077653408050537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,4,1,balanced,0.2155146598815918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,4,1,balanced,0.21156799793243408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,4,power_law_1.01,0.12228480577468873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,4,1,balanced,0.21228800217310587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,4,1,balanced,0.22714666525522867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,4,1,balanced,0.2510400017102559
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,4,1,balanced,0.2552853425343831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,32,8,power_law_1.01,0.1120959997177124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,4,1,balanced,0.24427199363708496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,4,1,balanced,0.29781333605448407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,4,1,balanced,0.3092533349990845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,4,1,balanced,0.384549339612325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,4,1,balanced,0.4136906862258911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,4,power_law_1.01,0.16250879764556886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,4,1,balanced,0.5661173264185587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,4,1,balanced,0.6949866612752279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,4,1,balanced,0.9939626852671305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,4,1,balanced,1.2716853618621826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,32,8,power_law_1.01,0.12175999879837036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,4,power_law_1.01,0.2132863998413086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,4,1,balanced,1.8827306429545085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,4,1,balanced,2.500666618347168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,4,power_law_1.01,0.25887360572814944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,4,1,balanced,3.12061341603597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,32,8,power_law_1.01,0.14318079948425294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,4,power_law_1.01,0.33383679389953613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,4,1,balanced,4.961893399556478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,32,8,power_law_1.01,0.16246399879455567
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,4,power_law_1.01,0.4348031997680664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,4,1,balanced,9.737306594848633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,32,8,power_law_1.01,0.2155776023864746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,4,power_law_1.01,0.5588031768798828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,32,8,power_law_1.01,0.2661439895629883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,4,power_law_1.01,0.7162752151489258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,32,8,power_law_1.01,0.3571840047836304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,4,power_law_1.01,0.05633919835090637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,4,power_law_1.01,1.0676159858703613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,4,power_law_1.01,0.06215680241584778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,32,8,power_law_1.01,0.4440192222595215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,4,power_law_1.01,0.07626879811286927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,32,8,power_law_1.01,0.6450175762176513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,4,power_law_1.01,2.2133056640625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,4,power_law_1.01,0.0664255976676941
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,32,8,power_law_1.01,0.8444607734680176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,4,power_law_1.01,0.08124160170555114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,4,power_law_1.01,0.09208319783210754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,32,8,power_law_1.01,1.0344511985778808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,4,power_law_1.01,0.10588799715042115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,64,power_law_1.2,0.03829759955406189
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,64,power_law_1.2,0.03922559916973114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,4,power_law_1.01,0.11965440511703491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,64,power_law_1.2,0.03200640082359314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,32,8,power_law_1.01,1.5515775680541992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,64,power_law_1.2,0.034041601419448855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,64,power_law_1.2,0.033107200264930726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,4,power_law_1.01,0.11981439590454102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,64,power_law_1.2,0.034016001224517825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,64,power_law_1.2,0.034764799475669864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,4,power_law_1.01,0.12277760505676269
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,64,power_law_1.2,0.03399679958820343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,64,power_law_1.2,0.03406080007553101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,64,power_law_1.2,0.03758080005645752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,4,power_law_1.01,0.12825599908828736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,64,power_law_1.2,0.04082559943199158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,32,8,power_law_1.01,3.0353599548339845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,64,power_law_1.2,0.04128639996051788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,4,power_law_1.01,0.13203840255737304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,64,power_law_1.2,0.04593279957771301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,64,power_law_1.2,0.04725759923458099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,4,power_law_1.01,0.13586560487747193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,64,power_law_1.2,0.0536575973033905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,64,power_law_1.2,0.05774719715118408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,4,power_law_1.01,0.1416767954826355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,64,power_law_1.2,0.058796799182891844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,64,power_law_1.2,0.0654911994934082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,64,power_law_1.2,0.07262719869613647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,32,1,power_law_1.01,0.09756799936294555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,4,power_law_1.01,0.15011199712753295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,64,power_law_1.2,0.0906495988368988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,64,power_law_1.2,0.11093759536743164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,4,power_law_1.01,0.15048320293426515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,64,power_law_1.2,0.16046719551086425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,64,power_law_1.2,0.16883840560913085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,32,1,power_law_1.01,0.0986624002456665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,4,power_law_1.01,0.16129280328750611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,64,power_law_1.2,0.23719680309295654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,4,power_law_1.01,0.1760256052017212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,64,power_law_1.2,0.26426880359649657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,64,power_law_1.2,0.4668543815612793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,32,1,power_law_1.01,0.10787839889526367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,4,power_law_1.01,0.1927616000175476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,64,power_law_1.2,0.7063231945037842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,4,power_law_1.01,0.2300096035003662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,8,power_law_1.2,0.035596799850463864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,32,1,power_law_1.01,0.12680959701538086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,64,power_law_1.2,0.8540608406066894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,4,power_law_1.01,0.2652928113937378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,8,power_law_1.2,0.0336896002292633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,8,power_law_1.2,0.03289600014686585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,64,power_law_1.2,1.087731170654297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,4,power_law_1.01,0.33544321060180665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,8,power_law_1.2,0.03445119857788086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,32,1,power_law_1.01,0.16306560039520263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,8,power_law_1.2,0.04076800048351288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,8,power_law_1.2,0.05081599950790405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,4,power_law_1.01,0.4010496139526367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,8,power_law_1.2,0.054553598165512085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,8,power_law_1.2,0.052025598287582395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,32,1,power_law_1.01,0.195961594581604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,64,power_law_1.2,2.594700813293457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,4,power_law_1.01,0.5434112071990966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,8,power_law_1.2,0.053388798236846925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,8,power_law_1.2,0.05196160078048706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,8,power_law_1.2,0.05397760272026062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,4,power_law_1.01,0.7153471946716309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,32,1,power_law_1.01,0.21175680160522461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,8,power_law_1.2,0.05427839756011963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,8,power_law_1.2,0.05609599947929382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,8,power_law_1.2,0.05793280005455017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,4,power_law_1.01,0.9968959808349609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,8,power_law_1.2,0.0630079984664917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,32,1,power_law_1.01,0.2548799991607666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,8,power_law_1.2,0.06428800225257873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,8,power_law_1.2,0.06561279892921448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,8,power_law_1.2,0.07320320010185241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,4,power_law_1.01,1.2743295669555663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,32,1,power_law_1.01,0.26320641040802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,8,power_law_1.2,0.07299839854240417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,8,power_law_1.2,0.08332160115242004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,8,power_law_1.2,0.08729599714279175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,4,power_law_1.01,1.6111679077148438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,8,power_law_1.2,0.11227519512176513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,32,1,power_law_1.01,0.27085440158843993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,8,power_law_1.2,0.13095680475234986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,8,power_law_1.2,0.15648000240325927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,32,1,power_law_1.01,0.281113600730896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,8,power_law_1.2,0.21994240283966066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,4,power_law_1.01,2.3315391540527344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,8,power_law_1.2,0.32065279483795167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,32,1,power_law_1.01,0.29212799072265627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,8,power_law_1.2,0.35831680297851565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,2,8,power_law_1.2,0.42465920448303224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,32,1,power_law_1.01,0.29679360389709475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,4,power_law_1.2,0.08796160221099854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,2,8,power_law_1.2,0.7566207885742188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,4,power_law_1.01,4.642208099365234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,32,1,power_law_1.01,0.31208961009979247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,4,power_law_1.2,0.07676159739494323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,2,8,power_law_1.2,1.2016192436218263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,32,1,power_law_1.01,0.32598400115966797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,4,power_law_1.2,0.08151040077209473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,32,1,power_law_1.01,0.3407423973083496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,4,power_law_1.2,0.10359679460525513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,32,1,power_law_1.01,0.35482239723205566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,4,power_law_1.2,0.12284799814224243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,32,1,power_law_1.01,0.41196160316467284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,4,power_law_1.2,0.1922047972679138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,32,1,power_law_1.01,0.446073579788208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,4,power_law_1.2,0.21011838912963868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,32,1,power_law_1.01,0.5429823875427247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,4,power_law_1.2,0.21385600566864013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,4,32,balanced,0.0232640008131663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,4,32,balanced,0.023205332458019257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,4,32,balanced,0.023760000864664715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,4,32,balanced,0.024800000091393787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,4,32,balanced,0.023770667612552643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,4,32,balanced,0.02548266698916753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,4,32,balanced,0.025648000339667004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,4,32,balanced,0.025279998779296875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,4,32,balanced,0.02550400048494339
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,4,32,balanced,0.027402666707833607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,4,32,balanced,0.02717333287000656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,4,32,balanced,0.027221334477265675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,4,32,balanced,0.027429332335789997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,4,32,balanced,0.03136533250411352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,4,32,balanced,0.032511999209721885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,4,32,balanced,0.03352533280849457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,4,32,balanced,0.03345066557327906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,4,power_law_1.2,0.22364161014556885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,4,32,balanced,0.03551466763019562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,32,balanced,0.03571200122435888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,32,1,power_law_1.01,0.6253376007080078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,32,balanced,0.03757333258787791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,32,balanced,0.03772799919048945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,32,balanced,0.04186133543650309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,32,balanced,0.041706666350364685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,32,balanced,0.05165866514046987
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,32,balanced,0.058677335580190025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,32,balanced,0.07735999921957652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,32,balanced,0.08963732918103536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,32,balanced,0.1037066678206126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,32,balanced,0.13962666193644205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,32,balanced,0.22573333978652954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,4,power_law_1.2,0.22472960948944093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,16,power_law_1.01,0.060038399696350095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,32,1,power_law_1.01,0.8153792381286621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,4,power_law_1.2,0.23734400272369385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,16,power_law_1.01,0.0616703987121582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,32,1,power_law_1.01,0.993990421295166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,16,power_law_1.01,0.06598399877548218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,4,power_law_1.2,0.24954240322113036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,16,power_law_1.01,0.05411199927330017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,32,1,power_law_1.01,1.3660608291625977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,4,power_law_1.2,0.2533440113067627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,16,power_law_1.01,0.056377601623535153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,4,power_law_1.2,0.2629760026931763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,32,1,power_law_1.01,1.7249664306640624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,16,power_law_1.01,0.05968639850616455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,16,power_law_1.01,0.06222720146179199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,4,power_law_1.2,0.2758016109466553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,32,1,power_law_1.01,2.4618303298950197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,16,power_law_1.01,0.06334720253944397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,4,power_law_1.2,0.28387200832366943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,8,power_law_1.01,0.04943360090255737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,16,power_law_1.01,0.06442239880561829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,8,power_law_1.01,0.07093759775161743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,4,power_law_1.2,0.28390400409698485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,32,1,power_law_1.01,3.1845951080322266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,16,power_law_1.01,0.06478719711303711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,8,power_law_1.01,0.052883201837539674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,16,power_law_1.01,0.0661184012889862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,4,power_law_1.2,0.32706561088562014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,8,power_law_1.01,0.06123520135879516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,8,power_law_1.01,0.06399359703063964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,16,power_law_1.01,0.06796799898147583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,32,1,power_law_1.01,3.910451126098633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,4,power_law_1.2,0.35731201171875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,8,power_law_1.01,0.07623680233955384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,16,power_law_1.01,0.06936320066452026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,8,power_law_1.01,0.0874176025390625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,4,power_law_1.2,0.4218560218811035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,8,power_law_1.01,0.09408640265464782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,16,power_law_1.01,0.07094399929046631
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,8,power_law_1.01,0.09269760251045227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,16,power_law_1.01,0.0769919991493225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,4,power_law_1.2,0.46776318550109863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,32,1,power_law_1.01,6.158758544921875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,8,power_law_1.01,0.09642239809036254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,8,power_law_1.01,0.0995519995689392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,16,power_law_1.01,0.07827839851379395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,4,power_law_1.2,0.5880447864532471
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,8,power_law_1.01,0.10017280578613282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,16,power_law_1.01,0.0843392014503479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,8,power_law_1.01,0.10381439924240113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,16,power_law_1.01,0.10003839731216431
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,4,power_law_1.2,0.6398719787597656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,8,power_law_1.01,0.10023679733276367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,8,power_law_1.01,0.11183359622955322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,16,power_law_1.01,0.11153279542922974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,8,power_law_1.01,0.11801600456237793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,4,power_law_1.2,0.8613247871398926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,32,1,power_law_1.01,11.818739318847657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,16,power_law_1.01,0.13404799699783326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,8,power_law_1.01,0.12545280456542968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,8,power_law_1.01,0.14087040424346925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,16,power_law_1.01,0.1549631953239441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,8,power_law_1.01,0.1589184045791626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,4,power_law_1.2,1.140345573425293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,16,power_law_1.01,0.20779519081115722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,8,power_law_1.01,0.1949056029319763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,8,power_law_1.01,0.2196671962738037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,16,power_law_1.01,0.2382335901260376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,4,power_law_1.2,1.5488896369934082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,8,power_law_1.01,0.27164158821105955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,16,power_law_1.01,0.31439359188079835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,8,power_law_1.01,0.32992000579833985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,8,power_law_1.01,0.4578495979309082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,16,power_law_1.01,0.395033597946167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,4,power_law_1.2,1.9478208541870117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,8,power_law_1.01,0.5770368099212646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,16,power_law_1.01,0.5764992237091064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,8,power_law_1.01,0.7760511875152588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,16,power_law_1.01,0.7497920036315918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,8,power_law_1.01,1.0450943946838378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,16,1,balanced,0.03761066744724909
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,16,1,balanced,0.0558186670144399
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,16,1,balanced,0.07751466830571492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,16,1,balanced,0.07821333408355713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,16,1,balanced,0.07784000039100647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,2,4,power_law_1.2,2.282099151611328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,16,1,balanced,0.07880533238252004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,16,1,balanced,0.08067733546098073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,16,1,balanced,0.07999466856320699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,16,1,balanced,0.08019733428955078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,16,1,balanced,0.08152533570925395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,16,1,balanced,0.08121599753697713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,16,1,balanced,0.08221333225568135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,16,1,balanced,0.08235733211040497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,16,1,balanced,0.08682133754094441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,16,1,balanced,0.09506133198738098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,16,1,balanced,0.09557867050170898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,16,power_law_1.01,0.8854911804199219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,16,1,balanced,0.10434133807818095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,16,1,balanced,0.1129813293615977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,8,power_law_1.01,1.220032024383545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,16,1,balanced,0.12844799955685934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,16,1,balanced,0.1551199952761332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,16,1,balanced,0.19214399655659994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,16,1,balanced,0.24664533138275146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,16,1,balanced,0.2892373402913411
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,16,1,balanced,0.3978293339411418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,16,1,balanced,0.4861546754837036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,16,1,balanced,0.7068053086598715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,16,1,balanced,0.9060640335083008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,16,1,balanced,1.1061333020528157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,16,1,balanced,1.7260799407958984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,8,power_law_1.01,2.0103872299194334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,16,power_law_1.01,1.367193603515625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,16,1,balanced,3.388186772664388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,2,4,power_law_1.2,3.6523391723632814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,8,power_law_1.01,3.9340606689453126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,16,power_law_1.01,2.7467840194702147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,2,4,power_law_1.2,7.686137390136719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,2,4,power_law_1.01,0.11163519620895386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,2,4,power_law_1.01,0.10024319887161255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,2,4,power_law_1.01,0.11730560064315795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,2,4,power_law_1.01,0.1183359980583191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,2,4,power_law_1.01,0.11247999668121338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,2,4,power_law_1.01,0.12726399898529053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,2,4,power_law_1.01,0.14507520198822021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,2,4,power_law_1.01,0.12940160036087037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,2,4,power_law_1.01,0.13897600173950195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,2,4,power_law_1.01,0.13763200044631957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,2,4,power_law_1.01,0.1381824016571045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,2,4,power_law_1.01,0.14257279634475709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,2,4,power_law_1.01,0.14805760383605956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,2,4,power_law_1.01,0.1715456008911133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,2,4,power_law_1.01,0.17872639894485473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,32,2,power_law_1.01,0.030156800150871278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,2,4,power_law_1.01,0.19625600576400756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,32,2,power_law_1.01,0.030092799663543703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,2,4,power_law_1.01,0.19541759490966798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,32,2,power_law_1.01,0.0297791987657547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,2,4,power_law_1.01,0.25771520137786863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,32,2,power_law_1.01,0.030086401104927062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,32,power_law_1.2,0.07687680125236511
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,32,2,power_law_1.01,0.030534398555755616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,2,4,power_law_1.01,0.30658559799194335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,32,2,power_law_1.01,0.03130879998207092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,32,2,power_law_1.01,0.032390400767326355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,2,4,power_law_1.01,0.4087552070617676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,32,2,power_law_1.01,0.033024001121521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,2,4,power_law_1.01,0.43614721298217773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,32,2,power_law_1.01,0.03284479975700379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,32,power_law_1.2,0.08346239924430847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,32,2,power_law_1.01,0.0336896002292633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,2,4,power_law_1.01,0.6515456199645996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,32,2,power_law_1.01,0.03378559947013855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,32,2,power_law_1.01,0.0344895988702774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,32,2,power_law_1.01,0.03612799942493439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,2,4,power_law_1.01,0.8078911781311036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,32,power_law_1.2,0.05048959851264954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,32,2,power_law_1.01,0.03929600119590759
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,32,2,power_law_1.01,0.043935999274253845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,32,2,power_law_1.01,0.045747199654579164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,2,4,power_law_1.01,1.142188835144043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,32,2,power_law_1.01,0.047443199157714847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,32,2,power_law_1.01,0.05751680135726929
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,32,power_law_1.2,0.05112959742546082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,32,2,power_law_1.01,0.06547200083732604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,2,4,power_law_1.01,2.1616447448730467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,32,2,power_law_1.01,0.07989760041236878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,32,2,power_law_1.01,0.10579839944839478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,32,2,power_law_1.01,0.14017280340194702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,32,power_law_1.2,0.05915520191192627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,32,2,power_law_1.01,0.15940480232238768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,2,4,power_law_1.01,2.440339279174805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,32,2,power_law_1.01,0.21101438999176025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,32,2,power_law_1.01,0.2881727933883667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,1,balanced,0.05062933266162872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,32,power_law_1.2,0.06555520296096802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,32,2,power_law_1.01,0.3831808090209961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,2,4,power_law_1.01,2.430860710144043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,1,balanced,0.054560000697771706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,1,balanced,0.0668213317791621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,1,balanced,0.0921493371327718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,1,balanced,0.13657066226005554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,32,2,power_law_1.01,0.5732800006866455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,1,balanced,0.15244799852371216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,1,balanced,0.15548800428708395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,1,balanced,0.16014933586120605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,1,balanced,0.1620213290055593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,1,balanced,0.16155733664830527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,1,balanced,0.16703999042510986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,32,2,power_law_1.01,0.6841792106628418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,1,balanced,0.17361066738764444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,32,power_law_1.2,0.06687359809875489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,1,balanced,0.17376534144083658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,1,balanced,0.18686399857203165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,1,balanced,0.19569599628448486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,1,balanced,0.20361600319544473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,1,balanced,0.21513599157333374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,2,4,power_law_1.01,4.21082878112793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,1,balanced,0.2550666729609172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,32,2,power_law_1.01,0.9080127716064453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,1,balanced,0.28037333488464355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,1,balanced,0.3957333167394002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,1,balanced,0.4234400192896525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,32,power_law_1.2,0.07082239985466003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,1,balanced,0.6308533350626627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,1,balanced,0.726149320602417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,1,balanced,1.1280960241953533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,1,balanced,1.3390612602233887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,32,2,power_law_1.01,2.362700843811035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,32,power_law_1.2,0.07192959785461425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,1,balanced,2.0600372950236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,2,4,power_law_1.01,6.0905406951904295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,1,balanced,2.5621066093444824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,32,power_law_1.2,0.07258880138397217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,1,balanced,3.069141387939453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,32,power_law_1.2,0.0752128005027771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,1,balanced,4.892330805460612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,32,power_law_1.2,0.07920640110969543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,32,power_law_1.2,0.08279039859771728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,1,balanced,9.433263778686523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,2,4,power_law_1.01,14.995916748046875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,32,power_law_1.2,0.0807424008846283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,32,power_law_1.2,0.0887935996055603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,32,power_law_1.2,0.09449599981307984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,32,power_law_1.2,0.09950079917907714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,32,power_law_1.2,0.11454720497131347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,32,power_law_1.2,0.1223039984703064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,32,power_law_1.2,0.15418879985809325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,32,power_law_1.2,0.17320319414138793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,1,power_law_1.2,0.049516800045967105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,32,power_law_1.2,0.22702720165252685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,1,power_law_1.2,0.07134079933166504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,32,power_law_1.2,0.27200000286102294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,1,power_law_1.2,0.07905920147895813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,32,power_law_1.2,0.3780479907989502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,64,balanced,0.06491200129191081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,64,balanced,0.06559999783833821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,1,power_law_1.2,0.12303359508514404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,64,balanced,0.06634133557478587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,64,balanced,0.06449600060780843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,4,balanced,0.07418666779994965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,64,balanced,0.07442133128643036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,64,balanced,0.09644266963005066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,4,balanced,0.10030399759610494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,64,balanced,0.10546666383743286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,64,balanced,0.10893332958221436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,4,balanced,0.1516693333784739
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,64,balanced,0.10698133707046509
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,64,balanced,0.10725866754849751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,64,balanced,0.11015466849009196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,4,balanced,0.24723732471466064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,64,balanced,0.10912533601125081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,64,balanced,0.11333333452542622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,4,balanced,0.43934933344523114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,64,balanced,0.11222933729489644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,64,balanced,0.11241066455841064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,64,balanced,0.11693867047627766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,64,balanced,0.12454932928085327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,4,balanced,0.82696533203125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,64,balanced,0.12601600090662637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,1,power_law_1.2,0.15972479581832885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,32,power_law_1.2,0.4793407917022705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,64,balanced,0.1299626628557841
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,64,balanced,0.14178133010864258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,4,balanced,0.8275093237559
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,64,balanced,0.1543786625067393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,64,balanced,0.18174932400385538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,4,balanced,0.8294133345286051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,64,balanced,0.20470933119455972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,64,balanced,0.2574933369954427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,4,balanced,0.8314080238342285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,64,balanced,0.3318293293317159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,4,balanced,0.8342933654785156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,64,balanced,0.4273013273874919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,64,balanced,0.5327039957046509
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,1,power_law_1.2,0.21473920345306396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,4,balanced,0.8331946531931559
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,64,balanced,0.6699146429697672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,4,balanced,0.8376586437225342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,32,power_law_1.2,0.7479104042053223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,4,balanced,0.8425599733988444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,64,balanced,1.0010826587677002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,4,balanced,0.841866652170817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,4,balanced,0.8543573220570883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,64,balanced,1.9074719746907551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,1,power_law_1.2,0.31972479820251465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,4,balanced,0.8554666837056478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,4,balanced,0.8623039722442627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,4,balanced,0.879413366317749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,4,balanced,0.899610678354899
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,1,power_law_1.2,0.34272639751434325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,4,balanced,0.9223039944966634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,32,power_law_1.2,0.9943424224853515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,4,balanced,0.9719839890797933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,4,balanced,1.01801602045695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,4,balanced,1.062058687210083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,1,power_law_1.2,0.35822720527648927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,4,balanced,1.2986079851786296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,4,balanced,1.3287413120269775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,4,balanced,1.7968212763468425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,32,power_law_1.2,1.190726375579834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,1,power_law_1.2,0.3625983953475952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,4,balanced,2.002357323964437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,4,balanced,2.635157267252604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,1,power_law_1.2,0.39229440689086914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,4,balanced,3.539519945780436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,1,power_law_1.2,0.40750718116760254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,32,power_law_1.2,1.953343963623047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,4,balanced,6.994768142700195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,1,power_law_1.2,0.4231103897094727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,1,power_law_1.2,0.43319039344787597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,1,power_law_1.2,0.4538303852081299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,32,power_law_1.2,3.4623233795166017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,1,power_law_1.2,0.4715007781982422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,1,power_law_1.2,0.5061439990997314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,1,power_law_1.2,0.5674176216125488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,1,power_law_1.2,0.6184063911437988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,16,power_law_1.2,0.08094720244407654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,1,power_law_1.2,0.7272895812988281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,1,power_law_1.2,0.8501312255859375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,16,power_law_1.2,0.07175679802894593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,1,power_law_1.2,1.0662272453308106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,16,power_law_1.2,0.06352639794349671
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,1,power_law_1.2,1.1980031967163085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,8,balanced,0.04539200166861216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,8,balanced,0.047557334105173744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,8,balanced,0.04894400139649709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,8,balanced,0.04959466556708018
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,8,balanced,0.05778666834036509
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,8,balanced,0.08262933293978374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,8,balanced,0.08333866794904073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,8,balanced,0.08390399813652039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,16,power_law_1.2,0.06624000072479248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,8,balanced,0.08513599634170532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,1,power_law_1.2,1.5698880195617675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,8,balanced,0.08570133646329244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,8,balanced,0.08795733253161113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,8,balanced,0.08755733569463094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,8,balanced,0.08993599812189738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,8,balanced,0.09061333537101746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,8,balanced,0.0978666643301646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,8,balanced,0.10171199838320415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,8,balanced,0.1072160005569458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,8,balanced,0.11746133367220561
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,8,balanced,0.12852266430854797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,8,balanced,0.1497173309326172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,8,balanced,0.1821546753247579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,8,balanced,0.22404799858729044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,1,power_law_1.2,1.9370624542236328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,16,power_law_1.2,0.07255679965019227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,8,balanced,0.2654506762822469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,8,balanced,0.35970131556193036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,8,balanced,0.45498132705688477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,8,balanced,0.642901341120402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,8,balanced,0.8258826732635498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,8,balanced,1.0101280212402344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,8,balanced,1.5525813102722168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,2,power_law_1.2,0.050892800092697144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,16,power_law_1.2,0.08676480054855347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,1,power_law_1.2,2.6594879150390627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,8,balanced,3.0379838943481445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,2,power_law_1.2,0.08483840227127075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,16,power_law_1.2,0.0884223997592926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,1,power_law_1.2,3.3911617279052733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,2,power_law_1.2,0.07498239874839782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,16,power_law_1.2,0.09175040125846863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,2,power_law_1.2,0.0871616005897522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,1,power_law_1.2,4.1534782409667965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,16,power_law_1.2,0.09548159837722778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,2,power_law_1.2,0.12478079795837402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,16,power_law_1.2,0.09374719858169556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,2,power_law_1.2,0.1581055998802185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,1,power_law_1.2,6.206329727172852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,2,power_law_1.2,0.18083839416503905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,16,power_law_1.2,0.09761919975280761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,2,power_law_1.2,0.2386944055557251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,16,power_law_1.2,0.10043519735336304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,2,power_law_1.2,0.24983038902282714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,16,power_law_1.2,0.10186879634857178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,2,power_law_1.2,0.2515007972717285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,1,power_law_1.2,11.967238616943359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,16,power_law_1.2,0.10632959604263306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,2,power_law_1.2,0.2671744108200073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,16,power_law_1.2,0.1130944013595581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,2,power_law_1.2,0.27985920906066897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,16,power_law_1.01,0.0575872004032135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,16,power_law_1.2,0.1152959942817688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,2,power_law_1.2,0.2916863918304443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,16,power_law_1.01,0.06450560092926025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,2,power_law_1.2,0.30951039791107177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,16,power_law_1.2,0.12444800138473511
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,16,power_law_1.01,0.053523200750350955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,2,power_law_1.2,0.32647678852081297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,16,power_law_1.2,0.14028799533843994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,2,power_law_1.2,0.33921279907226565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,16,power_law_1.01,0.05299839973449707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,16,power_law_1.2,0.16149120330810546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,2,power_law_1.2,0.3559423923492432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,16,power_law_1.01,0.05422719717025757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,2,power_law_1.2,0.3904063940048218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,16,power_law_1.2,0.20668160915374756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,16,power_law_1.01,0.058233600854873654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,1,balanced,0.07453866799672444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,1,balanced,0.09716266393661499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,1,balanced,0.1511893371740977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,2,power_law_1.2,0.43148159980773926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,1,balanced,0.2437600096066793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,1,balanced,0.4408053159713745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,1,balanced,0.8303306897481283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,1,balanced,1.2245439688364665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,16,power_law_1.2,0.23160960674285888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,16,power_law_1.01,0.05971840023994446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,1,balanced,1.2231733004252117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,1,balanced,1.2304373582204182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,1,balanced,1.2350880304972331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,2,power_law_1.2,0.5026944160461426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,1,balanced,1.2442293167114258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,1,balanced,1.2542400360107422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,1,balanced,1.257754643758138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,16,power_law_1.01,0.06106879711151123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,1,balanced,1.264677365620931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,16,power_law_1.2,0.3050623893737793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,1,balanced,1.2789973417917888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,2,power_law_1.2,0.5835328102111816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,1,balanced,1.2902666727701824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,1,balanced,1.3075679937998455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,1,balanced,1.3482346534729004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,1,balanced,1.386858622233073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,16,power_law_1.01,0.060703998804092406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,1,balanced,1.4698506991068523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,1,balanced,1.5452747344970703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,2,power_law_1.2,0.7360127925872803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,16,power_law_1.2,0.3400192022323608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,1,balanced,1.6732959747314453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,1,balanced,1.8187039693196614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,16,power_law_1.01,0.06098560094833374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,1,balanced,2.113957405090332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,1,balanced,2.5563413302103677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,2,power_law_1.2,0.909331226348877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,1,balanced,3.1292746861775718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,16,power_law_1.01,0.0627776026725769
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,16,power_law_1.2,0.46195201873779296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,1,balanced,4.175360043843587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,16,power_law_1.01,0.06394879817962647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,2,power_law_1.2,1.1842559814453124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,1,balanced,5.632138570149739
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,16,power_law_1.2,0.6142848014831543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,16,power_law_1.01,0.06474239826202392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,1,balanced,7.263834635416667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,2,power_law_1.2,1.3584768295288085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,16,power_law_1.01,0.06735360026359558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,2,power_law_1.2,1.8345407485961913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,16,power_law_1.2,0.883084774017334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,16,power_law_1.01,0.07406079769134521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,1,balanced,13.728293100992838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,16,power_law_1.01,0.07482240200042725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,2,power_law_1.2,2.445484733581543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,16,power_law_1.01,0.0824832022190094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,16,power_law_1.2,1.1491647720336915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,16,power_law_1.01,0.0960640013217926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,2,power_law_1.2,2.99484806060791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,16,power_law_1.01,0.10220160484313964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,16,power_law_1.01,0.1279744029045105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,2,16,power_law_1.2,1.4416704177856445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,2,power_law_1.2,4.379852676391602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,16,power_law_1.01,0.15096960067749024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,16,power_law_1.01,0.1838912010192871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,16,power_law_1.01,0.21487998962402344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,2,16,power_law_1.2,2.751692771911621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,8,balanced,0.051882664362589516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,8,balanced,0.05373333394527435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,8,balanced,0.055957332253456116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,8,balanced,0.056314667065938316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,8,balanced,0.0639626681804657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,8,balanced,0.07452266911665599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,8,balanced,0.08683733145395915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,8,balanced,0.0860746701558431
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,8,balanced,0.08683733145395915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,8,balanced,0.08851200342178345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,8,balanced,0.08726400136947632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,16,power_law_1.01,0.2896255970001221
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,8,balanced,0.09072533249855042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,8,balanced,0.09169600407282512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,8,balanced,0.09276266892751057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,8,balanced,0.09902933239936829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,8,balanced,0.10126399993896484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,8,balanced,0.10763200124104817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,8,balanced,0.12155733505884807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,2,power_law_1.2,8.093011474609375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,8,balanced,0.13144000371297201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,8,balanced,0.15760533014933267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,8,balanced,0.172650674978892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,8,balanced,0.21185066302617392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,8,balanced,0.25979199012120563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,8,balanced,0.3381439844767253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,8,balanced,0.4465706745783488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,8,balanced,0.609994649887085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,8,balanced,0.7969439824422201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,16,power_law_1.01,0.3700927972793579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,8,balanced,0.9832692941029867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,8,balanced,1.5202879905700684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,8,balanced,2.956218719482422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,16,power_law_1.01,0.5088895797729492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,16,power_law_1.01,0.6591936111450195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,16,power_law_1.01,0.7694719791412353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,2,16,power_law_1.2,5.000537490844726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,16,power_law_1.01,1.208140754699707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,16,power_law_1.01,2.5602495193481447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,32,balanced,0.04818666477998098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,32,balanced,0.04977599779764811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,32,balanced,0.05363733569780985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,32,balanced,0.06836266815662384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,32,balanced,0.08089600006739299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,32,balanced,0.08255466818809509
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,32,balanced,0.08249600231647491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,32,balanced,0.08074666559696198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,32,balanced,0.08162133395671844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,32,balanced,0.08081600069999695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,32,balanced,0.08087466657161713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,32,balanced,0.08459200461705525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,32,balanced,0.08375466863314311
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,32,balanced,0.08551999926567078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,32,balanced,0.09086933732032776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,32,balanced,0.09095999598503113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,32,balanced,0.09278399745623271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,32,balanced,0.11327466368675232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,32,balanced,0.11144000291824341
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,32,balanced,0.1351626714070638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,32,balanced,0.1524799962838491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,32,balanced,0.19922133286794028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,32,balanced,0.18186134099960327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,32,balanced,0.23846399784088135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,32,balanced,0.26467732588450116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,32,balanced,0.3601440191268921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,32,balanced,0.41792531808217365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,16,balanced,0.037674665451049805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,4,32,balanced,0.5074719985326132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,16,balanced,0.03540800015131632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,16,balanced,0.03533333291610082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,16,balanced,0.03749866783618927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,16,balanced,0.037658666570981346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,4,32,balanced,0.72652268409729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,16,balanced,0.03955200066169103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,16,balanced,0.03963200002908707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,16,balanced,0.03953066716591517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,16,balanced,0.04106133431196213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,16,balanced,0.03958933303753535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,16,balanced,0.03996799886226654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,16,balanced,0.04136000076929728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,16,balanced,0.04195733368396759
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,16,balanced,0.04365866879622141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,4,32,balanced,1.5454613367716472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,16,balanced,0.04541333516438802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,16,balanced,0.047775998711586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,16,balanced,0.04800533254941305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,16,balanced,0.052149335543314614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,16,balanced,0.052095999320348106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,16,balanced,0.0582239975531896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,16,balanced,0.06206400195757548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,16,balanced,0.07674133280913036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,16,balanced,0.09071999788284302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,16,balanced,0.11967999736467998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,16,balanced,0.15003732840220133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,16,balanced,0.19657599925994873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,16,balanced,0.24045334259668985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,16,balanced,0.29176533222198486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,16,balanced,0.43012801806132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,16,balanced,0.8144853115081787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,1,power_law_1.01,0.07370880246162415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,1,power_law_1.01,0.07950080037117005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,1,power_law_1.01,0.09039360284805298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,1,power_law_1.01,0.11932799816131592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,1,power_law_1.01,0.15194239616394042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,8,8,power_law_1.2,0.04470399916172028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,1,power_law_1.01,0.1872383952140808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,8,8,power_law_1.2,0.04618239998817444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,8,8,power_law_1.2,0.03455359935760498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,8,8,power_law_1.2,0.03493120074272156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,1,power_law_1.01,0.23221120834350586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,8,8,power_law_1.2,0.03451519906520843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,8,8,power_law_1.2,0.035097599029541016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,8,8,power_law_1.2,0.03528960049152374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,1,power_law_1.01,0.24915199279785155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,8,8,power_law_1.2,0.038150399923324585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,8,8,power_law_1.2,0.040575999021530154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,8,8,power_law_1.2,0.04472320079803467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,1,power_law_1.01,0.2533951997756958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,8,8,power_law_1.2,0.04293760061264038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,8,8,power_law_1.2,0.05159040093421936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,8,8,power_law_1.2,0.050835198163986205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,8,8,power_law_1.2,0.04950399994850159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,1,power_law_1.01,0.26379520893096925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,8,8,power_law_1.2,0.06373760104179382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,8,8,power_law_1.2,0.06507520079612732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,32,8,balanced,0.04177066683769226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,32,8,balanced,0.04365866879622141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,32,8,balanced,0.04561600089073181
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,32,8,balanced,0.04757333298524221
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,8,8,power_law_1.2,0.07445120215415954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,32,8,balanced,0.04985600213209788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,1,power_law_1.01,0.27130239009857177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,32,8,balanced,0.06021333237489065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,32,8,balanced,0.06617600222428639
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,32,8,balanced,0.068271999557813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,1,power_law_1.2,0.09810559749603272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,32,8,balanced,0.07030933101971944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,8,8,power_law_1.2,0.0886784017086029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,32,8,balanced,0.06965866684913635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,32,8,balanced,0.06870399912198384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,32,8,balanced,0.07053333520889282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,32,8,balanced,0.07243200143178304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,32,8,balanced,0.07682133217652638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,32,8,balanced,0.08250666658083598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,8,8,power_law_1.2,0.10046080350875855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,32,8,balanced,0.08493333061536153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,32,8,balanced,0.09273067116737366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,32,8,balanced,0.10761599739392598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,32,8,balanced,0.12144000331560771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,8,8,power_law_1.2,0.12357120513916016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,32,8,balanced,0.14805866281191507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,1,power_law_1.01,0.2773184061050415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,32,8,balanced,0.16899732748667398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,32,8,balanced,0.2156533400217692
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,32,8,balanced,0.2761066754659017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,8,8,power_law_1.2,0.14177279472351073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,32,8,balanced,0.3736533323923747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,32,8,balanced,0.48535998662312824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,1,power_law_1.2,0.14582400321960448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,8,8,power_law_1.2,0.17945599555969238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,32,8,balanced,0.6986666520436605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,32,8,balanced,0.8940853277842203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,1,power_law_1.01,0.2818943977355957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,8,8,power_law_1.2,0.2396159887313843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,32,8,balanced,1.1104053656260173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,8,8,power_law_1.2,0.3136447906494141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,32,8,balanced,1.7106560071309407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,8,8,power_law_1.2,0.40268797874450685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,1,power_law_1.01,0.29607040882110597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,1,power_law_1.2,0.20105600357055664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,32,8,balanced,3.3642454147338867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,8,8,power_law_1.2,0.5880064010620117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,1,power_law_1.01,0.30725760459899903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,8,8,power_law_1.2,0.7578368186950684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,1,power_law_1.2,0.3077440023422241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,8,8,power_law_1.2,0.8800704002380371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,1,power_law_1.01,0.31612799167633054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,8,8,power_law_1.2,1.2491456031799317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,1,power_law_1.01,0.33057279586791993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,1,power_law_1.2,0.5275263786315918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,8,8,power_law_1.2,2.635897636413574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,1,power_law_1.01,0.3814912080764771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,1,power_law_1.2,0.8101823806762696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,1,power_law_1.01,0.41590399742126466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,1,power_law_1.01,0.5080128192901612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,1,power_law_1.2,0.939417552947998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,1,power_law_1.01,0.5860095977783203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,1,power_law_1.2,1.3947392463684083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,1,power_law_1.01,0.7598976135253906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,1,power_law_1.2,1.473516845703125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,1,power_law_1.01,0.9268671989440918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,1,power_law_1.01,1.2644288063049316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,1,power_law_1.2,1.554918384552002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,2,8,balanced,0.0798773318529129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,2,8,balanced,0.12132267157236735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,2,8,balanced,0.07934933404127757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,2,8,balanced,0.08045333127180736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,2,8,balanced,0.08105599880218506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,2,8,balanced,0.08130133152008057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,2,8,balanced,0.08078933258851369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,2,8,balanced,0.07852266728878021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,2,8,balanced,0.080335999528567
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,2,8,balanced,0.0798773318529129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,1,power_law_1.01,1.633932876586914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,2,8,balanced,0.08282666901747386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,2,8,balanced,0.08265600105126698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,2,8,balanced,0.08156799773375194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,2,8,balanced,0.08493333061536153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,2,8,balanced,0.09517866373062134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,2,8,balanced,0.09307199716567993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,2,8,balanced,0.09483733773231506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,2,8,balanced,0.11374933520952861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,2,8,balanced,0.11372266213099162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,2,8,balanced,0.1276853382587433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,1,power_law_1.2,1.6514816284179688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,2,8,balanced,0.1523413360118866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,2,8,balanced,0.18563199043273926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,2,8,balanced,0.22856533527374268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,2,8,balanced,0.3065279920895894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,2,8,balanced,0.36926400661468506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,2,8,balanced,0.528005321820577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,2,8,balanced,0.7046826680501302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,2,8,balanced,0.8472266991933187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,2,8,balanced,1.35590394337972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,1,power_law_1.01,2.3305471420288084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,1,power_law_1.2,1.6979135513305663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,2,8,balanced,2.6007307370503745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,1,power_law_1.01,2.9391103744506837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,1,power_law_1.2,1.777644729614258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,4,power_law_1.2,0.1604864001274109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,1,power_law_1.2,1.8679168701171875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,4,power_law_1.2,0.12297600507736206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,1,power_law_1.01,3.587910461425781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,4,power_law_1.2,0.1525823950767517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,1,power_law_1.2,1.9327743530273438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,4,power_law_1.2,0.24341120719909667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,1,power_law_1.2,1.9400192260742188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,1,power_law_1.01,5.646931076049805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,4,power_law_1.2,0.2725311994552612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,4,power_law_1.2,0.4616511821746826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,1,power_law_1.2,2.09816951751709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,4,power_law_1.2,0.4639616012573242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,1,power_law_1.2,2.2612287521362306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,4,power_law_1.2,0.5158463954925537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,2,1,power_law_1.2,0.05251839756965637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,2,1,power_law_1.2,0.05907840132713318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,2,1,power_law_1.2,0.06528000235557556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,2,1,power_law_1.2,0.09753599762916565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,4,power_law_1.2,0.5312511920928955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,1,power_law_1.01,10.759295654296874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,2,1,power_law_1.2,0.09681280255317688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,1,power_law_1.2,2.424678421020508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,2,1,power_law_1.2,0.09873279929161072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,2,1,power_law_1.2,0.09829760193824769
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,2,1,power_law_1.2,0.09971839785575867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,2,1,power_law_1.2,0.1013375997543335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,4,power_law_1.2,0.5372352123260498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,2,1,power_law_1.2,0.10182399749755859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,2,1,power_law_1.2,0.10357760190963745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,2,1,power_law_1.2,0.1086016058921814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,2,1,power_law_1.2,0.10847359895706177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,2,1,power_law_1.2,0.1152448058128357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,1,power_law_1.2,2.618943977355957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,2,1,power_law_1.2,0.12872960567474365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,4,power_law_1.2,0.5300864219665528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,2,1,power_law_1.2,0.13274240493774414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,2,1,power_law_1.2,0.13781119585037233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,2,1,power_law_1.2,0.16441600322723388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,2,1,power_law_1.2,0.18766720294952394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,2,1,power_law_1.2,0.2342463970184326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,4,power_law_1.2,0.5639296054840088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,2,1,power_law_1.2,0.27966721057891847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,2,1,power_law_1.2,0.3752448081970215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,1,power_law_1.2,2.8039039611816405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,2,1,power_law_1.2,0.4614399909973145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,4,power_law_1.2,0.5511744022369385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,2,1,power_law_1.2,0.6483136177062988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,2,1,power_law_1.2,0.8088128089904785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,2,1,power_law_1.2,1.150278377532959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,4,power_law_1.2,0.593939208984375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,2,1,power_law_1.2,1.4831616401672363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,1,power_law_1.2,3.3575233459472655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,4,power_law_1.2,0.6169727802276611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,2,1,power_law_1.2,1.8327360153198242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,4,power_law_1.2,0.6228608131408692
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,2,1,power_law_1.2,2.8444543838500977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,1,power_law_1.2,3.4196033477783203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,4,power_law_1.2,0.6533120155334473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,2,1,power_law_1.2,5.543174362182617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,4,power_law_1.2,0.7388095855712891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,1,power_law_1.2,4.197715377807617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,4,power_law_1.2,0.7876992225646973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,4,power_law_1.2,0.9437760353088379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,2,power_law_1.2,0.032179200649261476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,2,power_law_1.2,0.0389055997133255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,4,power_law_1.2,0.8731391906738282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,1,power_law_1.2,4.651929473876953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,2,power_law_1.2,0.04442879855632782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,2,power_law_1.2,0.05550720095634461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,2,power_law_1.2,0.06431999802589417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,4,power_law_1.2,1.0522303581237793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,2,power_law_1.2,0.10478719472885131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,2,power_law_1.2,0.10631040334701539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,4,power_law_1.2,1.2048640251159668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,2,power_law_1.2,0.11008640527725219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,1,power_law_1.2,5.842303848266601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,2,power_law_1.2,0.11183359622955322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,4,power_law_1.2,1.4645183563232422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,2,power_law_1.2,0.11089919805526734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,2,power_law_1.2,0.11866240501403809
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,2,power_law_1.2,0.11822719573974609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,4,power_law_1.2,1.8292160034179688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,2,power_law_1.2,0.12263679504394531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,2,power_law_1.2,0.12624000310897826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,1,power_law_1.2,7.03597412109375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,2,power_law_1.2,0.1302847981452942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,4,power_law_1.2,2.642905616760254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,2,power_law_1.2,0.13786239624023439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,2,power_law_1.2,0.14905600547790526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,2,1,power_law_1.01,0.05115519762039185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,2,1,power_law_1.01,0.061561602354049685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,2,power_law_1.2,0.1691200017929077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,2,1,power_law_1.01,0.06936960220336914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,2,1,power_law_1.01,0.0932096004486084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,2,1,power_law_1.01,0.09644160270690919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,2,power_law_1.2,0.16135040521621705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,2,1,power_law_1.01,0.0985152006149292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,4,power_law_1.2,3.5345535278320312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,2,1,power_law_1.01,0.09901440143585205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,2,1,power_law_1.01,0.10101759433746338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,2,power_law_1.2,0.19037439823150634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,2,1,power_law_1.01,0.09945600032806397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,2,1,power_law_1.01,0.10064640045166015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,1,power_law_1.2,8.447615814208984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,2,1,power_law_1.01,0.10492160320281982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,2,power_law_1.2,0.21530239582061766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,2,1,power_law_1.01,0.10867199897766114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,2,1,power_law_1.01,0.11006720066070556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,2,1,power_law_1.01,0.1175104022026062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,2,power_law_1.2,0.2686527967453003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,2,1,power_law_1.01,0.12872960567474365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,2,1,power_law_1.01,0.13186559677124024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,2,1,power_law_1.01,0.14012160301208496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,2,1,power_law_1.01,0.16380159854888915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,2,power_law_1.2,0.2857664108276367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,4,power_law_1.2,3.913158416748047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,2,1,power_law_1.01,0.18628480434417724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,2,1,power_law_1.01,0.2351423978805542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,2,power_law_1.2,0.33761279582977294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,2,1,power_law_1.01,0.28139519691467285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,2,1,power_law_1.01,0.3734463930130005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,2,1,power_law_1.01,0.4613952159881592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,2,power_law_1.2,0.4229311943054199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,2,1,power_law_1.01,0.6471424102783203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,2,1,power_law_1.01,0.7985663890838623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,2,power_law_1.2,0.562879991531372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,2,1,power_law_1.01,1.1520895957946777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,2,1,power_law_1.01,1.4846015930175782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,4,power_law_1.2,6.2098430633544925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,1,power_law_1.2,11.628896331787109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,2,power_law_1.2,0.7569024085998535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,2,1,power_law_1.01,1.8327808380126953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,2,2,power_law_1.2,0.8028736114501953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,2,1,power_law_1.01,2.847084808349609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,2,2,power_law_1.2,1.3116800308227539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,2,1,power_law_1.01,5.565356826782226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,2,2,power_law_1.2,2.600204849243164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,4,power_law_1.2,12.492371368408204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,1,power_law_1.2,21.31360626220703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,1,2,power_law_1.2,0.12416640520095826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,1,2,power_law_1.2,0.13471360206604005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,1,2,power_law_1.2,0.1720703959465027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,1,2,power_law_1.2,0.23276159763336182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,1,2,power_law_1.2,0.23287680149078369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,1,2,power_law_1.2,0.23688960075378418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,1,2,power_law_1.2,0.24277119636535643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,1,2,power_law_1.2,0.2690623998641968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,4,8,power_law_1.01,0.027219200134277345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,1,power_law_1.01,0.049772799015045166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,1,2,power_law_1.2,0.28113279342651365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,1,2,power_law_1.2,0.31562879085540774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,4,8,power_law_1.01,0.023865599930286408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,1,2,power_law_1.2,0.27314560413360595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,1,2,power_law_1.2,0.2774784088134766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,4,8,power_law_1.01,0.022150400280952453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,1,2,power_law_1.2,0.2819839954376221
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,1,2,power_law_1.2,0.29843840599060056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,1,power_law_1.01,0.05389440059661865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,1,2,power_law_1.2,0.3037760019302368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,4,8,power_law_1.01,0.022355200350284578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,1,2,power_law_1.2,0.32688639163970945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,1,2,power_law_1.2,0.3739392042160034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,4,8,power_law_1.01,0.025356799364089966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,2,4,power_law_1.01,0.05161600112915039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,1,2,power_law_1.2,0.45728001594543455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,2,4,power_law_1.01,0.04853120148181915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,1,power_law_1.01,0.062463998794555664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,1,2,power_law_1.2,0.49536638259887694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,2,4,power_law_1.01,0.04542720019817352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,4,8,power_law_1.01,0.0265855997800827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,2,4,power_law_1.01,0.04733439981937408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,1,2,power_law_1.2,0.6356287956237793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,2,4,power_law_1.01,0.04775680005550385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,4,8,power_law_1.01,0.026643198728561402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,1,2,power_law_1.2,0.7662847995758056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,2,4,power_law_1.01,0.048332801461219786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,2,4,power_law_1.01,0.04837760031223297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,1,power_law_1.01,0.08321920037269592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,1,2,power_law_1.2,0.967251205444336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,2,4,power_law_1.01,0.04943360090255737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,4,8,power_law_1.01,0.02744320034980774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,2,4,power_law_1.01,0.05187839865684509
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,2,4,power_law_1.01,0.055961602926254274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,1,2,power_law_1.2,1.1195712089538574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,2,4,power_law_1.01,0.05624960064888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,4,8,power_law_1.01,0.029747200012207032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,1,power_law_1.01,0.09975680112838745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,2,4,power_law_1.01,0.05834239721298218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,2,4,power_law_1.01,0.06159999966621399
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,1,2,power_law_1.2,1.7347776412963867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,4,8,power_law_1.01,0.03028480112552643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,2,4,power_law_1.01,0.06592640280723572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,2,4,power_law_1.01,0.07520639896392822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,2,4,power_law_1.01,0.07446399927139283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,1,power_law_1.01,0.14590079784393312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,4,8,power_law_1.01,0.03155199885368347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,1,2,power_law_1.2,2.4151487350463867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,2,4,power_law_1.01,0.08759679794311523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,2,4,power_law_1.01,0.09949439764022827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,4,8,power_law_1.01,0.03292160034179688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,2,4,power_law_1.01,0.11452159881591797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,2,4,power_law_1.01,0.1408128023147583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,1,power_law_1.01,0.15256960391998292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,2,4,power_law_1.01,0.16844160556793214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,4,8,power_law_1.01,0.03284479975700379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,1,2,power_law_1.2,3.3991424560546877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,2,4,power_law_1.01,0.2201024055480957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,2,4,power_law_1.01,0.2766272068023682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,4,8,power_law_1.01,0.0338239997625351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,1,power_law_1.01,0.15733120441436768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,2,4,power_law_1.01,0.39955840110778806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,4,8,power_law_1.01,0.03744640052318573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,2,4,power_law_1.01,0.5086207866668702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,1,2,power_law_1.2,4.0031169891357425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,1,power_law_1.01,0.15861120223999023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,2,4,power_law_1.01,0.7515200138092041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,4,8,power_law_1.01,0.038227200508117676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,2,4,power_law_1.01,0.9697855949401856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,4,8,power_law_1.01,0.04320000112056732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,1,power_law_1.01,0.16406400203704835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,2,4,power_law_1.01,1.3433216094970704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,4,8,power_law_1.01,0.04764800071716309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,1,2,power_law_1.2,6.542675018310547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,1,power_law_1.01,0.16915839910507202
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,8,power_law_1.01,0.0531391978263855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,2,4,power_law_1.01,1.961952018737793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,1,power_law_1.01,0.17360639572143555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,8,power_law_1.01,0.06000000238418579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,8,power_law_1.01,0.06510080099105835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,1,power_law_1.01,0.1761728048324585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,8,power_law_1.01,0.08075519800186157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,2,4,power_law_1.01,4.363180923461914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,1,power_law_1.01,0.18750720024108886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,1,2,power_law_1.2,11.130169677734376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,8,power_law_1.01,0.0933247983455658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,1,power_law_1.01,0.19100799560546874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,8,power_law_1.01,0.12282240390777588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,1,power_law_1.01,0.20021119117736816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,8,power_law_1.01,0.16965119838714598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,1,power_law_1.01,0.21241600513458253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,8,power_law_1.01,0.21676158905029297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,1,power_law_1.01,0.24228479862213134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,8,power_law_1.01,0.28550400733947756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,1,power_law_1.01,0.2704319953918457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,1,power_law_1.01,0.33023359775543215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,8,power_law_1.01,0.3536704063415527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,1,power_law_1.01,0.36941440105438234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,1,2,power_law_1.2,23.35680694580078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,8,power_law_1.01,0.5289152145385743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,1,power_law_1.01,0.475545597076416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,1,power_law_1.01,0.5627007961273194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,1,power_law_1.01,0.7512191772460938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,8,power_law_1.01,1.1510911941528321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,1,power_law_1.01,0.941868782043457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,1,power_law_1.01,1.3301504135131836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,2,power_law_1.01,0.0755840003490448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,2,power_law_1.01,0.09409279823303222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,1,power_law_1.01,1.7284160614013673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,2,power_law_1.01,0.09260159730911255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,2,power_law_1.01,0.14391679763793946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,2688,1856,6,128,1,1,power_law_1.01,2.1077247619628907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,2,power_law_1.01,0.20529279708862305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,2,power_law_1.01,0.30836479663848876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,2,power_law_1.01,0.3360703945159912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,2,power_law_1.01,0.3435456037521362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,2688,1856,6,128,1,1,power_law_1.01,3.261113739013672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,2,power_law_1.01,0.3483968019485474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,2,power_law_1.01,0.349619197845459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,2,power_law_1.01,0.3577536106109619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,2,power_law_1.01,0.36815359592437746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,2,power_law_1.01,0.3688191890716553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,2,power_law_1.01,0.37763841152191163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,2688,1856,6,128,1,1,power_law_1.01,6.390707015991211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,128,power_law_1.01,0.056364798545837404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,2,power_law_1.01,0.3947200059890747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,2,power_law_1.01,0.4093632221221924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,2,power_law_1.01,0.414188814163208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,8,16,power_law_1.01,0.039776000380516055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,2,power_law_1.01,0.4424767971038818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,8,16,power_law_1.01,0.036032000184059144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,128,power_law_1.01,0.05925120115280151
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,8,16,power_law_1.01,0.0357120007276535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,2,power_law_1.01,0.4526847839355469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,8,16,power_law_1.01,0.03526400029659271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,2,power_law_1.01,0.5020415782928467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,8,16,power_law_1.01,0.03687039911746979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,128,power_law_1.01,0.05235840082168579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,2,power_law_1.01,0.5620351791381836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,8,16,power_law_1.01,0.038848000764846805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,2,power_law_1.01,0.6642560005187989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,8,16,power_law_1.01,0.0390720009803772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,8,16,power_law_1.01,0.03941119909286499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,2,power_law_1.01,0.7607168197631836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,128,power_law_1.01,0.05187839865684509
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,8,16,power_law_1.01,0.039136001467704774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,8,16,power_law_1.01,0.039103999733924866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,2,power_law_1.01,1.0179007530212403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,8,16,power_law_1.01,0.040627199411392215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,128,power_law_1.01,0.05370879769325256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,2,power_law_1.01,1.1599167823791503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,8,16,power_law_1.01,0.04232960045337677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,8,16,power_law_1.01,0.04222080111503601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,8,16,power_law_1.01,0.04321280121803284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,2,power_law_1.01,1.5737279891967773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,128,power_law_1.01,0.054604798555374146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,8,16,power_law_1.01,0.048147198557853696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,8,16,power_law_1.01,0.050361597537994386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,2,power_law_1.01,1.9330175399780274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,8,16,power_law_1.01,0.05288959741592407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,128,power_law_1.01,0.05400959849357605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,8,16,power_law_1.01,0.05676800012588501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,8,16,power_law_1.01,0.06117759943008423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,2,power_law_1.01,2.586195182800293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,8,16,power_law_1.01,0.06998400092124939
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,128,power_law_1.01,0.05427200198173523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,8,16,power_law_1.01,0.0836031973361969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,8,16,power_law_1.01,0.10087679624557495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,128,power_law_1.01,0.055302399396896365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,2,power_law_1.01,3.679923248291016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,8,16,power_law_1.01,0.12074879407882691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,8,16,power_law_1.01,0.17876479625701905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,8,16,power_law_1.01,0.1981951951980591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,128,power_law_1.01,0.05534719824790955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,8,16,power_law_1.01,0.32887680530548097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,8,power_law_1.01,0.15166720151901245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,128,power_law_1.01,0.0561024010181427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,8,16,power_law_1.01,0.41666560173034667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,2,power_law_1.01,6.565856170654297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,128,power_law_1.01,0.059680002927780154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,8,16,power_law_1.01,0.5539328098297119
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,8,power_law_1.01,0.24031999111175537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,128,power_law_1.01,0.05996800065040588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,8,16,power_law_1.01,0.6848447799682618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,128,power_law_1.01,0.06061440110206604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,8,power_law_1.01,0.4360832214355469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,8,16,power_law_1.01,1.4753984451293944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,128,power_law_1.01,0.06758400201797485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,8,power_law_1.01,0.14693119525909423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,128,power_law_1.01,0.06780800223350525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,8,power_law_1.01,0.2106623888015747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,128,power_law_1.01,0.07368959784507752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,128,power_law_1.01,0.08426880240440368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,8,power_law_1.01,0.3377088069915771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,128,power_law_1.01,0.09434880018234253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,32,8,power_law_1.01,0.04692479968070984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,8,power_law_1.01,0.3375040054321289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,32,8,power_law_1.01,0.05058559775352478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,128,power_law_1.01,0.11658240556716919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,32,8,power_law_1.01,0.04577920138835907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,8,power_law_1.01,0.4456831932067871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,128,power_law_1.01,0.1379520058631897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,32,8,power_law_1.01,0.04574080109596253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,32,8,power_law_1.01,0.0486272007226944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,8,power_law_1.01,0.4655295848846436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,8,power_law_1.2,0.05605760216712952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,128,power_law_1.01,0.1768895983695984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,32,8,power_law_1.01,0.052172797918319705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,8,power_law_1.2,0.0629696011543274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,32,8,power_law_1.01,0.05573760271072388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,8,power_law_1.01,0.48853759765625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,128,power_law_1.01,0.1858944058418274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,32,8,power_law_1.01,0.058310401439666745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,8,power_law_1.2,0.058982402086257935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,32,8,power_law_1.01,0.05863040089607239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,8,power_law_1.01,0.4669312000274658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,8,power_law_1.2,0.05736320018768311
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,32,8,power_law_1.01,0.0593280017375946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,128,power_law_1.01,0.24097280502319335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,32,8,power_law_1.01,0.061478400230407716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,8,power_law_1.2,0.06379520297050476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,8,power_law_1.01,0.5208831787109375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,32,8,power_law_1.01,0.06288639903068542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,8,power_law_1.2,0.07059839963912964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,128,power_law_1.01,0.3044543981552124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,32,8,power_law_1.01,0.06627200245857238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,8,power_law_1.01,0.5230400085449218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,32,8,power_law_1.01,0.06988800168037415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,8,power_law_1.2,0.0747648000717163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,32,8,power_law_1.01,0.07831680178642272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,8,power_law_1.2,0.08213760256767273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,8,power_law_1.01,0.5225024223327637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,128,power_law_1.01,0.3836416006088257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,32,8,power_law_1.01,0.08050559759140015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,8,power_law_1.2,0.08577280044555664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,32,8,power_law_1.01,0.08355200290679932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,8,power_law_1.01,0.545958423614502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,32,8,power_law_1.01,0.09747840166091919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,8,power_law_1.2,0.08777599930763244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,128,power_law_1.01,0.5072639942169189
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,32,8,power_law_1.01,0.11010559797286987
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,8,power_law_1.01,0.5383103847503662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,8,power_law_1.2,0.09049599766731262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,32,8,power_law_1.01,0.14351359605789185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,32,8,power_law_1.01,0.17971839904785156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,8,power_law_1.2,0.09149439930915833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,8,power_law_1.01,0.6016831874847413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,32,8,power_law_1.01,0.23797760009765626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,2,128,power_law_1.01,0.6137663841247558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,8,power_law_1.2,0.0963648021221161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,32,8,power_law_1.01,0.2770944118499756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,8,power_law_1.01,0.6241919994354248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,8,power_law_1.2,0.09968000054359435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,32,8,power_law_1.01,0.3876352071762085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,8,power_law_1.2,0.10784000158309937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,8,power_law_1.01,0.6557248115539551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,32,8,power_law_1.01,0.5051583766937255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,8,power_law_1.2,0.11174399852752685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,2,128,power_law_1.01,0.9093759536743165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,32,8,power_law_1.01,0.7249728202819824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,8,power_law_1.01,0.7659840106964111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,8,power_law_1.2,0.1244928002357483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,32,8,power_law_1.01,0.91909761428833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,8,power_law_1.2,0.13871359825134277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,8,power_law_1.01,0.7476223945617676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,8,power_law_1.2,0.1548799991607666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,32,8,power_law_1.01,1.2094528198242187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,8,power_law_1.01,0.9137344360351562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,8,power_law_1.2,0.18983039855957032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,32,8,power_law_1.01,1.7514432907104491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,8,power_law_1.2,0.1926591992378235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,8,power_law_1.01,0.9828096389770508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,8,power_law_1.2,0.24765439033508302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,2,128,power_law_1.01,1.8395776748657227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,8,power_law_1.01,1.2728896141052246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,8,power_law_1.2,0.3256256103515625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,32,8,power_law_1.01,3.3943550109863283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,8,power_law_1.2,0.41436161994934084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,8,power_law_1.01,1.3430144309997558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,8,power_law_1.2,0.5425727844238282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,8,power_law_1.01,1.6414527893066406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,8,power_law_1.2,0.782758378982544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,8,power_law_1.01,2.047750473022461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,8,power_law_1.2,1.0426624298095704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,8,power_law_1.2,1.2830656051635743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,8,power_law_1.01,2.311270332336426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,128,balanced,0.039733332892258964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,128,balanced,0.03958400090535482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,128,balanced,0.04185600082079569
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,128,balanced,0.03956799954175949
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,128,balanced,0.04178666571776072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,128,balanced,0.04159466673930486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,128,balanced,0.04179200033346812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,128,balanced,0.04182399809360504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,128,balanced,0.04188799858093262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,128,balanced,0.04178666571776072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,128,balanced,0.04173333446184794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,128,balanced,0.041738669077555336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,128,balanced,0.042175998290379844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,128,balanced,0.043680002291997276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,128,balanced,0.04584000011285146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,128,balanced,0.04379733403523763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,128,balanced,0.04756266872088114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,8,power_law_1.2,2.014182472229004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,128,balanced,0.04782933493455251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,128,balanced,0.04771733283996582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,128,balanced,0.05227200190226237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,128,balanced,0.05385600030422211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,128,balanced,0.058837334314982094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,128,balanced,0.06223999957243601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,128,balanced,0.07845333218574524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,128,balanced,0.08931733171145122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,128,balanced,0.11060266693433125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,128,balanced,0.1318666636943817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,2,1,power_law_1.2,0.05905280113220215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,128,balanced,0.15518400073051453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,8,power_law_1.01,3.0862848281860353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,128,balanced,0.22603732347488403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,2,1,power_law_1.2,0.08255360126495362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,128,balanced,0.39871466159820557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,2,1,power_law_1.2,0.08706560134887695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,2,1,power_law_1.2,0.1335744023323059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,2,1,power_law_1.2,0.13167359828948974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,2,1,power_law_1.2,0.13553919792175292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,2,1,power_law_1.2,0.1406399965286255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,2,1,power_law_1.2,0.14020479917526246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,2,1,power_law_1.2,0.14383360147476196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,2,1,power_law_1.2,0.14448000192642213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,2,1,power_law_1.2,0.14495999813079835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,2,1,power_law_1.2,0.15096960067749024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,2,1,power_law_1.2,0.1535104036331177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,2,1,power_law_1.2,0.16257280111312866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,2,1,power_law_1.2,0.17548160552978515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,2,1,power_law_1.2,0.18382079601287843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,8,power_law_1.2,3.9518142700195313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,2,1,power_law_1.2,0.18940800428390503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,2,1,power_law_1.2,0.22274560928344728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,2,1,power_law_1.2,0.2531136035919189
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,2,1,power_law_1.2,0.30718719959259033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,2,1,power_law_1.2,0.37046399116516116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,2,1,power_law_1.2,0.4875648021697998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,2,1,power_law_1.2,0.593503999710083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,8,power_law_1.01,6.429574584960937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,2,1,power_law_1.2,0.8241984367370605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,2,1,power_law_1.2,1.065939235687256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,2,1,power_law_1.2,1.5347968101501466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,2,1,power_law_1.2,1.985856056213379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,2,1,power_law_1.2,2.4614656448364256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,2,1,power_law_1.2,3.8154430389404297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,2,1,power_law_1.2,7.49438705444336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,16,power_law_1.01,0.10105600357055664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,16,power_law_1.01,0.13898240327835082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,16,power_law_1.01,0.06080639958381653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,16,power_law_1.01,0.07819520235061646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,16,power_law_1.01,0.08428159952163697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,64,power_law_1.2,0.0897599995136261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,16,power_law_1.01,0.10694400072097779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,16,power_law_1.01,0.10870399475097656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,64,power_law_1.2,0.08378880023956299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,16,power_law_1.01,0.11367679834365844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,64,power_law_1.2,0.07256320118904114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,16,power_law_1.01,0.11034239530563354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,128,power_law_1.01,0.062041598558425906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,16,power_law_1.01,0.11508480310440064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,64,power_law_1.2,0.07941759824752807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,16,power_law_1.01,0.10994559526443481
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,64,power_law_1.2,0.07708160281181335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,16,power_law_1.01,0.12106239795684814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,16,power_law_1.01,0.11674879789352417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,128,power_law_1.01,0.06104959845542908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,64,power_law_1.2,0.08654720187187195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,16,power_law_1.01,0.12408319711685181
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,64,power_law_1.2,0.09141119718551635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,16,power_law_1.01,0.13252480030059816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,128,power_law_1.01,0.059680002927780154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,16,power_law_1.01,0.13833600282669067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,64,power_law_1.2,0.08448640108108521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,8,balanced,0.05231999854246775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,8,balanced,0.053823997577031456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,8,balanced,0.05421333511670431
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,8,balanced,0.058176000912984215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,8,balanced,0.06235733131567637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,8,balanced,0.07834666470686595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,8,balanced,0.10336533188819885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,8,balanced,0.10256533821423848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,16,power_law_1.01,0.15117440223693848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,8,balanced,0.10400533676147461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,8,balanced,0.10371733705202739
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,8,balanced,0.10610666871070862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,8,balanced,0.10910933216412862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,8,balanced,0.10931733250617981
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,8,balanced,0.10961066683133443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,8,balanced,0.11749866604804993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,64,power_law_1.2,0.08892160058021545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,8,balanced,0.1209333340326945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,8,balanced,0.12708266576131186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,8,balanced,0.13607999682426453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,8,balanced,0.1469066639741262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,16,power_law_1.01,0.17428480386734008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,8,balanced,0.16709866126378378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,8,balanced,0.18323200941085815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,8,balanced,0.2224959929784139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,128,power_law_1.01,0.059571200609207155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,8,balanced,0.26156800985336304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,8,balanced,0.3446720043818156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,8,balanced,0.45735466480255127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,64,power_law_1.2,0.09211519956588746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,16,power_law_1.01,0.19267840385437013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,8,balanced,0.6153226693471273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,8,balanced,0.8007893562316895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,8,balanced,0.9921173254648844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,16,power_law_1.01,0.236844801902771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,8,balanced,1.516416072845459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,64,power_law_1.2,0.09671040177345276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,128,power_law_1.01,0.060083198547363284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,8,balanced,2.9428586959838867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,16,power_law_1.01,0.22623999118804933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,64,power_law_1.2,0.0944320023059845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,16,power_law_1.01,0.30034561157226564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,64,power_law_1.2,0.09557120203971863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,128,power_law_1.01,0.060729598999023436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,16,power_law_1.01,0.3397631883621216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,64,power_law_1.2,0.0914687991142273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,16,power_law_1.01,0.45168638229370117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,128,power_law_1.01,0.06149119734764099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,64,power_law_1.2,0.10085760354995728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,16,power_law_1.01,0.5472896099090576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,64,power_law_1.2,0.09966080188751221
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,128,power_law_1.01,0.06228479743003845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,16,power_law_1.01,0.7756351947784423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,64,power_law_1.2,0.11882239580154419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,128,power_law_1.01,0.06133120059967041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,64,power_law_1.2,0.1282047986984253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,16,power_law_1.01,1.0248319625854492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,64,power_law_1.2,0.14280320405960084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,128,power_law_1.01,0.061945599317550656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,16,power_law_1.01,1.4180480003356934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,64,power_law_1.2,0.17432960271835327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,128,power_law_1.01,0.06208639740943909
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,64,power_law_1.2,0.2085184097290039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,16,power_law_1.01,2.202412796020508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,64,power_law_1.2,0.2676032066345215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,128,power_law_1.01,0.06311039924621582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,64,power_law_1.2,0.32279679775238035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,128,power_law_1.01,0.0641152024269104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,64,power_law_1.2,0.4461952209472656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,64,power_law_1.01,0.06778879761695862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,16,power_law_1.01,3.595814514160156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,128,power_law_1.01,0.06442880034446716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,64,power_law_1.2,0.5696832180023194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,128,power_law_1.01,0.07002239823341369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,64,power_law_1.01,0.05294079780578613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,64,power_law_1.2,0.8500160217285156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,128,power_law_1.01,0.07028480172157288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,64,power_law_1.01,0.04690560102462769
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,64,power_law_1.2,1.209881591796875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,128,power_law_1.01,0.07253760099411011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,64,power_law_1.01,0.0469184011220932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,128,power_law_1.01,0.08009600043296813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,64,power_law_1.2,1.426630401611328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,64,power_law_1.01,0.04616959989070892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,64,power_law_1.01,0.04850560128688812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,128,power_law_1.01,0.08851199746131896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,64,power_law_1.2,2.3574783325195314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,64,power_law_1.01,0.04961279928684235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,128,power_law_1.01,0.10539519786834717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,64,power_law_1.01,0.05139840245246887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,128,power_law_1.01,0.12252160310745239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,64,power_law_1.01,0.05129600167274475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,128,power_law_1.01,0.15793919563293457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,4,balanced,0.04393066465854645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,4,balanced,0.04562666515509287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,64,power_law_1.2,4.757164764404297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,4,balanced,0.05195199946562449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,4,balanced,0.0754559983809789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,4,balanced,0.10258133212725322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,4,balanced,0.11578133702278137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,4,balanced,0.11564800143241882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,4,balanced,0.11966933806737264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,4,balanced,0.11980266372362773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,4,balanced,0.12149332960446675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,4,balanced,0.12111999591191609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,4,balanced,0.12231466174125671
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,4,balanced,0.12116799751917522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,4,balanced,0.12731732924779257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,4,balanced,0.13409066200256348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,64,power_law_1.01,0.05300480127334595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,4,balanced,0.1376586655775706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,4,balanced,0.13863466183344522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,4,balanced,0.15474133690198263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,4,balanced,0.15844800074895224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,128,power_law_1.01,0.1876799941062927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,4,balanced,0.18998400370279947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,4,balanced,0.20754132668177286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,4,balanced,0.27375467618306476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,4,balanced,0.2985333402951558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,4,balanced,0.4287946621576945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,4,balanced,0.4963039954503377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,4,balanced,0.6865493456522623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,4,balanced,0.8684853712717692
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,64,power_law_1.01,0.05248000025749207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,4,balanced,1.0226986408233643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,128,power_law_1.01,0.24572160243988037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,4,balanced,1.6015307108561199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,4,balanced,3.0641759236653647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,64,power_law_1.01,0.05374079942703247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,128,power_law_1.01,0.3160831928253174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,64,power_law_1.01,0.05615360140800476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,128,power_law_1.01,0.4814335823059082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,64,power_law_1.01,0.059008002281188965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,64,power_law_1.01,0.06533759832382202
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,128,power_law_1.01,0.5643904209136963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,64,power_law_1.01,0.06809599995613098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,128,power_law_1.01,0.6726016044616699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,64,power_law_1.01,0.07578880190849305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,64,power_law_1.01,0.09226239919662475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,128,power_law_1.01,1.1727423667907715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,16,power_law_1.01,0.09169920086860657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,64,power_law_1.01,0.1011199951171875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,16,power_law_1.01,0.096288001537323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,64,power_law_1.01,0.12762240171432496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,16,power_law_1.01,0.06365439891815186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,4,balanced,0.03299200038115183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,4,balanced,0.031370667119820915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,4,balanced,0.03341866781314214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,4,balanced,0.047775998711586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,4,balanced,0.06930666665236156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,16,power_law_1.01,0.0690496027469635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,4,balanced,0.07634666562080383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,4,balanced,0.07774400214354198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,4,balanced,0.07851733267307281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,4,balanced,0.07773866752783458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,4,balanced,0.07751999795436859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,4,balanced,0.07828799883524577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,4,balanced,0.0784693310658137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,4,balanced,0.07841066519419353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,4,balanced,0.08142399787902832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,4,balanced,0.08284799754619598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,64,power_law_1.01,0.14406399726867675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,4,balanced,0.08535466591517131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,4,balanced,0.0862666666507721
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,128,power_law_1.01,2.187008094787598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,4,balanced,0.10269332925478618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,16,power_law_1.01,0.07439360022544861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,4,balanced,0.09319466352462769
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,4,balanced,0.12545599540074667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,4,balanced,0.10879466931025188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,4,balanced,0.14098133643468222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,4,balanced,0.14172266920407614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,4,balanced,0.209824005762736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,4,balanced,0.20710933208465576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,4,balanced,0.2874506711959839
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,4,balanced,0.3300693432490031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,16,power_law_1.01,0.08860160112380981
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,2,4,balanced,0.3930720090866089
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,2,4,balanced,0.5822720130284628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,2,4,balanced,1.0877226988474529
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,64,power_law_1.01,0.19005440473556517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,16,power_law_1.01,0.09313279986381531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,16,power_law_1.01,0.09039360284805298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,64,power_law_1.01,0.23233919143676757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,16,power_law_1.01,0.09068160057067871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,16,power_law_1.01,0.09022719860076904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,16,power_law_1.01,0.09634559750556945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,64,power_law_1.01,0.3195519924163818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,16,power_law_1.01,0.09739519953727722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,16,power_law_1.01,0.09649279713630676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,64,power_law_1.01,0.41864957809448244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,16,power_law_1.01,0.09850879907608032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,16,power_law_1.01,0.10641280412673951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,16,power_law_1.01,0.10811519622802734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,64,power_law_1.01,0.5753024101257325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,16,power_law_1.01,0.11153279542922974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,16,power_law_1.01,0.12457599639892578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,16,power_law_1.01,0.13596160411834718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,64,power_law_1.01,0.7169151782989502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,16,power_law_1.01,0.16343679428100585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,16,power_law_1.01,0.1773695945739746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,64,power_law_1.01,0.9132287979125977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,16,power_law_1.01,0.22047359943389894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,16,power_law_1.01,0.2809344053268433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,1,power_law_1.2,0.035692799091339114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,2,power_law_1.01,0.06842880249023438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,16,power_law_1.01,0.39144320487976075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,1,power_law_1.2,0.042131200432777405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,64,power_law_1.01,1.3904704093933105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,2,power_law_1.01,0.09285119771957398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,16,power_law_1.01,0.48093438148498535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,2,power_law_1.01,0.08979840278625488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,1,power_law_1.2,0.055174398422241214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,16,power_law_1.01,0.6353536128997803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,2,power_law_1.01,0.11871999502182007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,1,power_law_1.2,0.07207040190696716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,16,power_law_1.01,0.9135040283203125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,2,power_law_1.01,0.17737599611282348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,1,power_law_1.2,0.09639040231704712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,64,power_law_1.01,2.784793663024902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,2,power_law_1.01,0.22766718864440919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,16,power_law_1.01,0.9975808143615723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,1,power_law_1.2,0.16159360408782958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,2,power_law_1.01,0.31580801010131837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,1,power_law_1.2,0.1704319953918457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,2,power_law_1.01,0.33717119693756104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,16,power_law_1.01,1.4759296417236327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,1,power_law_1.2,0.18332159519195557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,2,power_law_1.01,0.34484479427337644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,1,power_law_1.2,0.19185279607772826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,2,power_law_1.01,0.35595519542694093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,1,power_law_1.2,0.19248000383377076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,1,power_law_1.2,0.2016063928604126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,2,power_law_1.01,0.3688256025314331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,16,power_law_1.01,2.7810623168945314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,1,power_law_1.2,0.20894079208374022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,2,power_law_1.01,0.38667519092559816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,1,power_law_1.2,0.2112191915512085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,2,power_law_1.01,0.39842560291290285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,1,power_law_1.2,0.21668479442596436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,2,power_law_1.01,0.4041408061981201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,1,power_law_1.2,0.2289344072341919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,2,power_law_1.01,0.41603841781616213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,1,power_law_1.2,0.23482880592346192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,2,power_law_1.01,0.4195263862609863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,1,power_law_1.2,0.24235520362854004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,1,power_law_1.2,0.2764607906341553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,2,power_law_1.01,0.4459263801574707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,1,power_law_1.2,0.27587199211120605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,2,power_law_1.01,0.4907711982727051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,1,power_law_1.2,0.32311038970947265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,2,power_law_1.01,0.5042367935180664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,1,power_law_1.2,0.3462399959564209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,32,power_law_1.01,0.06183040142059326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,2,power_law_1.01,0.5806272029876709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,1,power_law_1.2,0.4073984146118164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,2,power_law_1.01,0.6130752086639404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,1,power_law_1.2,0.4648767948150635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,32,power_law_1.01,0.06646400094032287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,1,power_law_1.2,0.577561616897583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,2,power_law_1.01,0.7482111930847168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,1,power_law_1.2,0.679040002822876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,32,power_law_1.01,0.05756800174713135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,2,power_law_1.01,0.8571840286254883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,16,power_law_1.01,0.05504000186920166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,1,power_law_1.2,0.9238592147827148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,2,power_law_1.01,1.083801555633545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,32,power_law_1.01,0.051692801713943484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,16,power_law_1.01,0.05612800121307373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,1,power_law_1.2,1.1419967651367187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,16,power_law_1.01,0.045049598813056944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,2,power_law_1.01,1.3011263847351073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,32,power_law_1.01,0.05303679704666138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,16,power_law_1.01,0.04576640129089356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,2,1,power_law_1.2,1.3870976448059082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,32,power_law_1.01,0.05559039711952209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,2,power_law_1.01,1.787628746032715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,16,power_law_1.01,0.04663679897785187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,16,power_law_1.01,0.04982399940490723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,32,power_law_1.01,0.05649920105934143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,2,1,power_law_1.2,2.1107648849487304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,2,power_law_1.01,2.216543960571289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,16,power_law_1.01,0.051571202278137204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,32,power_law_1.01,0.05708799958229065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,16,power_law_1.01,0.05198079943656921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,16,power_law_1.01,0.052908802032470705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,2,power_law_1.01,2.708083152770996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,32,power_law_1.01,0.05816320180892944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,2,1,power_law_1.2,4.059212875366211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,16,power_law_1.01,0.05382400155067444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,32,power_law_1.01,0.0580735981464386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,16,power_law_1.01,0.055404800176620486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,16,power_law_1.01,0.056524801254272464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,2,power_law_1.01,3.954790496826172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,32,power_law_1.01,0.0591871976852417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,16,power_law_1.01,0.058483201265335086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,32,power_law_1.01,0.060652798414230345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,16,power_law_1.01,0.06099200248718262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,16,power_law_1.01,0.06680319905281067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,32,power_law_1.01,0.061887997388839724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,16,power_law_1.01,0.06912639737129211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,32,power_law_1.01,0.06406400203704835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,2,power_law_1.01,7.4349311828613285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,16,power_law_1.01,0.07178239822387696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,32,power_law_1.01,0.06918399930000305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,16,power_law_1.01,0.08457599878311158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,16,power_law_1.01,0.09598720073699951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,32,power_law_1.01,0.07020800113677979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,16,power_law_1.01,0.12575360536575317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,32,power_law_1.01,0.07499520182609558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,16,power_law_1.01,0.15688960552215575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,4,power_law_1.01,0.04080640077590943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,32,power_law_1.01,0.08511360287666321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,16,power_law_1.01,0.2034303903579712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,4,power_law_1.01,0.038700801134109494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,16,power_law_1.01,0.23843200206756593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,32,power_law_1.01,0.09512959718704224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,4,power_law_1.01,0.040115201473236085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,16,power_law_1.01,0.3370048046112061
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,4,power_law_1.01,0.04304639995098114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,32,power_law_1.01,0.11589759588241577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,4,power_law_1.01,0.047737601399421695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,16,power_law_1.01,0.42533121109008787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,4,power_law_1.01,0.05440639853477478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,32,power_law_1.01,0.13905279636383056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,4,power_law_1.01,0.05591679811477661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,16,power_law_1.01,0.6498239994049072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,4,power_law_1.01,0.05806080102920532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,32,power_law_1.01,0.1777151942253113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,4,power_law_1.01,0.05699840188026428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,16,power_law_1.01,0.7764607906341553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,8,power_law_1.01,0.033907198905944826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,4,power_law_1.01,0.05909759998321533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,32,power_law_1.01,0.20330240726470947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,8,power_law_1.01,0.05617920160293579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,4,power_law_1.01,0.059699201583862306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,8,power_law_1.01,0.0333759993314743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,16,power_law_1.01,0.9544256210327149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,32,power_law_1.01,0.27584640979766845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,4,power_law_1.01,0.06339840292930603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,8,power_law_1.01,0.035872000455856326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,8,power_law_1.01,0.037324801087379456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,4,power_law_1.01,0.06302080154418946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,8,power_law_1.01,0.050361597537994386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,32,power_law_1.01,0.340499210357666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,8,power_law_1.01,0.05162879824638367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,4,power_law_1.01,0.06912000179290771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,16,power_law_1.01,1.4717696189880372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,8,power_law_1.01,0.052691197395324706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,4,power_law_1.01,0.07484800219535828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,8,power_law_1.01,0.054022401571273804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,4,power_law_1.01,0.07943040132522583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,8,power_law_1.01,0.05244799852371216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,32,power_law_1.01,0.5129280090332031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,8,power_law_1.01,0.05596799850463867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,4,power_law_1.01,0.08446080088615418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,8,power_law_1.01,0.053907197713851926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,8,power_law_1.01,0.05617280006408691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,4,power_law_1.01,0.10351999998092651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,32,power_law_1.01,0.6132991790771485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,16,power_law_1.01,2.9447935104370115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,8,power_law_1.01,0.05788800120353699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,4,power_law_1.01,0.10472960472106933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,8,power_law_1.01,0.062483197450637816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,4,power_law_1.01,0.12732800245285034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,8,power_law_1.01,0.06577280163764954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,8,power_law_1.01,0.06874240040779114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,32,power_law_1.01,0.7838399887084961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,4,power_law_1.01,0.1537343978881836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,8,power_law_1.01,0.07423359751701356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,4,power_law_1.01,0.18696320056915283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,8,power_law_1.01,0.07592960000038147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,8,power_law_1.01,0.0875711977481842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,4,power_law_1.01,0.23612799644470214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,8,power_law_1.01,0.095769602060318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,32,power_law_1.01,1.161177635192871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,8,power_law_1.01,0.11808639764785767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,4,power_law_1.01,0.34369919300079343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,8,power_law_1.01,0.15575040578842164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,4,power_law_1.01,0.41224961280822753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,8,power_law_1.01,0.17932159900665284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,8,power_law_1.01,0.23099520206451415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,4,power_law_1.01,0.5905792236328125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,8,power_law_1.01,0.3087104082107544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,32,power_law_1.01,2.440921592712402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,4,power_law_1.01,0.8472319602966308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,8,power_law_1.01,0.36268160343170164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,1,8,power_law_1.01,0.448038387298584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,4,power_law_1.01,0.9810815811157226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,1,8,power_law_1.01,0.6947199821472168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,2,64,balanced,0.025258667767047882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,2,64,balanced,0.025216000775496166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,2,64,balanced,0.025125332176685333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,2,64,balanced,0.025472000241279602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,2,64,balanced,0.02531733363866806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,2,64,balanced,0.027248000105222065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,2,64,balanced,0.027376001079877216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,2,64,balanced,0.02735466758410136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,2,64,balanced,0.026837334036827087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,2,64,balanced,0.027327999472618103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,2,64,balanced,0.02943466603755951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,2,64,balanced,0.029296000798543293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,2,64,balanced,0.029285334050655365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,2,64,balanced,0.03141866624355316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,2,64,balanced,0.033471999069054924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,2,64,balanced,0.03374933451414108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,2,64,balanced,0.03551466763019562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,2,64,balanced,0.03614933292071024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,4,power_law_1.01,1.4982784271240235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,64,balanced,0.03583999971548716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,64,balanced,0.03754666695992152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,64,balanced,0.03766400118668874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,64,balanced,0.03948266555865606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,64,balanced,0.03997333347797394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,64,balanced,0.04599999884764353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,64,balanced,0.0498879998922348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,64,balanced,0.060165335734685264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,64,balanced,0.07349333167076111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,64,balanced,0.081194669008255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,64,balanced,0.11283199985822041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,64,balanced,0.17121599117914835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,1,8,power_law_1.01,1.453343963623047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,4,power_law_1.01,2.8762367248535154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,2,power_law_1.01,0.031865599751472476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,2,power_law_1.01,0.05483520030975342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,2,power_law_1.01,0.042080000042915344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,2,power_law_1.01,0.059494400024414064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,2,power_law_1.01,0.0699072003364563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,2,power_law_1.01,0.09800320267677307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,2,power_law_1.01,0.10598399639129638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,2,power_law_1.01,0.11189119815826416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,2,power_law_1.01,0.11052160263061524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,16,4,balanced,0.03781333317359289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,16,4,balanced,0.03595199932654699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,16,4,balanced,0.035445332527160645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,16,4,balanced,0.035749333600203194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,16,4,balanced,0.03719999889532725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,128,balanced,0.04065066576004028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,16,4,balanced,0.03572800010442734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,2,power_law_1.01,0.11192320585250855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,16,4,balanced,0.03734400123357773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,128,balanced,0.04019733270009359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,16,4,balanced,0.03579200059175491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,128,balanced,0.04043200115362803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,16,4,balanced,0.0377866675456365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,16,4,balanced,0.03756266583998998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,128,balanced,0.04050133377313614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,16,4,balanced,0.03565866748491923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,128,balanced,0.04027199993530909
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,16,4,balanced,0.0378506655494372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,128,balanced,0.04194133480389913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,16,4,balanced,0.03956799954175949
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,128,balanced,0.04174399872620901
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,16,4,balanced,0.03942933430274328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,128,balanced,0.043466667334238686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,16,4,balanced,0.041637333730856575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,128,balanced,0.04302933315436045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,16,4,balanced,0.04166933397452036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,128,balanced,0.041759997606277466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,16,4,balanced,0.043706665436426796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,2,power_law_1.01,0.11644799709320068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,128,balanced,0.04262933135032654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,16,4,balanced,0.04878933231035868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,128,balanced,0.04381866753101349
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,16,4,balanced,0.050101334849993386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,128,balanced,0.04404800136884054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,16,4,balanced,0.056090667843818665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,128,balanced,0.0459146648645401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,16,4,balanced,0.06009600063165029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,128,balanced,0.046154667933781944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,16,4,balanced,0.07871999839941661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,128,balanced,0.046154667933781944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,16,4,balanced,0.09282132983207703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,128,balanced,0.04772266745567322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,16,4,balanced,0.1232319970925649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,128,balanced,0.05019199848175049
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,128,balanced,0.05013866722583771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,16,4,balanced,0.1448906660079956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,128,balanced,0.056373332937558494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,2,power_law_1.01,0.11771520376205444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,16,4,balanced,0.1974560022354126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,128,balanced,0.06022400160630544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,128,balanced,0.07687999804814656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,16,4,balanced,0.24406399329503378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,128,balanced,0.09687999884287517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,16,4,balanced,0.29657600323359173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,128,balanced,0.13154133160909018
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,128,balanced,0.15652799606323242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,16,4,balanced,0.4387893279393514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,128,balanced,0.2119413415590922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,2,power_law_1.01,0.11761280298233032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,128,balanced,0.2714453339576721
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,16,4,balanced,0.8166080315907797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,128,balanced,0.32678399483362836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,128,balanced,0.4901386499404907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,128,balanced,0.9306293328603109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,2,power_law_1.01,0.12282880544662475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,2,power_law_1.01,0.1292799949645996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,2,power_law_1.01,0.04764800071716309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,2,power_law_1.01,0.13263360261917115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,2,power_law_1.01,0.14523520469665527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,2,power_law_1.01,0.16755839586257934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,2,power_law_1.01,0.0549888014793396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,2,power_law_1.01,0.1547968029975891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,2,power_law_1.01,0.17740800380706787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,2,power_law_1.01,0.20821759700775147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,2,power_law_1.01,0.05669119954109192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,2,power_law_1.01,0.2465343952178955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,2,power_law_1.01,0.2559295892715454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,2,power_law_1.01,0.06984959840774536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,2,power_law_1.01,0.33240959644317625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,2,power_law_1.01,0.3979583978652954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,2,power_law_1.01,0.0898751974105835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,2,power_law_1.01,0.5569664001464844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,2,power_law_1.01,0.6600255966186523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,2,power_law_1.01,0.11002240180969239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,2,2,power_law_1.01,0.806272029876709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,2,power_law_1.01,0.1372480034828186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,2,2,power_law_1.01,1.1838975906372071
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,16,4,power_law_1.01,0.03519999980926514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,16,4,power_law_1.01,0.03052160143852234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,2,power_law_1.01,0.14491519927978516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,16,4,power_law_1.01,0.02971520125865936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,16,4,power_law_1.01,0.029657599329948426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,32,power_law_1.01,0.09359999895095825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,16,4,power_law_1.01,0.030323201417922975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,16,4,power_law_1.01,0.030451199412345885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,16,4,power_law_1.01,0.03167999982833862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,16,4,power_law_1.01,0.031667199730873105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,16,4,power_law_1.01,0.032876798510551454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,16,4,power_law_1.01,0.03327359855175018
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,2,power_law_1.01,0.14948480129241942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,16,4,power_law_1.01,0.03290880024433136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,16,4,power_law_1.01,0.033907198905944826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,32,power_law_1.01,0.1333631992340088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,2,2,power_law_1.01,2.3979776382446287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,16,4,power_law_1.01,0.0350847989320755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,16,4,power_law_1.01,0.0366784006357193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,16,4,power_law_1.01,0.04505600035190582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,16,4,power_law_1.01,0.047712001204490664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,16,4,power_law_1.01,0.050246399641036985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,2,power_law_1.01,0.1511615991592407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,16,4,power_law_1.01,0.055718398094177245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,16,4,power_law_1.01,0.06355199813842774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,32,power_law_1.01,0.10919040441513062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,16,4,power_law_1.01,0.0785152018070221
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,16,4,power_law_1.01,0.09163519740104675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,16,4,power_law_1.01,0.13018239736557008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,16,4,power_law_1.01,0.12683520317077637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,2,power_law_1.01,0.15847040414810182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,16,4,power_law_1.01,0.1655743956565857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,32,power_law_1.01,0.07152640223503112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,16,4,power_law_1.01,0.23469440937042235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,16,4,power_law_1.01,0.30566399097442626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,16,4,power_law_1.01,0.4468095779418945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,2,power_law_1.01,0.16750719547271728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,32,power_law_1.01,0.07428479790687562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,16,4,power_law_1.01,0.6071936130523682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,16,4,power_law_1.01,0.7999936103820801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,2,power_law_1.01,0.17128959894180298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,32,power_law_1.01,0.08507519960403442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,16,4,power_law_1.01,1.7422143936157226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,2,power_law_1.01,0.18468480110168456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,32,power_law_1.01,0.08419200181961059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,2,power_law_1.01,0.19526400566101074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,32,power_law_1.01,0.0955456018447876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,2,power_law_1.01,0.2061824083328247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,32,power_law_1.01,0.09881600141525268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,2,power_law_1.01,0.21276800632476806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,32,power_law_1.01,0.10025600194931031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,2,power_law_1.01,0.2463360071182251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,32,power_law_1.01,0.10180480480194092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,8,2,power_law_1.2,0.04611839950084686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,8,2,power_law_1.2,0.05727360248565674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,8,2,power_law_1.2,0.058950400352478026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,2,power_law_1.01,0.26855039596557617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,8,2,power_law_1.2,0.07508479952812194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,32,power_law_1.01,0.10695040225982666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,8,2,power_law_1.2,0.07438719868659974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,8,2,power_law_1.2,0.07461119890213012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,8,2,power_law_1.2,0.08004480004310607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,8,2,power_law_1.2,0.07909759879112244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,8,2,power_law_1.2,0.08081279993057251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,8,2,power_law_1.2,0.08065279722213745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,2,power_law_1.01,0.32921600341796875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,32,power_law_1.01,0.10907520055770874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,8,2,power_law_1.2,0.08102399706840516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,8,2,power_law_1.2,0.08606719970703125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,8,2,power_law_1.2,0.08543360233306885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,8,2,power_law_1.2,0.09467520117759705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,8,2,power_law_1.2,0.09830399751663207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,8,2,power_law_1.2,0.10377600193023681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,32,power_law_1.01,0.10809600353240967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,2,power_law_1.01,0.38048639297485354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,8,2,power_law_1.2,0.12428799867630005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,8,2,power_law_1.2,0.1294592022895813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,8,2,power_law_1.2,0.14545279741287231
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,8,2,power_law_1.2,0.2021631956100464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,8,2,power_law_1.2,0.21116158962249756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,32,power_law_1.01,0.11557120084762573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,8,2,power_law_1.2,0.31062400341033936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,2,power_law_1.01,0.5189695835113526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,8,2,power_law_1.2,0.37463040351867677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,8,2,power_law_1.2,0.4655168056488037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,8,2,power_law_1.2,0.6778304100036621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,32,power_law_1.01,0.11698559522628785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,1,power_law_1.2,0.034067198634147644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,8,2,power_law_1.2,0.8734975814819336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,2,power_law_1.01,0.6118015766143798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,8,2,power_law_1.2,1.1850815773010255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,1,power_law_1.2,0.035206401348114015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,32,power_law_1.01,0.12431360483169555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,8,2,power_law_1.2,1.3206527709960938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,1,power_law_1.2,0.041196799278259276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,2,power_law_1.01,0.7989439964294434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,1,power_law_1.2,0.04787839949131012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,8,2,power_law_1.2,2.1928640365600587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,32,power_law_1.01,0.13047679662704467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,1,power_law_1.2,0.05543680191040039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,1,power_law_1.2,0.08901119828224183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,2,power_law_1.01,1.0040575981140136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,32,power_law_1.01,0.1504320025444031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,8,2,power_law_1.2,5.1042945861816404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,1,power_law_1.2,0.09221760034561158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,1,power_law_1.2,0.09583359956741333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,32,power_law_1.01,0.17430399656295775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,1,power_law_1.2,0.09617919921875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,2,power_law_1.01,1.4448127746582031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,1,power_law_1.2,0.09835519790649414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,1,power_law_1.2,0.1004863977432251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,32,power_law_1.01,0.1979647994041443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,1,power_law_1.2,0.10510079860687256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,1,power_law_1.2,0.1078336000442505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,2,power_law_1.01,1.8561344146728516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,32,power_law_1.01,0.25382399559020996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,1,power_law_1.2,0.13040000200271606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,1,power_law_1.2,0.1375040054321289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,1,power_law_1.2,0.14195200204849243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,32,power_law_1.01,0.3070528030395508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,1,power_law_1.2,0.15695359706878662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,2,power_law_1.01,2.2641151428222654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,1,power_law_1.2,0.17466880083084108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,32,power_law_1.01,0.4134655952453613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,1,power_law_1.2,0.18549760580062866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,1,power_law_1.2,0.23763198852539064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,1,power_law_1.2,0.2881407976150513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,32,power_law_1.01,0.5254464149475098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,1,power_law_1.2,0.37978239059448243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,2,power_law_1.01,3.475724792480469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,1,power_law_1.2,0.5015744209289551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,32,power_law_1.01,0.7356671810150146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,1,power_law_1.2,0.7118207931518554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,1,power_law_1.2,0.8223872184753418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,32,power_law_1.01,0.960051155090332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,16,power_law_1.2,0.05543680191040039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,1,power_law_1.2,1.182419204711914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,16,power_law_1.2,0.06430720090866089
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,2,power_law_1.01,6.867929840087891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,1,power_law_1.2,1.4847807884216309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,32,power_law_1.01,1.1874752044677734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,16,power_law_1.2,0.05178239941596985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,16,power_law_1.2,0.05238400101661682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,1,power_law_1.2,1.8383808135986328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,16,power_law_1.2,0.05373439788818359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,32,power_law_1.01,1.8476991653442383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,16,power_law_1.2,0.05486080050468445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,1,power_law_1.2,2.9048831939697264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,16,power_law_1.2,0.05950719714164734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,16,power_law_1.2,0.060083198547363284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,16,power_law_1.2,0.06001920104026794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,32,power_law_1.01,3.7459583282470703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,1,power_law_1.2,5.60181770324707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,16,power_law_1.2,0.060159999132156375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,16,power_law_1.2,0.06217600107192993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,16,power_law_1.2,0.06355199813842774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,16,power_law_1.2,0.06519039869308471
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,16,power_law_1.2,0.06609280109405517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,128,balanced,0.037392000357309975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,128,balanced,0.03959999978542328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,128,balanced,0.03864533454179764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,128,balanced,0.038032000263532005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,128,balanced,0.03692266593376795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,128,balanced,0.03844800094763438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,128,balanced,0.04074133435885111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,128,balanced,0.03862400104602178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,128,balanced,0.03863999992609024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,16,power_law_1.2,0.07274240255355835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,128,balanced,0.039488000174363456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,128,balanced,0.038346665600935616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,128,balanced,0.04033066580692927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,128,balanced,0.04030400017897288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,128,balanced,0.040522667268911995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,128,balanced,0.04407466451327006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,128,balanced,0.04238933324813843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,128,balanced,0.044266665975252785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,128,balanced,0.044351999958356224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,128,balanced,0.04673600196838379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,128,balanced,0.0484746644894282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,128,balanced,0.05106666684150696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,16,power_law_1.2,0.07750399708747864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,128,balanced,0.05486399928728739
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,128,balanced,0.05753066639105479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,128,balanced,0.06976533432801564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,128,balanced,0.08003733555475871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,128,balanced,0.10204799969991048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,128,balanced,0.12196266651153564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,128,balanced,0.14221866925557455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,128,balanced,0.20057066281636557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,16,power_law_1.2,0.08140159845352173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,128,balanced,0.3585386673609416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,16,power_law_1.2,0.0942080020904541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,16,power_law_1.2,0.10839680433273316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,16,power_law_1.2,0.13054720163345337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,16,power_law_1.2,0.15208319425582886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,16,power_law_1.2,0.18592640161514282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,4,2,balanced,0.04165333261092504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,4,2,balanced,0.03961066653331121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,4,2,balanced,0.04404800136884054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,4,2,balanced,0.04776533444722494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,4,2,balanced,0.060080001751581825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,4,2,balanced,0.06002133091290792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,4,2,balanced,0.06011199951171875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,4,2,balanced,0.062037333846092224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,4,2,balanced,0.060533334811528526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,4,2,balanced,0.06214400132497152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,4,2,balanced,0.06208533545335134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,4,2,balanced,0.06452266871929169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,16,power_law_1.2,0.23656320571899414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,4,2,balanced,0.06640533109505971
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,4,2,balanced,0.06624533236026764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,4,2,balanced,0.07039999961853027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,4,2,balanced,0.07400000095367432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,4,2,balanced,0.07852266728878021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,4,2,balanced,0.08897599577903748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,4,2,balanced,0.0981333355108897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,32,1,power_law_1.2,0.05244160294532776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,4,2,balanced,0.1213653286298116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,4,2,balanced,0.13917332887649536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,4,2,balanced,0.19389333327611288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,4,2,balanced,0.23317333062489828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,4,2,balanced,0.33264533678690594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,4,2,balanced,0.42933865388234455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,4,2,balanced,0.6319946845372518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,16,power_law_1.2,0.3362943887710571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,4,2,balanced,0.833733320236206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,4,2,balanced,1.0445386568705242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,32,1,power_law_1.2,0.05623679757118225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,4,2,balanced,1.6480587323506672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,16,power_law_1.2,0.3919935941696167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,4,2,balanced,3.196197191874186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,32,1,power_law_1.2,0.061900800466537474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,16,power_law_1.2,0.5573823928833008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,32,1,power_law_1.2,0.07698559761047363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,16,power_law_1.2,0.738105583190918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,32,1,power_law_1.2,0.09613440036773682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,16,power_law_1.2,1.0025279998779297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,32,1,power_law_1.2,0.11936000585556031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,32,1,power_law_1.2,0.13694080114364623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,4,power_law_1.2,0.08648959994316101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,16,power_law_1.2,1.450233554840088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,32,1,power_law_1.2,0.16388479471206666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,32,1,power_law_1.2,0.17494399547576905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,32,1,power_law_1.2,0.17985279560089112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,4,power_law_1.2,0.07878400087356567
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,16,power_law_1.2,3.0258432388305665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,32,1,power_law_1.2,0.19364479780197144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,1,power_law_1.2,0.03340800106525421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,1,power_law_1.2,0.03857919871807099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,32,1,power_law_1.2,0.2076416015625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,4,power_law_1.2,0.07747200131416321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,1,power_law_1.2,0.04862079918384552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,1,power_law_1.2,0.05985280275344849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,32,1,power_law_1.2,0.21694719791412354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,1,power_law_1.2,0.07028480172157288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,1,power_law_1.2,0.11333760023117065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,32,1,power_law_1.2,0.23513600826263428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,1,power_law_1.2,0.12024960517883301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,4,power_law_1.2,0.10466560125350952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,1,power_law_1.2,0.12291840314865113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,32,1,power_law_1.2,0.2602880001068115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,1,power_law_1.2,0.12874239683151245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,1,power_law_1.2,0.13003519773483277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,1,power_law_1.2,0.13537280559539794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,32,1,power_law_1.2,0.27447679042816164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,1,power_law_1.2,0.14345599412918092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,4,power_law_1.2,0.12339199781417846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,1,power_law_1.2,0.14613759517669678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,32,1,power_law_1.2,0.30599040985107423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,1,power_law_1.2,0.15299839973449708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,1,power_law_1.2,0.15969280004501343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,32,1,power_law_1.2,0.3813375949859619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,1,power_law_1.2,0.16794240474700928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,1,power_law_1.2,0.1822208046913147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,4,power_law_1.2,0.18881920576095582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,1,power_law_1.2,0.2071615934371948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,32,1,power_law_1.2,0.37694079875946046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,1,power_law_1.2,0.2287424087524414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,1,power_law_1.2,0.28506879806518554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,32,1,power_law_1.2,0.47367038726806643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,8,power_law_1.2,0.06747519969940186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,1,power_law_1.2,0.32692480087280273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,4,power_law_1.2,0.20781440734863282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,1,power_law_1.2,0.42508797645568847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,32,1,power_law_1.2,0.5611648082733154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,1,power_law_1.2,0.5202623844146729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,8,power_law_1.2,0.09189760088920593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,1,power_law_1.2,0.709113597869873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,32,1,power_law_1.2,0.7389503955841065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,8,power_law_1.2,0.08777599930763244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,4,power_law_1.2,0.21273601055145264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,1,power_law_1.2,0.8935808181762696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,32,1,power_law_1.2,0.9310527801513672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,1,power_law_1.2,1.2766976356506348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,8,power_law_1.2,0.06716799736022949
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,4,power_law_1.2,0.22187519073486328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,1,power_law_1.2,1.654207992553711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,32,1,power_law_1.2,1.2542079925537108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,8,power_law_1.2,0.0821120023727417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,1,power_law_1.2,2.025215911865234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,8,power_law_1.2,0.09105280041694641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,32,1,power_law_1.2,1.5704383850097656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,4,power_law_1.2,0.22704639434814453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,8,power_law_1.2,0.10008959770202637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,1,power_law_1.2,3.0513919830322265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,32,1,power_law_1.2,2.2470016479492188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,8,power_law_1.2,0.11909760236740112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,4,power_law_1.2,0.2391103982925415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,8,power_law_1.2,0.11962239742279053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,32,1,power_law_1.2,2.8916479110717774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,4,power_law_1.2,0.2471168041229248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,8,power_law_1.2,0.12579200267791749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,1,power_law_1.2,5.883276748657226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,8,power_law_1.2,0.12744959592819213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,32,1,power_law_1.2,3.550815963745117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,4,power_law_1.2,0.24995200634002684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,8,power_law_1.2,0.13200000524520875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,8,power_law_1.2,0.1356735944747925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,4,power_law_1.2,0.25939838886260985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,32,1,power_law_1.2,5.741708755493164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,8,power_law_1.2,0.14413440227508545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,8,power_law_1.2,0.1505344033241272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,4,power_law_1.2,0.2682368040084839
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,8,power_law_1.2,0.1573248028755188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,8,power_law_1.2,0.16795519590377808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,4,power_law_1.2,0.27672319412231444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,8,power_law_1.2,0.18622080087661744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,32,1,power_law_1.2,11.136691284179687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,4,power_law_1.2,0.29071359634399413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,8,power_law_1.2,0.2006848096847534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,4,power_law_1.2,0.06777600049972535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,8,power_law_1.2,0.23826560974121094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,4,power_law_1.2,0.34897921085357664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,4,power_law_1.2,0.06711680293083191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,8,power_law_1.2,0.2733824014663696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,8,power_law_1.2,0.345632004737854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,4,power_law_1.2,0.06250240206718445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,4,power_law_1.2,0.3610624074935913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,8,power_law_1.2,0.39246718883514403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,4,power_law_1.2,0.07927680015563965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,4,power_law_1.2,0.4307583808898926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,8,power_law_1.2,0.5231872081756592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,4,power_law_1.2,0.09084799885749817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,8,power_law_1.2,0.6901055812835694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,4,power_law_1.2,0.4674560070037842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,4,power_law_1.2,0.12774399518966675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,1,power_law_1.01,0.03372800052165985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,8,power_law_1.2,0.9050944328308106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,4,power_law_1.2,0.13763840198516847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,1,power_law_1.01,0.05559039711952209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,4,power_law_1.2,0.6165887832641601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,1,power_law_1.01,0.055174398422241214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,1,power_law_1.01,0.08604159951210022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,4,power_law_1.2,0.14275200366973878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,8,power_law_1.2,1.1710016250610351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,1,power_law_1.01,0.10712319612503052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,1,power_law_1.01,0.16894079446792604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,4,power_law_1.2,0.14612480401992797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,4,power_law_1.2,0.658790397644043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,1,power_law_1.01,0.18102400302886962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,8,power_law_1.2,1.3396096229553223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,1,power_law_1.01,0.18376319408416747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,1,power_law_1.01,0.19304959774017333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,4,power_law_1.2,0.14821120500564575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,1,power_law_1.01,0.18904320001602173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,1,power_law_1.01,0.20476160049438477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,1,power_law_1.01,0.2048703908920288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,4,power_law_1.2,0.887615966796875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,4,power_law_1.2,0.15594880580902098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,8,power_law_1.2,2.207910346984863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,1,power_law_1.01,0.2076672077178955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,1,power_law_1.01,0.2166975975036621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,1,power_law_1.01,0.22894721031188964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,4,power_law_1.2,0.15877120494842528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,1,power_law_1.01,0.23013761043548583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,1,power_law_1.01,0.24174079895019532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,4,power_law_1.2,1.1169535636901855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,1,power_law_1.01,0.2676032066345215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,4,power_law_1.2,0.16351360082626343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,1,power_law_1.01,0.26670079231262206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,1,power_law_1.01,0.3040895938873291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,8,power_law_1.2,4.171526336669922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,1,power_law_1.01,0.33642239570617677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,4,power_law_1.2,0.16926720142364501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,1,power_law_1.01,0.40549120903015134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,4,power_law_1.2,1.5790975570678711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,1,power_law_1.01,0.446284818649292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,4,power_law_1.2,0.1805888056755066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,1,power_law_1.01,0.5827839851379395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,1,power_law_1.01,0.695142412185669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,4,power_law_1.2,0.1843135952949524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,1,power_law_1.01,0.9533184051513672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,1,power_law_1.01,1.194655990600586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,4,power_law_1.2,0.19682559967041016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,4,power_law_1.2,1.8696895599365235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,1,1,power_law_1.01,1.4450176239013672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,4,power_law_1.2,0.22707839012145997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,1,1,power_law_1.01,2.1948736190795897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,4,power_law_1.2,0.25241599082946775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,4,4,power_law_1.2,2.1697919845581053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,4,power_law_1.2,0.3110912084579468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,1,1,power_law_1.01,4.215526580810547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,2,8,power_law_1.01,0.04539520144462585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,2,8,power_law_1.01,0.05188480019569397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,2,8,power_law_1.01,0.04496639966964722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,2,8,power_law_1.01,0.04506239891052246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,2,8,power_law_1.01,0.04539520144462585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,4,power_law_1.2,0.38487679958343507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,2,8,power_law_1.01,0.04625920057296753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,2,8,power_law_1.01,0.045798400044441225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,2,8,power_law_1.01,0.0466048002243042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,2,8,power_law_1.01,0.05240319967269898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,2,8,power_law_1.01,0.053439998626708986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,2,8,power_law_1.01,0.05406079888343811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,2,8,power_law_1.01,0.061318397521972656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,2,8,power_law_1.01,0.061292797327041626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,2,8,power_law_1.01,0.06413440108299255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,4,power_law_1.2,0.48981761932373047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,2,8,power_law_1.01,0.074726402759552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,2,8,power_law_1.01,0.07548800110816956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,2,8,power_law_1.01,0.08447999954223633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,2,8,power_law_1.01,0.10493439435958862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,4,4,power_law_1.2,3.38260498046875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,2,8,power_law_1.01,0.12373759746551513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,2,8,power_law_1.01,0.1445248007774353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,2,8,power_law_1.01,0.16919679641723634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,2,8,power_law_1.01,0.21201279163360595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,2,8,power_law_1.01,0.3043776035308838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,4,power_law_1.2,0.5392000198364257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,2,8,power_law_1.01,0.38334081172943113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,2,8,power_law_1.01,0.5102975845336915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,2,8,power_law_1.01,0.7803391933441162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,2,8,power_law_1.01,0.946662425994873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,2,8,power_law_1.01,1.1811840057373046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,4,power_law_1.2,0.7385024070739746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,2,8,power_law_1.01,1.9337600708007812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,2,8,power_law_1.01,3.834688186645508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,4,power_law_1.2,0.9031167984008789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,8,power_law_1.01,0.0361407995223999
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,8,power_law_1.01,0.029728001356124877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,4,power_law_1.2,1.2703935623168945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,4,4,power_law_1.2,7.350860595703125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,8,power_law_1.01,0.029369598627090453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,8,power_law_1.01,0.02975359857082367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,8,power_law_1.01,0.03116160035133362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,8,power_law_1.01,0.03366400003433227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,8,power_law_1.01,0.033228799700737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,4,power_law_1.2,1.63439998626709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,8,power_law_1.01,0.03400320112705231
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,8,power_law_1.01,0.03409920036792755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,8,power_law_1.01,0.03475199937820435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,8,power_law_1.01,0.03521279990673065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,8,power_law_1.01,0.03619199991226196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,8,power_law_1.01,0.036959999799728395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,8,power_law_1.01,0.03842560052871704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,4,4,power_law_1.2,1.9941120147705078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,8,power_law_1.01,0.04160000085830688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,8,power_law_1.01,0.043398401141166686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,8,power_law_1.01,0.04594559967517853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,8,power_law_1.01,0.052934402227401735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,2,power_law_1.01,0.04981119930744171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,8,power_law_1.01,0.05976319909095764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,8,power_law_1.01,0.06483200192451477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,2,power_law_1.01,0.05273600220680237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,8,power_law_1.01,0.07391999959945679
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,2,power_law_1.01,0.057011198997497556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,8,power_law_1.01,0.0885312020778656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,4,4,power_law_1.2,3.1851455688476564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,8,power_law_1.01,0.11957759857177734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,32,8,power_law_1.2,0.03893760144710541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,2,power_law_1.01,0.06961280107498169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,8,power_law_1.01,0.17624959945678711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,32,8,power_law_1.2,0.03780480027198792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,32,8,power_law_1.2,0.03470079898834229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,2,power_law_1.01,0.0896127998828888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,8,power_law_1.01,0.2082304000854492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,32,8,power_law_1.2,0.03514240086078644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,32,8,power_law_1.2,0.034944000840187076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,2,power_law_1.01,0.10805120468139648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,8,power_law_1.01,0.3156928062438965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,32,8,power_law_1.2,0.0354559987783432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,32,8,power_law_1.2,0.03578880131244659
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,2,power_law_1.01,0.13446400165557862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,32,8,power_law_1.2,0.035174399614334106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,8,power_law_1.01,0.3929152011871338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,32,8,power_law_1.2,0.03560320138931274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,32,8,power_law_1.2,0.03589119911193848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,2,power_law_1.01,0.14305280447006224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,32,8,power_law_1.2,0.03700479865074158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,8,power_law_1.01,0.45348482131958007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,32,8,power_law_1.2,0.03681280016899109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,2,power_law_1.01,0.14483200311660765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,32,8,power_law_1.2,0.03688960075378418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,32,8,power_law_1.2,0.03699199855327606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,32,8,power_law_1.2,0.041068801283836366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,2,power_law_1.01,0.14602880477905272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,32,8,power_law_1.2,0.04255360066890716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,8,power_law_1.01,0.7969215869903564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,32,8,power_law_1.2,0.04417920112609863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,32,8,power_law_1.2,0.04982399940490723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,2,power_law_1.01,0.15589760541915892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,32,8,power_law_1.2,0.05466880202293396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,32,8,power_law_1.2,0.06458240151405334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,2,power_law_1.01,0.17214720249176024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,32,8,power_law_1.2,0.07056000232696533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,32,8,power_law_1.2,0.08715519905090333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,4,4,power_law_1.2,6.948639678955078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,32,8,power_law_1.2,0.10806399583816528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,2,power_law_1.01,0.17454080581665038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,8,power_law_1.01,1.4979392051696778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,32,8,power_law_1.2,0.1486143946647644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,32,8,power_law_1.2,0.18200960159301757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,2,power_law_1.01,0.19198720455169677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,32,8,power_law_1.2,0.25391359329223634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,32,8,power_law_1.2,0.32042880058288575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,2,power_law_1.01,0.20118401050567628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,32,8,power_law_1.2,0.38229119777679443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,32,8,power_law_1.2,0.5956160068511963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,2,power_law_1.01,0.21578879356384278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,2,power_law_1.01,0.21349759101867677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,32,8,power_law_1.2,1.166975975036621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,2,power_law_1.01,0.24849278926849366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,2,power_law_1.01,0.2566976070404053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,2,power_law_1.01,0.3141887903213501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,2,power_law_1.01,0.3461951971054077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,2,power_law_1.01,0.4398784160614014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,2,power_law_1.01,0.530291223526001
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,2,power_law_1.01,0.7171072006225586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,8,balanced,0.04577066500981649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,8,balanced,0.05211733281612396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,8,balanced,0.07075733443101247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,2,power_law_1.01,0.9005887985229493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,8,balanced,0.09702400366465251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,8,balanced,0.1446293294429779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,8,balanced,0.2435413400332133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,8,balanced,0.24316267172495523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,8,balanced,0.24572267134984335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,8,balanced,0.24636266628901163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,8,balanced,0.24676799774169922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,8,balanced,0.2480000058809916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,32,power_law_1.01,0.12711679935455322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,8,balanced,0.24978667497634888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,8,balanced,0.25014400482177734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,8,balanced,0.2508106629053752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,8,balanced,0.2562026580174764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,8,balanced,0.25923200448354083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,8,balanced,0.26186132431030273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,8,balanced,0.27191466093063354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,8,balanced,0.2811573346455892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,8,balanced,0.2961813410123189
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,8,balanced,0.322762668132782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,2,power_law_1.01,1.2838656425476074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,8,balanced,0.35118401050567627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,8,balanced,0.38222400347391766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,8,balanced,0.46964800357818604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,32,power_law_1.01,0.1976639986038208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,8,balanced,0.5189973513285319
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,2,balanced,0.031957333286603294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,8,balanced,0.7133119901021322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,2,balanced,0.03443199892838796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,2,balanced,0.035829332967599235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,2,balanced,0.048351998130480446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,8,balanced,0.8037866751352946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,2,balanced,0.0702453354994456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,2,balanced,0.08134933312733968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,2,balanced,0.0803466687599818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,8,balanced,1.0726453463236492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,2,balanced,0.08012799918651581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,2,balanced,0.08052800099054973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,2,balanced,0.08011200030644734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,2,balanced,0.08169066905975342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,2,balanced,0.08274133503437042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,8,balanced,1.4675413767496746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,2,balanced,0.0831413318713506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,2,power_law_1.01,1.6462720870971679
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,2,balanced,0.08194666604201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,32,power_law_1.01,0.15282560586929322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,2,balanced,0.08753066261609395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,2,balanced,0.08739733695983887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,2,balanced,0.0916266640027364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,2,balanced,0.10604266325632732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,2,balanced,0.10030933221181233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,8,balanced,2.8221492767333984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,2,balanced,0.13798933227856955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,2,balanced,0.12851732969284058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,2,balanced,0.1683893402417501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,2,balanced,0.18284799655278525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,2,balanced,0.25251734256744385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,2,balanced,0.2890560030937195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,2,balanced,0.40901867548624676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,2,balanced,0.5014400084813436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,32,power_law_1.01,0.07886080145835876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,4,2,balanced,0.6081013282140096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,4,2,balanced,0.9205493132273356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,2,power_law_1.01,1.9932863235473632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,4,2,balanced,1.7607520421346028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,32,power_law_1.01,0.08995199799537659
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,32,power_law_1.01,0.12402559518814087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,2,power_law_1.01,2.980371284484863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,32,power_law_1.01,0.12046079635620117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,32,power_law_1.01,0.13318400382995604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,32,power_law_1.01,0.13935359716415405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,2,power_law_1.01,6.084844970703125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,32,power_law_1.01,0.14206080436706542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,32,power_law_1.01,0.14479360580444336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,32,power_law_1.01,0.14487040042877197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,32,power_law_1.01,0.15226880311965943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,32,power_law_1.2,0.0778432011604309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,32,power_law_1.01,0.15700479745864868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,256,power_law_1.2,0.062412798404693604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,32,power_law_1.01,0.15760639905929566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,32,power_law_1.2,0.06857600212097167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,32,power_law_1.01,0.15985920429229736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,256,power_law_1.2,0.062028801441192626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,32,power_law_1.2,0.05454720258712768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,32,power_law_1.01,0.15908479690551758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,32,power_law_1.2,0.06005120277404785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,256,power_law_1.2,0.062489598989486694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,32,power_law_1.01,0.174399995803833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,32,power_law_1.2,0.06356480121612548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,32,power_law_1.01,0.1937280058860779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,256,power_law_1.2,0.0646336019039154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,8,power_law_1.2,0.05920640230178833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,32,power_law_1.01,0.22958719730377197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,8,power_law_1.2,0.05944960117340088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,32,power_law_1.2,0.07406079769134521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,256,power_law_1.2,0.061952000856399535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,8,power_law_1.2,0.05375999808311462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,32,power_law_1.01,0.2595711946487427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,32,power_law_1.2,0.07757440209388733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,256,power_law_1.2,0.06284160017967225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,8,power_law_1.2,0.05546240210533142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,32,power_law_1.01,0.32467200756073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,8,power_law_1.2,0.05745279788970947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,256,power_law_1.2,0.0645632028579712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,32,power_law_1.2,0.0791808009147644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,8,power_law_1.2,0.06309120059013366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,32,power_law_1.01,0.3866559982299805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,8,power_law_1.2,0.06786559820175171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,256,power_law_1.2,0.06398720145225525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,32,power_law_1.2,0.07886719703674316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,32,power_law_1.01,0.5232384204864502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,8,power_law_1.2,0.06738560199737549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,256,power_law_1.2,0.06379520297050476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,32,power_law_1.2,0.077702397108078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,8,power_law_1.2,0.06991999745368957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,32,power_law_1.01,0.6483776092529296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,256,power_law_1.2,0.06542080044746398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,8,power_law_1.2,0.06988160014152527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,32,power_law_1.2,0.08133760094642639
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,8,power_law_1.2,0.0710591971874237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,32,power_law_1.01,0.89617919921875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,256,power_law_1.2,0.06509439945220948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,8,power_law_1.2,0.07464320063591004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,32,power_law_1.2,0.08407679796218873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,256,power_law_1.2,0.06412799954414368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,8,power_law_1.2,0.07635200023651123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,32,power_law_1.01,1.1523903846740722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,32,power_law_1.2,0.08558080196380616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,8,power_law_1.2,0.07863680124282837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,256,power_law_1.2,0.0658623993396759
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,8,power_law_1.2,0.08780800104141236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,32,power_law_1.01,1.4094847679138183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,32,power_law_1.2,0.08935040235519409
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,256,power_law_1.2,0.06468480229377746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,8,power_law_1.2,0.09080960154533387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,8,power_law_1.2,0.09889280200004577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,256,power_law_1.2,0.06780160069465638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,32,power_law_1.2,0.09579520225524903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,8,power_law_1.2,0.11169919967651368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,32,power_law_1.01,2.126188850402832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,256,power_law_1.2,0.06735360026359558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,8,power_law_1.2,0.13319679498672485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,32,power_law_1.2,0.09889280200004577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,256,power_law_1.2,0.08018559813499451
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,8,power_law_1.2,0.16002559661865234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,32,power_law_1.2,0.10497920513153076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,8,power_law_1.2,0.18238719701766967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,256,power_law_1.2,0.08555520176887513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,8,power_law_1.2,0.24029440879821778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,32,power_law_1.01,4.086867141723633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,32,power_law_1.2,0.1210368037223816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,256,power_law_1.2,0.1015552043914795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,8,power_law_1.2,0.2822848081588745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,8,power_law_1.2,0.377836799621582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,256,power_law_1.2,0.13571200370788575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,32,power_law_1.2,0.12988799810409546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,8,power_law_1.2,0.505683183670044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,256,power_law_1.2,0.15972479581832885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,32,power_law_1.2,0.16355199813842775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,8,power_law_1.2,0.6648704051971436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,256,power_law_1.2,0.21585919857025146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,32,power_law_1.2,0.19125759601593018
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,8,power_law_1.2,0.9293503761291504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,256,power_law_1.2,0.2612096071243286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,32,power_law_1.2,0.24729599952697753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,8,power_law_1.2,1.134432029724121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,256,power_law_1.2,0.35935359001159667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,32,power_law_1.2,0.2772288084030151
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,256,power_law_1.2,0.4913792133331299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,8,power_law_1.2,1.9550975799560546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,256,power_law_1.2,0.7354559898376465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,32,power_law_1.2,0.3846463918685913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,8,power_law_1.2,0.046777600049972536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,8,power_law_1.2,0.04297600090503693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,256,power_law_1.2,0.9824576377868652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,8,power_law_1.2,0.04392319917678833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,8,power_law_1.2,4.154828643798828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,8,power_law_1.2,0.047679999470710756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,32,power_law_1.2,0.4787007808685303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,8,power_law_1.2,0.05095040202140808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,8,power_law_1.2,0.05568000078201294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,256,power_law_1.2,1.2554752349853515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,8,power_law_1.2,0.060236799716949466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,8,power_law_1.2,0.061478400230407716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,32,power_law_1.2,0.6559679985046387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,8,power_law_1.2,0.06459519863128663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,8,power_law_1.2,0.06330239772796631
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,256,power_law_1.2,2.041299247741699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,8,power_law_1.2,0.06387839913368225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,8,power_law_1.2,0.06715520024299622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,32,power_law_1.2,0.7745024204254151
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,8,power_law_1.2,0.06830719709396363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,8,power_law_1.2,0.07203840017318726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,8,power_law_1.2,0.07973120212554932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,8,power_law_1.2,0.08391039967536926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,256,power_law_1.2,4.13504638671875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,8,power_law_1.2,0.08800640106201171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,2,32,power_law_1.2,0.9583359718322754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,8,power_law_1.2,0.0954688012599945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,8,power_law_1.2,0.10927360057830811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,8,power_law_1.2,0.13691519498825072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,8,power_law_1.2,0.1686527967453003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,2,balanced,0.07281599938869476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,2,balanced,0.11955199639002483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,2,balanced,0.1476426621278127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,2,balanced,0.24432533979415894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,2,balanced,0.43832000096638996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,2,balanced,0.8266932964324951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,8,power_law_1.2,0.2157439947128296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,2,balanced,1.21888534228007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,2,balanced,1.21998397509257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,2,balanced,1.2190186977386475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,8,power_law_1.2,0.23149440288543702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,2,balanced,1.2206453482309978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,2,balanced,1.2235840161641438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,2,balanced,1.2325759728749592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,8,power_law_1.2,0.3538496017456055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,2,balanced,1.2339093685150146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,2,32,power_law_1.2,1.4525568008422851
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,2,balanced,1.2405386765797932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,2,balanced,1.2497119903564453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,2,balanced,1.2552639643351238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,8,power_law_1.2,0.5026239871978759
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,2,balanced,1.2699039777119954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,2,balanced,1.2946826616923015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,2,balanced,1.3196319739023845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,2,balanced,1.3760159810384114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,8,power_law_1.2,0.6640704154968262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,2,balanced,1.4171627362569172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,2,balanced,1.502090613047282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,2,balanced,1.5945013364156086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,2,balanced,1.7610559463500977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,8,power_law_1.2,0.8280575752258301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,2,balanced,2.0227306683858237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,2,balanced,2.32370662689209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,2,balanced,3.034480094909668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,8,power_law_1.2,0.9782848358154297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,2,balanced,4.115333239237468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,2,balanced,5.154170672098796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,8,power_law_1.2,1.6826047897338867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,2,32,power_law_1.2,3.7092288970947265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,2,balanced,9.74891217549642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,128,power_law_1.2,0.08837119936943054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,8,power_law_1.2,3.08276481628418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,128,power_law_1.2,0.08866559863090515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,128,power_law_1.2,0.08094080090522766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,128,power_law_1.2,0.07478399872779846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,128,power_law_1.2,0.08010879755020142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,128,power_law_1.2,0.08144000172615051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,128,power_law_1.2,0.08511360287666321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,16,1,power_law_1.2,0.036800000071525577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,16,1,power_law_1.2,0.036908799409866334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,16,1,power_law_1.2,0.038553598523139956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,16,1,power_law_1.2,0.04234879910945892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,16,power_law_1.2,0.04981119930744171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,128,power_law_1.2,0.08622080087661743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,16,1,power_law_1.2,0.043756800889968875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,16,1,power_law_1.2,0.043628799915313723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,16,1,power_law_1.2,0.0450111985206604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,16,power_law_1.2,0.044377601146698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,16,1,power_law_1.2,0.04481280148029328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,16,1,power_law_1.2,0.04700160026550293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,4,power_law_1.2,0.07297279834747314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,16,1,power_law_1.2,0.04865919947624207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,16,1,power_law_1.2,0.049568000435829165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,16,power_law_1.2,0.04357120096683502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,16,1,power_law_1.2,0.05433599948883057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,16,1,power_law_1.2,0.05697280168533325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,4,power_law_1.2,0.07108479738235474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,16,1,power_law_1.2,0.06213120222091675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,16,1,power_law_1.2,0.07092480063438415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,16,power_law_1.2,0.04519680142402649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,128,power_law_1.2,0.08387839794158936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,16,1,power_law_1.2,0.07212799787521362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,16,1,power_law_1.2,0.08428800106048584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,4,power_law_1.2,0.06416000127792358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,16,1,power_law_1.2,0.09548799991607666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,16,power_law_1.2,0.04509440064430237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,16,1,power_law_1.2,0.10686719417572021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,16,1,power_law_1.2,0.12398719787597656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,4,power_law_1.2,0.0825215995311737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,16,1,power_law_1.2,0.1385599970817566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,16,power_law_1.2,0.04848639965057373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,16,1,power_law_1.2,0.18328959941864015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,16,1,power_law_1.2,0.2205631971359253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,4,power_law_1.2,0.10086400508880615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,128,power_law_1.2,0.08352000117301941
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,16,power_law_1.2,0.04975999891757965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,16,1,power_law_1.2,0.3045439958572388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,16,1,power_law_1.2,0.38995840549468996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,4,power_law_1.2,0.15456000566482545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,16,power_law_1.2,0.05085440278053284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,16,1,power_law_1.2,0.5550528049468995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,4,power_law_1.2,0.16484479904174804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,16,1,power_law_1.2,0.7179967880249023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,16,power_law_1.2,0.050995200872421265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,16,1,power_law_1.2,0.8833600044250488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,4,power_law_1.2,0.16374399662017822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,128,power_law_1.2,0.08608639836311341
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,16,power_law_1.2,0.05130879878997803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,4,power_law_1.2,0.17139840126037598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,16,1,power_law_1.2,1.3697152137756348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,16,power_law_1.2,0.052716797590255736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,4,power_law_1.2,0.17070720195770264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,16,power_law_1.2,0.052832001447677614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,128,power_law_1.2,0.08911359906196595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,4,power_law_1.2,0.18069119453430177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,16,1,power_law_1.2,2.6810815811157225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,16,power_law_1.2,0.05549439787864685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,4,power_law_1.2,0.1813055992126465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,16,power_law_1.2,0.05797119736671448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,4,power_law_1.2,0.1850111961364746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,128,power_law_1.2,0.09162880182266235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,16,power_law_1.2,0.06480640172958374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,4,power_law_1.2,0.19228800535202026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,16,power_law_1.2,0.0688256025314331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,4,power_law_1.2,0.2068416118621826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,16,power_law_1.2,0.06851840019226074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,128,power_law_1.2,0.08668799996376038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,4,power_law_1.2,0.21521921157836915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,16,power_law_1.2,0.08294399976730346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,4,power_law_1.2,0.21875839233398436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,16,power_law_1.2,0.08632320165634155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,4,power_law_1.2,0.24500479698181152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,128,power_law_1.2,0.09493119716644287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,16,power_law_1.2,0.11038719415664673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,4,power_law_1.2,0.27670400142669677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,16,power_law_1.2,0.13333760499954223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,4,power_law_1.2,0.3205120086669922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,16,power_law_1.2,0.15352959632873536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,128,power_law_1.2,0.09550079703330994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,4,power_law_1.2,0.38224000930786134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,16,power_law_1.2,0.22313599586486815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,4,power_law_1.2,0.47599358558654786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,16,power_law_1.2,0.2869312047958374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,128,power_law_1.2,0.09780480265617371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,4,power_law_1.2,0.5278528213500977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,16,power_law_1.2,0.3404927968978882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,4,power_law_1.2,0.7152448177337647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,128,power_law_1.2,0.10670080184936523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,16,power_law_1.2,0.5499008178710938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,4,power_law_1.2,0.9456192016601562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,16,power_law_1.2,0.7121088027954101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,4,power_law_1.2,1.4208895683288574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,128,power_law_1.2,0.11427199840545654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,8,32,power_law_1.2,0.03929600119590759
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,8,32,power_law_1.2,0.036313599348068236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,16,power_law_1.2,0.7988800048828125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,8,32,power_law_1.2,0.03425279855728149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,4,power_law_1.2,1.7094976425170898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,8,32,power_law_1.2,0.033881598711013795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,128,power_law_1.2,0.13279999494552613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,8,32,power_law_1.2,0.03496319949626923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,16,power_law_1.2,1.420518398284912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,8,32,power_law_1.2,0.03592959940433502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,4,power_law_1.2,2.12225284576416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,8,32,power_law_1.2,0.03619199991226196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,128,power_law_1.2,0.15590399503707886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,8,32,power_law_1.2,0.03627519905567169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,8,32,power_law_1.2,0.036735999584198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,8,32,power_law_1.2,0.0367680013179779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,4,power_law_1.2,3.0863679885864257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,8,32,power_law_1.2,0.039212799072265624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,128,power_law_1.2,0.20106880664825438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,16,power_law_1.2,2.393779182434082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,8,32,power_law_1.2,0.03852800130844116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,8,32,power_law_1.2,0.03973760008811951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,8,32,power_law_1.2,0.0418368011713028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,128,power_law_1.2,0.24638080596923828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,8,32,power_law_1.2,0.0452672004699707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,8,32,power_law_1.2,0.045952001214027406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,8,32,power_law_1.2,0.049542400240898135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,4,power_law_1.2,6.819654083251953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,128,power_law_1.2,0.33023359775543215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,8,32,power_law_1.2,0.054022401571273804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,8,32,power_law_1.2,0.0554751992225647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,8,32,power_law_1.2,0.06508799791336059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,8,32,power_law_1.2,0.08058239817619324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,128,power_law_1.2,0.4073791980743408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,8,32,power_law_1.2,0.0986240029335022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,8,32,power_law_1.2,0.13480960130691527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,8,32,power_law_1.2,0.1874943971633911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,128,power_law_1.2,0.5843135833740234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,8,32,power_law_1.2,0.2014591932296753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,8,32,power_law_1.2,0.2981120109558105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,128,power_law_1.2,0.7675903797149658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,8,32,power_law_1.2,0.42879362106323243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,8,32,power_law_1.2,0.5262847900390625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,32,power_law_1.2,0.03611519932746887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,128,power_law_1.2,0.9807999610900879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,32,power_law_1.2,0.045587199926376346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,8,32,power_law_1.2,0.7493887901306152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,32,power_law_1.2,0.03155840039253235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,32,power_law_1.2,0.0320576012134552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,32,power_law_1.2,0.03159680068492889
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,4,2,power_law_1.2,0.04521600008010864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,4,2,power_law_1.2,0.04547840058803558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,32,power_law_1.2,0.032576000690460204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,4,2,power_law_1.2,0.05616000294685364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,128,power_law_1.2,1.4597824096679688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,4,2,power_law_1.2,0.06231039762496948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,32,power_law_1.2,0.033267199993133545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,4,2,power_law_1.2,0.06458240151405334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,4,2,power_law_1.2,0.06476160287857055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,8,32,power_law_1.2,2.3405759811401365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,4,2,power_law_1.2,0.0646016001701355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,32,power_law_1.2,0.03367680013179779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,4,2,power_law_1.2,0.06554880142211914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,4,2,power_law_1.2,0.06698240041732788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,4,2,power_law_1.2,0.06715520024299622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,32,power_law_1.2,0.03400959968566895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,4,2,power_law_1.2,0.069651198387146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,4,2,power_law_1.2,0.07052800059318542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,4,2,power_law_1.2,0.0707584023475647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,32,power_law_1.2,0.03446399867534637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,4,2,power_law_1.2,0.0732800006866455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,4,2,power_law_1.2,0.08568959832191467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,4,2,power_law_1.2,0.08962560296058655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,32,power_law_1.2,0.03808000087738037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,4,2,power_law_1.2,0.09589759707450866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,4,2,power_law_1.2,0.1140671968460083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,32,power_law_1.2,0.038515201210975646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,4,2,power_law_1.2,0.13346560001373292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,4,2,power_law_1.2,0.16572799682617187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,4,2,power_law_1.2,0.18196480274200438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,32,power_law_1.2,0.041459199786186215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,4,2,power_law_1.2,0.24868481159210204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,4,2,power_law_1.2,0.2953536033630371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,128,power_law_1.2,3.6117183685302736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,32,power_law_1.2,0.04497919976711273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,4,2,power_law_1.2,0.3951103925704956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,32,power_law_1.2,0.051737600564956666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,4,2,power_law_1.2,0.516051197052002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,4,2,power_law_1.2,0.8440064430236817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,32,power_law_1.2,0.05170559883117676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,4,2,power_law_1.2,0.9170559883117676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,32,power_law_1.2,0.054041600227355956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,4,2,power_law_1.2,1.3445119857788086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,32,power_law_1.2,0.06269440054893494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,32,power_law_1.2,0.06840959787368775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,4,2,power_law_1.2,1.8768192291259767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,32,power_law_1.2,0.07921280264854431
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,32,power_law_1.2,0.09196799993515015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,4,2,power_law_1.2,4.172294235229492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,32,power_law_1.2,0.13444479703903198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,32,power_law_1.2,0.13729920387268066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,32,power_law_1.2,0.2043839931488037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,32,power_law_1.2,0.2603967905044556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,4,balanced,0.04832000037034353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,32,power_law_1.2,0.3402112007141113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,4,balanced,0.048751999934514366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,4,balanced,0.057536001006762184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,4,balanced,0.07522133489449818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,4,balanced,0.10662399729092915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,4,balanced,0.1572213371594747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,4,balanced,0.20688533782958984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,4,balanced,0.20586133003234863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,4,balanced,0.21289600928624472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,4,balanced,0.21264000733693442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,4,balanced,0.21094399690628052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,4,balanced,0.21760533253351846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,4,balanced,0.22300799687703451
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,32,power_law_1.2,0.5318272113800049
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,4,balanced,0.2241493264834086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,4,balanced,0.23149333397547403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,4,balanced,0.23537067572275797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,4,balanced,0.24566932519276938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,4,balanced,0.26386133829752606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,4,balanced,0.28040534257888794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,4,balanced,0.31410666306813556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,4,balanced,0.34545600414276123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,4,balanced,0.41493332386016846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,4,balanced,0.4939839839935303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,4,balanced,0.6337493260701498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,32,power_law_1.2,0.6637760162353515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,4,balanced,0.7995680173238119
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,4,balanced,1.0865920384724934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,4,balanced,1.3875253995259602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,4,balanced,1.7189173698425293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,4,balanced,2.597818692525228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,32,power_law_1.2,0.9585727691650391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,4,balanced,5.002122561136882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,4,power_law_1.01,0.05185920000076294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,32,power_law_1.2,2.3111167907714845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,4,power_law_1.01,0.07207679748535156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,1,power_law_1.2,0.05258880257606506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,1,power_law_1.2,0.06076800227165222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,4,power_law_1.01,0.06119679808616638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,1,power_law_1.2,0.08319360017776489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,1,power_law_1.2,0.10253440141677857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,4,power_law_1.01,0.06944640278816223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,1,power_law_1.2,0.13515520095825195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,1,power_law_1.2,0.23084800243377684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,1,power_law_1.2,0.24079999923706055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,4,power_law_1.01,0.08903679847717286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,1,power_law_1.2,0.25481600761413575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,1,power_law_1.2,0.2531840085983276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,4,power_law_1.01,0.10474879741668701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,1,power_law_1.2,0.26654078960418703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,1,power_law_1.2,0.27148799896240233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,1,power_law_1.2,0.2770751953125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,4,power_law_1.01,0.1290176033973694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,1,power_law_1.2,0.28105599880218507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,1,power_law_1.2,0.2878592014312744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,4,power_law_1.01,0.13288960456848145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,1,power_law_1.2,0.30593280792236327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,32,balanced,0.040565334260463715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,32,balanced,0.0378506655494372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,32,balanced,0.03606933355331421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,32,balanced,0.0360000009338061
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,1,power_law_1.2,0.3104703903198242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,32,balanced,0.04012266546487808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,32,balanced,0.039936001102129616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,32,balanced,0.03991466760635376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,32,balanced,0.039962666730086006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,32,balanced,0.039749334255854286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,32,balanced,0.040149333576361336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,32,balanced,0.0401706670721372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,32,balanced,0.04028266668319702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,32,balanced,0.042133331298828125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,32,balanced,0.04206933577855428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,1,power_law_1.2,0.3328831911087036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,32,balanced,0.04615999758243561
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,4,power_law_1.01,0.1347584009170532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,32,balanced,0.04399999976158142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,32,balanced,0.04623466730117798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,32,balanced,0.04822400212287903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,32,balanced,0.052154665191968284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,32,balanced,0.058304001887639366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,32,balanced,0.060640002290407814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,32,balanced,0.08031466603279114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,32,balanced,0.08819199601809184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,32,balanced,0.11344533165295918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,1,power_law_1.2,0.38215041160583496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,32,balanced,0.12577066818873087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,32,balanced,0.16305599610010782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,32,balanced,0.19801600774129233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,32,balanced,0.23305600881576538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,32,balanced,0.34337600072224933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,8,power_law_1.2,0.03400320112705231
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,1,power_law_1.2,0.434611177444458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,32,balanced,0.623583992322286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,4,power_law_1.01,0.13571840524673462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,8,power_law_1.2,0.03173120021820068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,1,power_law_1.2,0.5122560024261474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,8,power_law_1.2,0.03150080144405365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,1,power_law_1.2,0.5825407981872559
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,4,power_law_1.01,0.14410239458084106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,8,power_law_1.2,0.03294720053672791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,1,power_law_1.2,0.7455359935760498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,8,power_law_1.2,0.03573760092258453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,4,power_law_1.01,0.1484287977218628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,8,power_law_1.2,0.038848000764846805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,1,power_law_1.2,0.8974592208862304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,8,power_law_1.2,0.039705601334571836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,1,power_law_1.2,1.2120320320129394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,8,power_law_1.2,0.04092159867286682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,4,power_law_1.01,0.1536960005760193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,8,power_law_1.2,0.04224640130996704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,1,power_law_1.2,1.5176383972167968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,8,power_law_1.2,0.04081920087337494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,4,power_law_1.01,0.16051199436187744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,8,power_law_1.2,0.04570879936218262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,1,power_law_1.2,2.112518310546875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,8,power_law_1.2,0.04917759895324707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,4,power_law_1.01,0.17252479791641234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,8,power_law_1.2,0.05050879716873169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,8,power_law_1.2,0.050291198492050174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,1,power_law_1.2,2.706675148010254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,8,power_law_1.2,0.057734400033950806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,4,power_law_1.01,0.17306239604949952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,8,power_law_1.2,0.05961599946022034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,8,power_law_1.2,0.060166400671005246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,4,power_law_1.01,0.18298879861831666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,1,power_law_1.2,3.3149440765380858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,8,power_law_1.2,0.07214720249176025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,8,power_law_1.2,0.08726400136947632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,4,power_law_1.01,0.20324480533599854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,8,power_law_1.2,0.10808320045471191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,64,balanced,0.03166399896144867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,64,balanced,0.02940266579389572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,64,balanced,0.02940800040960312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,64,balanced,0.03173333406448364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,64,balanced,0.02998399982849757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,64,balanced,0.03176533430814743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,64,balanced,0.02994133283694585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,8,power_law_1.2,0.11958400011062623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,64,balanced,0.03143999973932902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,64,balanced,0.033573334415753685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,64,balanced,0.031354665756225586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,64,balanced,0.0316746657093366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,64,balanced,0.03333866596221924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,64,balanced,0.033376000821590424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,64,balanced,0.03554133325815201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,64,balanced,0.03547733277082443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,64,balanced,0.03777066618204117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,64,balanced,0.03765333443880081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,64,balanced,0.03961600114901861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,64,balanced,0.03830400109291077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,64,balanced,0.04166933397452036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,1,power_law_1.2,5.126822280883789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,8,power_law_1.2,0.17694720029830932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,64,balanced,0.043706665436426796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,4,power_law_1.01,0.21943039894104005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,64,balanced,0.04577599962552389
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,64,balanced,0.05013866722583771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,64,balanced,0.06404800216356914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,64,balanced,0.08251200119654338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,64,balanced,0.11597333351771037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,64,balanced,0.1470186710357666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,8,power_law_1.2,0.2187903881072998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,64,balanced,0.17699732383092245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,64,balanced,0.24829866488774618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,64,balanced,0.43940265973409015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,4,power_law_1.01,0.26044158935546874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,8,power_law_1.2,0.2649728059768677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,8,power_law_1.2,0.4218751907348633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,4,power_law_1.01,0.30561280250549316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,8,power_law_1.2,0.6069888114929199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,4,power_law_1.01,0.37355520725250246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,1,power_law_1.2,9.748512268066406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,8,power_law_1.2,0.7216383934020996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,4,power_law_1.01,0.47349119186401367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,8,power_law_1.2,0.876467227935791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,4,power_law_1.01,0.6293248176574707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,8,power_law_1.2,1.5411007881164551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,4,power_law_1.01,0.8211775779724121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,4,power_law_1.01,1.159513568878174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,8,power_law_1.2,3.168383979797363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,4,power_law_1.01,1.4305983543395997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,32,power_law_1.2,0.08213760256767273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,32,power_law_1.2,0.06395519971847534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,32,power_law_1.2,0.05856000185012818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,4,power_law_1.01,1.720479965209961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,32,power_law_1.2,0.07163519859313965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,32,power_law_1.2,0.07054719924926758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,8,balanced,0.04172799984614054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,8,balanced,0.04565866788228353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,8,balanced,0.04390400151411692
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,8,balanced,0.05179200073083242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,8,balanced,0.0739519993464152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,8,balanced,0.10326932867368062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,32,power_law_1.2,0.07831040024757385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,8,balanced,0.10481066505114238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,8,balanced,0.10577066739400227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,8,balanced,0.10504532853762309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,8,balanced,0.10444800059000652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,8,balanced,0.10912000139554341
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,8,balanced,0.10915199915568034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,8,balanced,0.11131733655929565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,8,balanced,0.11150933305422465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,8,balanced,0.12078932921091716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,8,balanced,0.12176000078519185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,32,power_law_1.2,0.08002560138702393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,8,balanced,0.12737066547075906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,4,power_law_1.01,2.6067968368530274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,8,balanced,0.1409280002117157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,8,balanced,0.1527733306090037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,8,balanced,0.17870400349299112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,8,balanced,0.19946134090423584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,8,balanced,0.24597867329915366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,8,balanced,0.2912213404973348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,32,power_law_1.2,0.0815999984741211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,8,balanced,0.39397335052490234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,8,balanced,0.479039986928304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,8,balanced,0.6794079939524332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,8,balanced,0.8616426785786947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,32,power_law_1.2,0.08202239871025085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,8,balanced,1.0614720185597737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,8,balanced,1.6094080607096355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,32,power_law_1.2,0.08455039858818054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,8,balanced,3.1378774642944336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,32,power_law_1.2,0.08394880294799804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,32,power_law_1.2,0.08812159895896912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,32,power_law_1.2,0.08933759927749634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,4,power_law_1.01,5.4271808624267575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,32,power_law_1.2,0.09573760032653808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,32,balanced,0.033786666889985405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,32,balanced,0.03162133445342382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,32,balanced,0.02942399928967158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,32,balanced,0.03159466634194056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,32,balanced,0.031311998764673867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,32,balanced,0.029792000850041706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,32,power_law_1.2,0.10712319612503052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,32,balanced,0.03134933362404505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,32,balanced,0.0315733328461647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,32,balanced,0.03177600105603536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,32,balanced,0.031194667021433514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,32,balanced,0.03182933231194814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,32,balanced,0.03136000037193298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,32,balanced,0.03332266708215078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,32,balanced,0.033999999364217125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,32,balanced,0.035530666510264076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,32,balanced,0.03740799923737844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,32,balanced,0.035589332381884255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,32,balanced,0.03818133225043615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,32,balanced,0.03775999943415324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,32,balanced,0.04154133299986521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,32,balanced,0.041802664597829185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,32,power_law_1.2,0.10650880336761474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,32,balanced,0.04598399996757507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,32,balanced,0.04571733375390371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,32,balanced,0.05390933156013489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,32,balanced,0.06244266529877981
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,32,balanced,0.07846400141716003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,32,balanced,0.09226133426030476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,4,32,balanced,0.10755200187365214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,4,32,balanced,0.13818666338920593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,4,32,balanced,0.22410666942596436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,32,power_law_1.2,0.1109760046005249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,32,power_law_1.2,0.12762880325317383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,32,power_law_1.2,0.15472639799118043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,32,power_law_1.2,0.19570560455322267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,32,power_law_1.2,0.1980736017227173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,32,power_law_1.2,0.27124478816986086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,32,power_law_1.2,0.31320960521698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,32,power_law_1.2,0.4813248157501221
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,32,power_law_1.2,0.598195219039917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,4,64,balanced,0.024133334557215374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,4,64,balanced,0.02369600037733714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,32,power_law_1.2,0.8966079711914062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,4,64,balanced,0.025744001070658367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,4,64,balanced,0.023786666492621105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,4,64,balanced,0.023557332654794056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,4,64,balanced,0.025578667720158894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,4,64,balanced,0.026730666557947796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,4,64,balanced,0.025258667767047882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,4,64,balanced,0.025498665869235992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,4,64,balanced,0.025407999753952026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,4,64,balanced,0.027477333943049114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,4,64,balanced,0.027248000105222065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,4,64,balanced,0.027290667096773785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,4,64,balanced,0.029951999584833782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,4,64,balanced,0.03169066707293192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,4,64,balanced,0.03365866591533025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,4,64,balanced,0.03204799940188726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,4,64,balanced,0.033717334270477295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,64,balanced,0.03549866626660029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,64,balanced,0.035461333890755974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,64,balanced,0.03755199909210205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,64,balanced,0.037871999045213066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,64,balanced,0.03979733337958654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,64,balanced,0.04572799801826477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,64,balanced,0.047279998660087585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,32,power_law_1.2,1.0118656158447266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,64,balanced,0.05824000140031179
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,64,balanced,0.06706133484840393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,64,balanced,0.07867200175921123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,64,balanced,0.1049066682656606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,64,balanced,0.16107733050982156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,1,power_law_1.01,0.03498240113258362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,32,power_law_1.2,1.473087978363037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,1,power_law_1.01,0.04414080083370209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,32,power_law_1.2,2.3869184494018554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,32,power_law_1.01,0.07496960163116455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,1,power_law_1.01,0.051520001888275144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,1,power_law_1.01,0.06821119785308838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,32,power_law_1.01,0.0605184018611908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,1,power_law_1.01,0.08283519744873047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,32,power_law_1.01,0.05459840297698974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,1,power_law_1.01,0.11788159608840942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,32,power_law_1.2,4.508505630493164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,32,power_law_1.01,0.055961602926254274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,1,power_law_1.01,0.12327040433883667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,32,power_law_1.01,0.055961602926254274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,1,power_law_1.01,0.1266816020011902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,1,power_law_1.01,0.13283200263977052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,32,power_law_1.01,0.05860480070114136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,1,power_law_1.01,0.1361407995223999
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,32,power_law_1.01,0.06113280057907104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,1,power_law_1.01,0.14120320081710816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,32,power_law_1.01,0.06039680242538452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,2,4,power_law_1.2,0.05998079776763916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,1,power_law_1.01,0.14877439737319947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,2,4,power_law_1.2,0.07383679747581481
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,2,4,power_law_1.2,0.052902400493621826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,32,power_law_1.01,0.0593280017375946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,2,4,power_law_1.2,0.05919359922409058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,1,power_law_1.01,0.14911359548568726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,2,4,power_law_1.2,0.06058239936828613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,2,4,power_law_1.2,0.05985280275344849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,2,4,power_law_1.2,0.0617792010307312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,32,power_law_1.01,0.05958399772644043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,2,4,power_law_1.2,0.06343680024147033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,1,power_law_1.01,0.1566655993461609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,2,4,power_law_1.2,0.06510720252990723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,2,4,power_law_1.2,0.06753919720649719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,2,4,power_law_1.2,0.06846079826354981
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,32,power_law_1.01,0.06379520297050476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,2,4,power_law_1.2,0.07251840233802795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,1,power_law_1.01,0.16838400363922118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,2,4,power_law_1.2,0.07417600154876709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,2,4,power_law_1.2,0.07560960054397584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,2,4,power_law_1.2,0.09472000002861022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,32,power_law_1.01,0.06459519863128663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,2,4,power_law_1.2,0.0954367995262146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,1,power_law_1.01,0.1756991982460022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,2,4,power_law_1.2,0.10558719635009765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,2,4,power_law_1.2,0.132096004486084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,2,4,power_law_1.2,0.14581120014190674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,2,4,power_law_1.2,0.19434239864349365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,32,power_law_1.01,0.0650111973285675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,1,power_law_1.01,0.19029760360717773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,2,4,power_law_1.2,0.23808000087738038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,2,4,power_law_1.2,0.3164799928665161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,2,4,power_law_1.2,0.3806207895278931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,32,power_law_1.01,0.06903679966926575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,1,power_law_1.01,0.22810239791870118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,2,4,power_law_1.2,0.5797311782836914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,2,4,power_law_1.2,0.797657585144043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,1,power_law_1.01,0.25073919296264646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,32,power_law_1.01,0.0732159972190857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,2,4,power_law_1.2,1.1989248275756836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,1,balanced,0.050213331977526345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,1,balanced,0.05459199845790863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,1,balanced,0.07074133555094402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,1,balanced,0.09872532884279887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,1,balanced,0.14614933729171753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,2,4,power_law_1.2,1.341478443145752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,1,balanced,0.18052266041437784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,1,power_law_1.01,0.31902079582214354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,1,balanced,0.18211734294891357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,32,power_law_1.01,0.07464320063591004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,1,balanced,0.18309332927068075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,1,balanced,0.1842986742655436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,1,balanced,0.18600533405939737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,1,balanced,0.18658665815989176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,1,balanced,0.1896053353945414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,1,balanced,0.1933599909146627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,1,balanced,0.19567465782165527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,2,4,power_law_1.2,1.8100608825683593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,1,balanced,0.20197333892186484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,1,balanced,0.20480000972747803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,1,balanced,0.21566933393478394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,1,balanced,0.23369600375493368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,1,power_law_1.01,0.3634752035140991
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,1,balanced,0.25176533063252765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,32,power_law_1.01,0.08464000225067139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,1,balanced,0.36927998065948486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,1,balanced,0.3344586690266927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,1,balanced,0.49302931626637775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,1,balanced,0.527237335840861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,1,balanced,0.7654826641082764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,2,4,power_law_1.2,2.701420783996582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,1,balanced,0.9126079877217611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,1,power_law_1.01,0.4814015865325928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,32,power_law_1.01,0.09705600142478943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,1,balanced,1.3381919860839844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,1,balanced,1.6704959869384766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,2688,1856,6,128,1,1,balanced,2.0942400296529136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,1,power_law_1.01,0.5963200092315674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,32,power_law_1.01,0.10355199575424194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,2,4,power_law_1.2,5.2152446746826175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,2688,1856,6,128,1,1,balanced,3.242527961730957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,32,power_law_1.01,0.12327040433883667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,1,power_law_1.01,0.8298687934875488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,2688,1856,6,128,1,1,balanced,6.42790412902832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,32,power_law_1.01,0.13435519933700563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,1,power_law_1.01,1.0486463546752929
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,32,power_law_1.01,0.17644799947738649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,1,power_law_1.01,1.4927616119384766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,32,power_law_1.01,0.21105918884277344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,1,power_law_1.01,1.9776128768920898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,32,power_law_1.01,0.2805248022079468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,32,power_law_1.01,0.34639999866485593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,1,power_law_1.01,2.431865692138672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,32,power_law_1.01,0.4942592144012451
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,1,power_law_1.01,3.7216190338134765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,32,power_law_1.01,0.5540800094604492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,32,power_law_1.01,0.7556863784790039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,32,power_law_1.01,1.1341952323913573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,1,power_law_1.01,7.307456207275391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,16,power_law_1.2,0.15000959634780883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,32,4,power_law_1.01,0.06155520081520081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,16,power_law_1.2,0.24373118877410888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,32,4,power_law_1.01,0.06603519916534424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,32,power_law_1.01,2.163680076599121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,16,power_law_1.2,0.19633920192718507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,32,4,power_law_1.01,0.08028159737586975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,16,power_law_1.2,0.111244797706604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,32,4,power_law_1.01,0.06636800169944763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,16,power_law_1.2,0.14755200147628783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,32,4,power_law_1.01,0.07436800003051758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,16,power_law_1.2,0.16864639520645142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,32,4,power_law_1.01,0.08865919709205627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,16,power_law_1.2,0.1820032000541687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,32,4,power_law_1.01,0.0949184000492096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,16,power_law_1.2,0.24184958934783934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,32,4,power_law_1.01,0.10220160484313964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,8,1,power_law_1.01,0.03979519903659821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,16,power_law_1.2,0.2450495958328247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,8,1,power_law_1.01,0.04785279929637909
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,8,1,power_law_1.01,0.052902400493621826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,8,1,power_law_1.01,0.07624319791793824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,32,4,power_law_1.01,0.1083840012550354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,8,1,power_law_1.01,0.07804800271987915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,16,power_law_1.2,0.264684796333313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,8,1,power_law_1.01,0.07876480221748353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,8,1,power_law_1.01,0.07818880081176757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,8,1,power_law_1.01,0.07938560247421264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,32,4,power_law_1.01,0.10667519569396973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,8,1,power_law_1.01,0.08421120047569275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,16,power_law_1.2,0.27464320659637453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,8,1,power_law_1.01,0.08327040076255798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,8,1,power_law_1.01,0.08383359909057617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,32,4,power_law_1.01,0.1087231993675232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,8,1,power_law_1.01,0.08918399810791015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,32,power_law_1.2,0.039801600575447085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,16,power_law_1.2,0.28488318920135497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,8,1,power_law_1.01,0.08980479836463928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,32,power_law_1.2,0.034892800450325015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,8,1,power_law_1.01,0.09537280201911927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,32,4,power_law_1.01,0.11534080505371094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,8,1,power_law_1.01,0.11015679836273193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,16,power_law_1.2,0.2920896053314209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,32,power_law_1.2,0.03491199910640717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,8,1,power_law_1.01,0.11198079586029053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,8,1,power_law_1.01,0.11314560174942016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,32,power_law_1.2,0.03394559919834137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,8,1,power_law_1.01,0.13333120346069335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,32,4,power_law_1.01,0.11784960031509399
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,16,power_law_1.2,0.28741118907928465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,8,1,power_law_1.01,0.14901119470596313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,32,power_law_1.2,0.034892800450325015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,8,1,power_law_1.01,0.1893887996673584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,32,power_law_1.2,0.03611519932746887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,8,1,power_law_1.01,0.22826879024505614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,32,4,power_law_1.01,0.12142080068588257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,8,1,power_law_1.01,0.30266239643096926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,32,power_law_1.2,0.036287999153137206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,16,power_law_1.2,0.31466240882873536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,8,1,power_law_1.01,0.36239359378814695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,32,power_law_1.2,0.036294400691986084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,8,1,power_law_1.01,0.5025343894958496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,32,4,power_law_1.01,0.1295040011405945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,32,power_law_1.2,0.03648000061511993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,8,1,power_law_1.01,0.6356927871704101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,16,power_law_1.2,0.3019711971282959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,8,1,power_law_1.01,0.9135616302490235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,32,power_law_1.2,0.03688960075378418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,32,4,power_law_1.01,0.1301375985145569
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,32,power_law_1.2,0.037510401010513304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,8,1,power_law_1.01,1.1793984413146972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,16,power_law_1.2,0.33797121047973633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,32,power_law_1.2,0.03868800103664398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,8,1,power_law_1.01,1.452620792388916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,32,4,power_law_1.01,0.14013439416885376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,32,power_law_1.2,0.03950720131397247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,16,power_law_1.2,0.37505919933319093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,32,power_law_1.2,0.0412416011095047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,8,1,power_law_1.01,2.2803327560424806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,32,4,power_law_1.01,0.1587007999420166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,32,power_law_1.2,0.04526079893112182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,16,power_law_1.2,0.40867199897766116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,32,power_law_1.2,0.04666880071163178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,32,power_law_1.2,0.04936319887638092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,32,4,power_law_1.01,0.1707584023475647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,8,1,power_law_1.01,4.435859298706054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,32,power_law_1.2,0.05624960064888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,16,power_law_1.2,0.4889472007751465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,32,power_law_1.2,0.05663359761238098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,32,4,power_law_1.01,0.211296010017395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,32,power_law_1.2,0.06974080204963684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,16,power_law_1.2,0.5157055854797363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,32,power_law_1.2,0.08014079928398132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,32,4,power_law_1.01,0.238700795173645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,32,power_law_1.2,0.11327999830245972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,16,power_law_1.2,0.6357312202453613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,32,power_law_1.2,0.13678079843521118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,32,4,power_law_1.01,0.3096640110015869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,32,power_law_1.2,0.17261439561843872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,16,power_law_1.2,0.7002111911773682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,32,power_law_1.2,0.23578879833221436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,32,4,power_law_1.01,0.3756799936294556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,32,power_law_1.2,0.3305279970169067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,16,power_law_1.2,0.9111040115356446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,32,4,power_law_1.01,0.5250944137573242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,32,power_law_1.2,0.4336063861846924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,16,power_law_1.2,0.8608960151672364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,32,power_law_1.2,0.6386943817138672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,32,4,power_law_1.01,0.6709311962127685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,16,power_law_1.2,1.1891200065612793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,32,power_law_1.2,0.9047871589660644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,32,4,power_law_1.01,0.985267162322998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,2,8,balanced,0.06145066519578298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,2,8,balanced,0.08705600102742513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,2,8,balanced,0.061386664708455406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,2,8,balanced,0.06102933486302694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,2,8,balanced,0.061349332332611084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,2,8,balanced,0.06301333506902058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,16,power_law_1.2,1.6367168426513672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,2,8,balanced,0.061333333452542625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,2,8,balanced,0.06483733157316844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,2,8,balanced,0.06753600140412648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,2,8,balanced,0.06592533489068349
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,32,4,power_law_1.01,1.2278271675109864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,2,8,balanced,0.06477866570154826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,2,8,balanced,0.06575466692447662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,2,8,balanced,0.06554666658242543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,2,8,balanced,0.06492800017197926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,2,8,balanced,0.07362133264541626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,2,8,balanced,0.07341866691907246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,2,8,balanced,0.07634133100509644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,32,power_law_1.2,1.8880575180053711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,2,8,balanced,0.084197332461675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,2,8,balanced,0.08176533381144206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,2,8,balanced,0.10071466366449992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,2,8,balanced,0.10447466373443604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,2,8,balanced,0.14402133226394653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,2,8,balanced,0.15371732910474142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,2,8,balanced,0.2125920057296753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,2,8,balanced,0.24597332874933878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,2,8,balanced,0.3455520073572795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,2,8,balanced,0.4466559886932373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,2,8,balanced,0.543994665145874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,16,power_law_1.2,1.957049560546875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,2,8,balanced,0.8183733622233073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,32,4,power_law_1.01,1.5429887771606445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,2,8,balanced,1.6005493799845378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,16,power_law_1.2,2.819987106323242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,32,4,power_law_1.01,2.4304576873779298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,16,power_law_1.2,4.961740875244141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,32,4,power_law_1.01,4.588025665283203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,1,power_law_1.01,0.09644799828529357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,1,power_law_1.01,0.14653439521789552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,1,power_law_1.2,0.09597439765930176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,1,power_law_1.01,0.15803519487380982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,1,power_law_1.2,0.10287359952926636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,1,power_law_1.01,0.33545598983764646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,1,power_law_1.2,0.1472383975982666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,1,power_law_1.01,0.536684799194336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,1,power_law_1.2,0.210534405708313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,1,power_law_1.01,0.7689663887023925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,1,power_law_1.2,0.28448638916015623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,1,power_law_1.01,1.0869183540344238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,1,power_law_1.2,0.5183551788330079
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,1,power_law_1.2,0.5759679794311523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,1,power_law_1.01,1.1504704475402832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,8,balanced,0.039919999738534294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,8,balanced,0.04386133452256521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,8,balanced,0.043925335009892784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,2,8,power_law_1.2,0.07800959944725036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,8,balanced,0.04625066618124644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,8,balanced,0.052058666944503784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,8,balanced,0.054485330979029335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,8,balanced,0.0581226646900177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,2,8,power_law_1.2,0.11934720277786255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,8,balanced,0.05622933308283488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,8,balanced,0.05621333420276642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,8,balanced,0.058261334896087646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,2,8,power_law_1.2,0.07746559977531434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,8,balanced,0.05820266902446747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,8,balanced,0.05823466678460439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,8,balanced,0.06041066845258077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,2,8,power_law_1.2,0.07777919769287109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,8,balanced,0.06368533273537953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,1,power_law_1.01,1.2123968124389648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,8,balanced,0.06654933094978333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,2,8,power_law_1.2,0.07835519909858704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,8,balanced,0.07176533341407776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,8,balanced,0.07057066758473714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,8,balanced,0.08097066481908162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,2,8,power_law_1.2,0.08188160061836243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,8,balanced,0.09479999542236328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,8,balanced,0.12165333827336629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,8,balanced,0.1427839994430542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,2,8,power_law_1.2,0.09923200011253357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,8,balanced,0.19142399231592813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,1,power_law_1.2,0.6126016139984131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,8,balanced,0.21596799294153848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,2,8,power_law_1.2,0.0885312020778656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,8,balanced,0.31304534276326496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,2,8,power_law_1.2,0.09162880182266235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,8,balanced,0.37539732456207275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,8,balanced,0.5323359966278076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,2,8,power_law_1.2,0.09425920248031616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,1,power_law_1.01,1.2452544212341308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,8,balanced,0.6810452938079834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,2,8,power_law_1.2,0.09052799940109253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,8,balanced,0.8295626640319824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,2,8,power_law_1.2,0.09886720180511474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,2,8,power_law_1.2,0.09807999730110169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,8,balanced,1.2811306317647297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,2,8,power_law_1.2,0.12741119861602784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,2,8,power_law_1.2,0.16106879711151123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,1,power_law_1.2,0.6367680072784424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,2,8,power_law_1.2,0.16836479902267457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,8,balanced,2.5015947024027505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,1,power_law_1.01,1.2983488082885741
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,2,8,power_law_1.2,0.14639359712600708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,2,8,power_law_1.2,0.17666560411453247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,2,8,power_law_1.2,0.22960000038146972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,2,8,power_law_1.2,0.34088959693908694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,2,8,power_law_1.2,0.3615295886993408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,1,power_law_1.2,0.6614975929260254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,1,power_law_1.01,1.3298175811767579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,2,8,power_law_1.2,0.46616320610046386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,2,8,power_law_1.2,0.7098432064056397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,2,8,power_law_1.2,1.0783295631408691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,1,power_law_1.01,1.3739839553833009
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,1,power_law_1.2,0.6759935855865479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,2,8,power_law_1.2,1.1814144134521485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,2,8,power_law_1.2,1.7386240005493163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,1,power_law_1.01,1.4093759536743165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,1,power_law_1.2,0.6892799854278564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,2,8,power_law_1.2,2.565465545654297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,1,power_law_1.01,1.4107839584350585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,1,power_law_1.2,0.6906496047973633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,2,8,power_law_1.2,2.9750463485717775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,1,power_law_1.01,1.4520768165588378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,1,power_law_1.2,0.7150527954101562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,2,8,power_law_1.2,4.8881278991699215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,1,power_law_1.01,1.5389887809753418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,1,power_law_1.2,0.7311488151550293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,1,power_law_1.01,1.629465675354004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,1,power_law_1.2,0.7622975826263427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,2,8,power_law_1.2,9.557516479492188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,1,power_law_1.2,0.7834432125091553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,1,power_law_1.01,1.6450624465942383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,1,power_law_1.2,0.8406847953796387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,1,power_law_1.01,1.856287956237793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,32,balanced,0.052341332038243614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,32,balanced,0.044122666120529175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,32,balanced,0.04271999994913737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,32,balanced,0.04460266729195913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,32,balanced,0.048058668772379555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,32,balanced,0.06012799839178721
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,32,balanced,0.06211733321348826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,32,balanced,0.0625600020090739
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,1,power_law_1.2,0.9387904167175293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,32,balanced,0.06251200040181477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,32,balanced,0.06117333471775055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,32,balanced,0.06660266717274983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,32,balanced,0.06311999758084615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,32,balanced,0.06665599842866261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,32,balanced,0.06663466493288676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,32,balanced,0.07102933526039124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,32,balanced,0.07102933526039124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,32,balanced,0.07412800192832947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,32,balanced,0.07850666840871175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,1,power_law_1.01,1.9264320373535155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,32,balanced,0.080485333998998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,32,balanced,0.09275733431180318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,32,balanced,0.09583466251691182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,32,balanced,0.11761599779129028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,32,balanced,0.1323946714401245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,32,balanced,0.17325333754221597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,32,balanced,0.18525334199269614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,1,power_law_1.2,1.0997183799743653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,32,balanced,0.23847466707229614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,32,balanced,0.30634133021036786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,32,balanced,0.3636053403218587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,32,balanced,0.5290559927622477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,32,balanced,0.9927732944488525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,1,power_law_1.01,2.2352447509765625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,1,power_law_1.2,1.1624256134033204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,1,power_law_1.2,1.4308608055114747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,1,power_law_1.01,2.427987289428711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,1,power_law_1.2,1.6856319427490234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,1,power_law_1.01,2.9389184951782226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,1,power_law_1.2,2.184998321533203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,1,power_law_1.01,3.3691455841064455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,1,power_law_1.2,2.8377471923828126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,2,balanced,0.033941333492596946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,2,balanced,0.032842665910720825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,2,balanced,0.035573333501815796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,1,power_law_1.01,4.313241577148437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,2,balanced,0.04067733387152354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,2,balanced,0.05937600135803223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,2,balanced,0.06214400132497152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,2,balanced,0.06358933448791504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,2,balanced,0.06121066709359487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,2,balanced,0.06413333117961884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,2,balanced,0.06462400158246358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,2,balanced,0.06623999774456024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,2,balanced,0.0684746652841568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,2,balanced,0.06834666430950165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,2,balanced,0.07270933190981548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,2,balanced,0.07803733150164287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,2,balanced,0.08063466846942902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,2,balanced,0.0846560001373291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,2,balanced,0.09497599800427754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,1,power_law_1.2,3.822969436645508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,2,balanced,0.10311466455459595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,2,balanced,0.129530668258667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,2,balanced,0.14121599992116293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,2,balanced,0.18604799111684164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,2,balanced,0.21924267212549844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,2,balanced,0.3253493309020996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,2,balanced,0.41437331835428876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,2,balanced,0.6095893383026123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,2,balanced,0.8075146675109863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,2,balanced,0.999077320098877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,2,balanced,1.5711092948913574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,1,power_law_1.01,5.329049682617187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,2,balanced,2.984837214152018
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,1,power_law_1.2,4.71794548034668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,16,balanced,0.04299733539422353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,16,balanced,0.04567466676235199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,16,balanced,0.04401599864164988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,16,balanced,0.04789866507053375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,16,balanced,0.07573333382606506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,16,balanced,0.0827466646830241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,16,balanced,0.08019199967384338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,16,balanced,0.08134399851163228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,16,balanced,0.08361599842707317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,16,balanced,0.0817440003156662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,16,balanced,0.08255999783674876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,16,balanced,0.08217599987983704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,16,balanced,0.08301866551240285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,16,balanced,0.08509866396586101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,16,balanced,0.0888213316599528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,16,balanced,0.08989333113034566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,16,balanced,0.09248532851537068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,16,balanced,0.10467200477917989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,16,balanced,0.10207466284434001
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,16,balanced,0.12107200423876445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,1,power_law_1.01,6.3282623291015625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,16,balanced,0.1223520040512085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,16,balanced,0.15043733517328897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,16,balanced,0.16195199886957803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,16,balanced,0.2262293299039205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,16,balanced,0.2512106696764628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,1,power_law_1.2,5.6966911315917965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,16,balanced,0.33312533299128216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,16,balanced,0.4040000041325887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,16,balanced,0.4628320137659709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,16,balanced,0.712112029393514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,16,balanced,1.3286186854044597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,1,power_law_1.01,9.311302185058594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,1,power_law_1.2,8.625107574462891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,1,4,power_law_1.01,0.13446400165557862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,1,4,power_law_1.01,0.10969599485397338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,1,4,power_law_1.01,0.11989120244979859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,1,4,power_law_1.01,0.13388799428939818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,1,4,power_law_1.01,0.1349184036254883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,1,power_law_1.2,16.298892211914062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,1,4,power_law_1.01,0.1390720009803772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,1,power_law_1.01,17.271852111816408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,1,4,power_law_1.01,0.15441919565200807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,1,4,power_law_1.01,0.16327680349349977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,1,4,power_law_1.01,0.18218239545822143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,1,4,power_law_1.01,0.212774395942688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,1,4,power_law_1.01,0.1530176043510437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,1,4,power_law_1.01,0.164518404006958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,1,4,power_law_1.01,0.1674496054649353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,1,4,power_law_1.01,0.19050240516662598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,1,4,power_law_1.01,0.20264320373535155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,1,4,power_law_1.01,0.22646400928497315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,1,4,power_law_1.01,0.23206400871276855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,1,4,power_law_1.01,0.26966400146484376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,1,balanced,0.03772799919048945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,1,4,power_law_1.01,0.4037631988525391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,1,balanced,0.06900266806284587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,1,balanced,0.07709333300590515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,1,balanced,0.12098133563995361
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,1,balanced,0.2044853369394938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,1,4,power_law_1.01,0.5220352172851562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,1,balanced,0.20808533827463785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,1,balanced,0.20879999796549478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,1,balanced,0.21073599656422934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,1,balanced,0.21264533201853433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,1,balanced,0.22155199448267618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,1,4,power_law_1.01,0.6072959899902344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,1,balanced,0.2166666587193807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,1,balanced,0.22064000368118286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,1,balanced,0.22219733397165933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,1,balanced,0.22649067640304565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,1,balanced,0.23679467042287192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,1,4,power_law_1.01,0.8575615882873535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,1,balanced,0.24435200293858847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,1,balanced,0.2515146732330322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,1,balanced,0.27881066004435223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,1,balanced,0.29677865902582806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,1,balanced,0.35381333033243817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,1,4,power_law_1.01,1.055116844177246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,1,balanced,0.38822933038075763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,1,balanced,0.510586659113566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,1,balanced,0.5865813493728638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,1,4,power_law_1.01,1.4765503883361817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,1,balanced,0.8132320245107015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,1,balanced,1.012837330500285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,1,balanced,1.4563840230305989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,1,4,power_law_1.01,1.9515008926391602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,1,balanced,1.8121600151062012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,1,balanced,2.21946128209432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,1,4,power_law_1.01,2.835487937927246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,1,balanced,3.4028746287027993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,1,4,power_law_1.01,3.2726718902587892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,1,balanced,6.57102902730306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,1,4,power_law_1.01,4.556435012817383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,1,4,power_law_1.01,7.5829315185546875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,8,balanced,0.04607999821503957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,8,balanced,0.048138668139775596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,8,balanced,0.049770668148994446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,8,balanced,0.07544533411661784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,8,balanced,0.10603732864061992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,8,balanced,0.11528000235557556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,8,balanced,0.11793599526087443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,8,balanced,0.11888532837231953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,8,balanced,0.11877866586049397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,8,balanced,0.11947199702262878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,8,balanced,0.12215466300646464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,1,1,power_law_1.2,0.21755518913269042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,8,balanced,0.12174399693806966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,8,balanced,0.1236799955368042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,8,balanced,0.12683733304341635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,8,balanced,0.13219199577967325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,8,balanced,0.13540266950925192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,8,balanced,0.13724266489346823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,1,1,power_law_1.2,0.2723711967468262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,8,balanced,0.1532799998919169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,8,balanced,0.15798933307329813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,8,balanced,0.18669867515563965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,8,balanced,0.20010133584340414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,8,balanced,0.2528853416442871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,1,1,power_law_1.2,0.4477503776550293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,8,balanced,0.28382933139801025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,8,balanced,0.4129653374354045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,8,balanced,0.4736106793085734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,8,balanced,0.6889866987864176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,1,1,power_law_1.2,0.6676991939544678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,8,balanced,0.8349760373433431
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,8,balanced,0.9834826787312826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,1,1,power_law_1.2,0.7187903881072998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,8,balanced,1.5288267135620117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,1,4,power_law_1.01,15.772479248046874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,1,1,power_law_1.2,0.7385983943939209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,8,balanced,2.919557253519694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,1,1,power_law_1.2,0.7495232105255127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,1,1,power_law_1.2,0.7574399948120117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,1,1,power_law_1.2,0.7676095962524414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,1,1,power_law_1.2,0.7879680156707763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,1,1,power_law_1.2,0.8207488059997559
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,1,1,power_law_1.2,0.8121472358703613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,1,1,power_law_1.2,0.8328831672668457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,1,1,power_law_1.2,0.8437888145446777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,1,1,power_law_1.2,0.8845120429992676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,1,1,power_law_1.2,0.9179583549499511
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,1,1,power_law_1.2,0.9820672035217285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,1,1,power_law_1.2,1.0872703552246095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,1,1,power_law_1.2,1.2363712310791015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,1,1,power_law_1.2,1.504319953918457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,256,power_law_1.01,0.06427519917488098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,1,1,power_law_1.2,1.6947839736938477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,1,1,power_law_1.2,2.2654272079467774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,1,1,power_law_1.2,2.7922496795654297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,256,power_law_1.01,0.06762239933013917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,1,1,power_law_1.2,3.9663230895996096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,256,power_law_1.01,0.05934720039367676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,1,1,power_law_1.2,5.257414245605469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,256,power_law_1.01,0.06144639849662781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,1,1,power_law_1.2,7.508659362792969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,2,1,power_law_1.01,0.07782400250434876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,2,1,power_law_1.01,0.09845119714736938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,256,power_law_1.01,0.060524797439575194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,2,1,power_law_1.01,0.11347839832305909
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,2,1,power_law_1.01,0.2052288055419922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,2,1,power_law_1.01,0.19791359901428224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,2,1,power_law_1.01,0.20766079425811768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,2,1,power_law_1.01,0.21098880767822265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,64,power_law_1.2,0.10483839511871337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,2,1,power_law_1.01,0.2167423963546753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,2,1,power_law_1.01,0.2222975969314575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,256,power_law_1.01,0.06789759993553161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,2,1,power_law_1.01,0.22156798839569092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,1,1,power_law_1.2,9.896646118164062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,2,1,power_law_1.01,0.21985280513763428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,2,1,power_law_1.01,0.22298240661621094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,2,1,power_law_1.01,0.23025920391082763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,2,1,power_law_1.01,0.25367040634155275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,2,1,power_law_1.01,0.27681920528411863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,64,power_law_1.2,0.08699520230293274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,2,1,power_law_1.01,0.28846719264984133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,2,1,power_law_1.01,0.28424320220947263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,256,power_law_1.01,0.06225919723510742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,2,1,power_law_1.01,0.32899839878082277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,2,1,power_law_1.01,0.36945281028747556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,2,1,power_law_1.01,0.45582079887390137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,2,1,power_law_1.01,0.541542387008667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,64,power_law_1.2,0.07327359914779663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,2,1,power_law_1.01,0.7189311981201172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,256,power_law_1.01,0.064300799369812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,1,1,power_law_1.2,12.580960083007813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,2,1,power_law_1.01,0.8826560020446778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,2,1,power_law_1.01,1.2425663948059082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,2,1,power_law_1.01,1.5946816444396972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,64,power_law_1.2,0.07916799783706666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,256,power_law_1.01,0.06362239718437195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,2,1,power_law_1.01,2.33251838684082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,2,1,power_law_1.01,3.0649152755737306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,64,power_law_1.2,0.08817920088768005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,256,power_law_1.01,0.06313599944114685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,2,1,power_law_1.01,3.8276031494140623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,256,power_law_1.01,0.06490240097045899
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,64,power_law_1.2,0.09874560236930847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,1,1,power_law_1.2,19.78954315185547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,2,1,power_law_1.01,5.87509765625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,256,power_law_1.01,0.06611199975013733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,64,power_law_1.2,0.10076800584793091
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,256,power_law_1.01,0.07010560035705567
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,64,power_law_1.2,0.1055616021156311
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,2,1,power_law_1.01,11.429062652587891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,256,power_law_1.01,0.06726400256156921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,64,power_law_1.2,0.10391039848327636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,256,power_law_1.01,0.07186560034751892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,64,power_law_1.2,0.10325759649276733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,256,power_law_1.01,0.07214720249176025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,64,power_law_1.2,0.10269440412521362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,1,1,power_law_1.2,39.597650146484376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,256,power_law_1.01,0.07581440210342408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,64,power_law_1.2,0.10819200277328492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,256,power_law_1.01,0.09192320108413696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,64,power_law_1.2,0.11052800416946411
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,256,power_law_1.01,0.10689280033111573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,64,power_law_1.2,0.11026560068130493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,256,power_law_1.01,0.1357759952545166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,64,power_law_1.2,0.11345280408859253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,2,power_law_1.2,0.10164480209350586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,256,power_law_1.01,0.16286720037460328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,64,power_law_1.2,0.1316799998283386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,2,power_law_1.2,0.08569599986076355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,2,power_law_1.2,0.122707200050354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,64,power_law_1.2,0.1502079963684082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,2,power_law_1.2,0.17189120054244994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,256,power_law_1.01,0.21679999828338622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,2,power_law_1.2,0.26823039054870607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,2,power_law_1.2,0.45217280387878417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,64,power_law_1.2,0.17683199644088746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,256,power_law_1.01,0.23471360206604003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,2,power_law_1.2,0.49171199798583987
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,2,power_law_1.2,0.5021247863769531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,64,power_law_1.2,0.16726399660110475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,2,power_law_1.2,0.5296448230743408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,8,2,power_law_1.01,0.05898879766464234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,256,power_law_1.01,0.3023360013961792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,2,power_law_1.2,0.5476607799530029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,64,power_law_1.2,0.22001280784606933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,2,power_law_1.2,0.5562496185302734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,8,2,power_law_1.01,0.061887997388839724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,2,power_law_1.2,0.5527935981750488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,256,power_law_1.01,0.39220480918884276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,8,2,power_law_1.01,0.06723200082778931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,2,power_law_1.2,0.5743552207946777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,64,power_law_1.2,0.23882880210876464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,2,power_law_1.2,0.5898176193237304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,8,2,power_law_1.01,0.07807999849319458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,2,power_law_1.2,0.6176447868347168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,64,power_law_1.2,0.2998208045959473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,8,2,power_law_1.01,0.09073280096054077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,2,power_law_1.2,0.5952000141143798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,256,power_law_1.01,0.5313856124877929
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,2,power_law_1.2,0.6704895973205567
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,8,2,power_law_1.01,0.11446399688720703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,64,power_law_1.2,0.3146752119064331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,2,power_law_1.2,0.7567232131958008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,8,2,power_law_1.01,0.12193280458450317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,2,power_law_1.2,0.7444032192230224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,8,2,power_law_1.01,0.12535040378570556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,256,power_law_1.01,0.6695295810699463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,2,power_law_1.2,0.8579263687133789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,64,power_law_1.2,0.41566081047058107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,8,2,power_law_1.01,0.1273344039916992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,2,power_law_1.2,1.0185536384582519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,8,2,power_law_1.01,0.13182719945907592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,2,power_law_1.2,1.2691328048706054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,64,power_law_1.2,0.5174015998840332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,8,2,power_law_1.01,0.13334399461746216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,2,power_law_1.2,1.2104576110839844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,1,256,power_law_1.01,0.841977596282959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,8,2,power_law_1.01,0.13542399406433106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,2,power_law_1.2,1.5782912254333497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,8,2,power_law_1.01,0.13745919466018677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,2,power_law_1.2,1.8666175842285155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,64,power_law_1.2,0.8292991638183593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,8,2,power_law_1.01,0.14364800453186036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,2,power_law_1.2,2.6558464050292967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,8,2,power_law_1.01,0.15034879446029664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,8,2,power_law_1.01,0.15631359815597534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,1,256,power_law_1.01,1.3099648475646972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,64,power_law_1.2,1.1246015548706054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,2,power_law_1.2,3.3261249542236326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,8,2,power_law_1.01,0.16547839641571044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,8,2,power_law_1.01,0.19033600091934205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,2,power_law_1.2,4.5171966552734375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,8,2,power_law_1.01,0.21687679290771483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,8,2,power_law_1.01,0.266374397277832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,1,64,power_law_1.2,1.1910400390625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,8,2,power_law_1.01,0.30967040061950685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,2,power_law_1.2,5.8254142761230465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,8,2,power_law_1.01,0.416864013671875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,8,2,power_law_1.01,0.5193024158477784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,1,256,power_law_1.01,2.7437376022338866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,8,2,power_law_1.01,0.7142848014831543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,1,64,power_law_1.2,1.6982656478881837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,8,2,power_law_1.01,0.9853952407836915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,2,power_law_1.2,12.607788848876954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,8,2,power_law_1.01,1.2993535995483398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,8,2,power_law_1.01,1.7109056472778321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,8,2,power_law_1.01,2.0856319427490235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,1,64,power_law_1.2,4.212006378173828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,8,2,power_law_1.01,3.001113510131836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,4,power_law_1.2,0.055622398853302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,8,2,power_law_1.01,6.335468673706055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,4,power_law_1.2,0.06358399987220764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,8,8,power_law_1.01,0.034272000193595886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,4,power_law_1.2,0.05943040251731872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,8,8,power_law_1.01,0.03583999872207642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,8,8,power_law_1.01,0.031200000643730165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,8,8,power_law_1.01,0.03036159873008728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,8,8,power_law_1.01,0.030707201361656188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,8,8,power_law_1.01,0.031071999669075014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,4,power_law_1.2,0.06672639846801758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,8,8,power_law_1.01,0.03253760039806366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,8,8,power_law_1.01,0.03325439989566803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,8,8,power_law_1.01,0.03287039995193482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,8,8,power_law_1.01,0.03689599931240082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,8,8,power_law_1.01,0.03519999980926514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,4,power_law_1.2,0.0773311972618103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,8,8,power_law_1.01,0.040915200114250184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,8,8,power_law_1.01,0.04056319892406464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,8,8,power_law_1.01,0.03961600065231323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,8,8,power_law_1.01,0.048876801133155824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,8,8,power_law_1.01,0.05120000243186951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,4,power_law_1.2,0.09371520280838012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,8,8,power_law_1.01,0.05629439949989319
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,8,8,power_law_1.01,0.058182400465011594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,8,8,power_law_1.01,0.06974719762802124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,8,8,power_law_1.01,0.08794239759445191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,8,8,power_law_1.01,0.0955839991569519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,4,power_law_1.2,0.11333760023117065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,8,8,power_law_1.01,0.1412608027458191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,8,8,power_law_1.01,0.1454208016395569
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,8,8,power_law_1.01,0.21415040493011475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,4,power_law_1.2,0.1152575969696045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,8,8,power_law_1.01,0.2536895990371704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,8,8,power_law_1.01,0.41139841079711914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,8,8,power_law_1.01,0.44046721458435056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,4,power_law_1.2,0.11986559629440308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,128,power_law_1.01,0.10900479555130005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,8,8,power_law_1.01,0.5887743949890136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,8,8,power_law_1.01,0.8752384185791016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,4,power_law_1.2,0.12100479602813721
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,128,power_law_1.01,0.12735359668731688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,4,power_law_1.2,0.12436480522155761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,8,8,power_law_1.01,1.7009727478027343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,4,power_law_1.2,0.13134080171585083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,128,power_law_1.01,0.10647679567337036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,4,power_law_1.2,0.1298367977142334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,128,power_law_1.01,0.09388800263404846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,4,power_law_1.2,0.14117759466171265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,32,power_law_1.2,0.03503359854221344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,128,power_law_1.01,0.09998720288276672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,4,power_law_1.2,0.15593600273132324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,32,power_law_1.2,0.032262399792671204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,4,power_law_1.2,0.1583616018295288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,128,power_law_1.01,0.10268160104751586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,32,power_law_1.2,0.032364800572395325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,4,power_law_1.2,0.16362240314483642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,32,power_law_1.2,0.03129599988460541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,128,power_law_1.01,0.10701440572738648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,4,power_law_1.2,0.18171520233154298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,32,power_law_1.2,0.03201920092105866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,128,power_law_1.01,0.11047680377960205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,4,power_law_1.2,0.19845759868621826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,32,power_law_1.2,0.03370879888534546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,4,power_law_1.2,0.23608319759368895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,128,power_law_1.01,0.11353600025177002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,32,power_law_1.2,0.03394559919834137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,4,power_law_1.2,0.2678976058959961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,32,power_law_1.2,0.03442560136318207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,128,power_law_1.01,0.1144320011138916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,4,power_law_1.2,0.346233606338501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,32,power_law_1.2,0.03484799861907959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,128,power_law_1.01,0.11800960302352906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,32,power_law_1.2,0.0351936012506485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,4,power_law_1.2,0.4058112144470215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,8,power_law_1.01,0.0551360011100769
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,32,power_law_1.2,0.03606399893760681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,128,power_law_1.01,0.12031359672546386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,4,power_law_1.2,0.5582143783569335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,32,power_law_1.2,0.03770880103111267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,8,power_law_1.01,0.0637503981590271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,128,power_law_1.01,0.11953279972076417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,4,power_law_1.2,0.7080959796905517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,32,power_law_1.2,0.04073599874973297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,8,power_law_1.01,0.05863040089607239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,32,power_law_1.2,0.04247680008411407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,128,power_law_1.01,0.11712640523910522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,8,power_law_1.01,0.05912960171699524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,4,power_law_1.2,1.0482239723205566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,32,power_law_1.2,0.04470399916172028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,128,power_law_1.01,0.11681280136108399
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,8,power_law_1.01,0.06771199703216553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,32,power_law_1.2,0.04784640073776245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,4,power_law_1.2,1.278553581237793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,8,power_law_1.01,0.0747648000717163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,32,power_law_1.2,0.049472001194953916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,128,power_law_1.01,0.12067199945449829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,32,power_law_1.2,0.05406079888343811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,8,power_law_1.01,0.08720639944076539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,128,power_law_1.01,0.11864320039749146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,4,power_law_1.2,1.595244789123535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,32,power_law_1.2,0.06387199759483338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,8,power_law_1.01,0.08723199963569642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,32,power_law_1.2,0.07768959999084472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,128,power_law_1.01,0.122707200050354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,8,power_law_1.01,0.08932480216026306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,32,power_law_1.2,0.10343680381774903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,8,power_law_1.01,0.0911679983139038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,4,power_law_1.2,2.5700351715087892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,128,power_law_1.01,0.1345023989677429
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,32,power_law_1.2,0.12588800191879274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,8,power_law_1.01,0.09013760089874268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,32,power_law_1.2,0.16327040195465087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,128,power_law_1.01,0.1639232039451599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,8,power_law_1.01,0.09411200284957885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,32,power_law_1.2,0.223142409324646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,8,power_law_1.01,0.09600639939308167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,128,power_law_1.01,0.1942720055580139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,32,power_law_1.2,0.3297343969345093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,8,power_law_1.01,0.09983360171318054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,4,power_law_1.2,5.098015975952149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,128,power_law_1.01,0.2301248073577881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,32,power_law_1.2,0.5004608154296875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,8,power_law_1.01,0.10619519948959351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,128,power_law_1.01,0.27797760963439944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,8,power_law_1.01,0.10890239477157593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,32,power_law_1.2,0.5968704223632812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,8,power_law_1.01,0.11955840587615967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,128,power_law_1.01,0.36177279949188235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,32,power_law_1.2,0.8625280380249023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,8,power_law_1.01,0.13395839929580688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,8,power_law_1.01,0.14851839542388917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,128,power_law_1.01,0.4780288219451904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,32,power_law_1.2,1.555519962310791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,8,power_law_1.01,0.17383040189743043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,128,power_law_1.01,0.658073616027832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,8,power_law_1.01,0.20362238883972167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,8,power_law_1.01,0.2564800024032593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,128,power_law_1.01,0.8186880111694336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,32,power_law_1.2,2.5754112243652343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,8,power_law_1.01,0.2733504056930542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,8,power_law_1.01,0.3925631999969482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,128,power_law_1.01,1.0512960433959961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,8,power_law_1.01,0.46651520729064944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,8,power_law_1.01,0.05326719880104065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,8,power_law_1.01,0.059436798095703125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,8,power_law_1.01,0.709772777557373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,128,power_law_1.01,1.6170368194580078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,8,power_law_1.01,0.04757120013237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,8,power_law_1.01,0.046598398685455324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,8,power_law_1.01,0.05909119844436646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,8,power_law_1.01,0.8907711982727051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,8,power_law_1.01,0.07226240038871765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,8,power_law_1.01,0.07169920206069946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,8,power_law_1.01,0.07240960001945496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,8,power_law_1.01,1.0164031982421875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,8,power_law_1.01,0.07683839797973632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,128,power_law_1.01,2.942348861694336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,8,power_law_1.01,0.07397119998931885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,8,power_law_1.01,0.08007680177688599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,8,power_law_1.01,0.08061439990997314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,8,power_law_1.01,1.7112319946289063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,8,power_law_1.01,0.08303999900817871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,8,power_law_1.01,0.08670079708099365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,8,power_law_1.01,0.09055359959602356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,8,power_law_1.01,0.09384959936141968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,8,power_law_1.01,0.09848960041999817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,8,power_law_1.01,0.11255680322647095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,1,power_law_1.2,0.05395200252532959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,8,power_law_1.01,3.2933376312255858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,8,power_law_1.01,0.12547199726104735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,8,power_law_1.01,0.1477311968803406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,8,power_law_1.01,0.16373759508132935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,1,power_law_1.2,0.05847039818763733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,8,power_law_1.01,0.20830719470977782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,8,power_law_1.01,0.2459968090057373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,8,power_law_1.01,0.35715839862823484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,8,power_law_1.01,0.44250879287719724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,1,power_law_1.2,0.06861439943313599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,8,power_law_1.01,0.6051712036132812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,16,1,power_law_1.2,0.035417601466178894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,8,power_law_1.01,0.7711679935455322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,16,1,power_law_1.2,0.040345600247383116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,16,1,power_law_1.2,0.03863039910793305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,16,1,power_law_1.2,0.05434880256652832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,1,power_law_1.2,0.08428159952163697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,16,1,power_law_1.2,0.05549439787864685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,16,1,power_law_1.2,0.05623040199279785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,16,1,power_law_1.2,0.05609599947929382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,8,power_law_1.01,1.0987648010253905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,16,1,power_law_1.2,0.05737599730491638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,16,1,power_law_1.2,0.057017600536346434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,16,1,power_law_1.2,0.05860480070114136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,16,1,power_law_1.2,0.058713597059249875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,16,1,power_law_1.2,0.06291840076446534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,16,1,power_law_1.2,0.0654528021812439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,16,1,power_law_1.2,0.06659200191497802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,16,1,power_law_1.2,0.07279360294342041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,16,1,power_law_1.2,0.07654399871826172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,16,1,power_law_1.2,0.07975040078163147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,8,power_law_1.01,1.4385472297668458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,1,power_law_1.2,0.1097983956336975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,16,1,power_law_1.2,0.09049599766731262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,16,1,power_law_1.2,0.09416959881782531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,16,1,power_law_1.2,0.11643520593643189
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,16,1,power_law_1.2,0.1421504020690918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,16,1,power_law_1.2,0.17497600317001344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,16,1,power_law_1.2,0.21487998962402344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,16,1,power_law_1.2,0.29754879474639895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,16,1,power_law_1.2,0.35269761085510254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,2,2,power_law_1.2,0.03173120021820068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,16,1,power_law_1.2,0.49582719802856445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,1,power_law_1.2,0.13708159923553467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,16,1,power_law_1.2,0.6390783786773682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,2,2,power_law_1.2,0.03273600041866302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,8,power_law_1.01,2.9211519241333006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,16,1,power_law_1.2,0.7731135845184326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,2,2,power_law_1.2,0.03705599904060364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,16,1,power_law_1.2,1.1844351768493653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,2,2,power_law_1.2,0.03845120072364807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,2,2,power_law_1.2,0.046937599778175354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,1,power_law_1.2,0.1575808048248291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,16,1,power_law_1.2,2.293913650512695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,2,2,power_law_1.2,0.06876159906387329
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,2,2,power_law_1.2,0.07275519967079162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,2,2,power_law_1.2,0.07390720248222352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,2,2,power_law_1.2,0.07780479788780212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,1,power_law_1.2,0.1986240029335022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,2,2,power_law_1.2,0.07608960270881653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,2,2,power_law_1.2,0.07996799945831298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,2,2,power_law_1.2,0.08254719972610473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,1,power_law_1.2,0.20864639282226563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,2,2,power_law_1.2,0.08361600041389465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,2,2,power_law_1.2,0.08287360072135926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,2,2,power_law_1.2,0.09241600036621093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,2,2,power_law_1.2,0.0933184027671814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,1,power_law_1.2,0.2149183988571167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,2,2,power_law_1.2,0.10128639936447144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,2,2,power_law_1.2,0.11091840267181396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,2,power_law_1.2,0.11546239852905274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,1,power_law_1.2,0.23053441047668458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,2,power_law_1.2,0.1456704020500183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,2,power_law_1.2,0.15453439950942993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,2,power_law_1.2,0.18873599767684937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,1,power_law_1.2,0.2439807891845703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,2,power_law_1.2,0.2348031997680664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,2,power_law_1.2,0.32896640300750735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,1,power_law_1.2,0.25919361114501954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,2,power_law_1.2,0.4121984004974365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,2,power_law_1.2,0.5496831893920898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,1,power_law_1.2,0.27587199211120605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,16,4,power_law_1.01,0.03230080008506775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,16,4,power_law_1.01,0.03243519961833954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,16,4,power_law_1.01,0.03111039996147156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,2,power_law_1.2,0.7047616004943847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,16,4,power_law_1.01,0.030995199084281923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,16,4,power_law_1.01,0.03203200101852417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,1,power_law_1.2,0.08863999843597412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,16,4,power_law_1.01,0.0331712007522583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,16,4,power_law_1.01,0.03408640027046204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,16,4,power_law_1.01,0.034431999921798705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,16,4,power_law_1.01,0.03492479920387268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,1,power_law_1.2,0.29807999134063723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,16,4,power_law_1.01,0.0377344012260437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,2,power_law_1.2,0.9361536026000976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,16,4,power_law_1.01,0.03624959886074066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,16,4,power_law_1.01,0.037785598635673524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,16,4,power_law_1.01,0.040031999349594116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,16,4,power_law_1.01,0.04264959990978241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,1,power_law_1.2,0.09961599707603455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,16,4,power_law_1.01,0.04965760111808777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,16,4,power_law_1.01,0.05309439897537231
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,16,4,power_law_1.01,0.05600000023841858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,16,4,power_law_1.01,0.07215999960899352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,16,4,power_law_1.01,0.08484479784965515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,1,power_law_1.2,0.3173311948776245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,16,4,power_law_1.01,0.10351359844207764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,2,power_law_1.2,1.4679743766784668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,16,4,power_law_1.01,0.12486399412155151
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,16,4,power_law_1.01,0.17098239660263062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,1,power_law_1.2,0.14479360580444336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,16,4,power_law_1.01,0.21407999992370605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,16,4,power_law_1.01,0.3072704076766968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,16,4,power_law_1.01,0.4377151966094971
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,1,power_law_1.2,0.3155839920043945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,16,4,power_law_1.01,0.5300352096557617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,16,4,power_law_1.01,0.7606336116790772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,1,power_law_1.2,0.22293760776519775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,16,4,power_law_1.01,0.9963199615478515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,2,power_law_1.2,3.0512704849243164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,1,power_law_1.2,0.37304959297180174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,16,4,power_law_1.01,1.525996780395508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,1,power_law_1.2,0.31931519508361816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,16,4,power_law_1.01,2.926438331604004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,1,power_law_1.2,0.42259840965270995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,1,power_law_1.2,0.5840767860412598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,1,power_law_1.2,0.5212096214294434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,1,power_law_1.2,0.6608831882476807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,1,power_law_1.2,0.6001535892486572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,1,power_law_1.2,0.6908160209655761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,1,power_law_1.2,0.7912191867828369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,1,power_law_1.2,0.7226240158081054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,1,power_law_1.2,0.9773119926452637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,1,power_law_1.2,0.738431978225708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,1,power_law_1.2,0.7851583957672119
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,1,power_law_1.2,1.3162367820739747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,1,power_law_1.2,0.8048831939697265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,256,balanced,0.06225066880385081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,256,balanced,0.06420266628265381
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,256,balanced,0.06411199768384297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,1,power_law_1.2,1.6269887924194335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,256,balanced,0.06235733131567637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,256,balanced,0.06419200201829274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,256,balanced,0.06474666794141133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,256,balanced,0.06383466720581055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,256,balanced,0.06614933411280315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,256,balanced,0.0682773341735204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,256,balanced,0.0689386675755183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,256,balanced,0.06870933373769124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,256,balanced,0.07051200171311696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,256,balanced,0.06911466519037883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,256,balanced,0.06895466645558675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,256,balanced,0.07541333138942719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,256,balanced,0.07663466533025105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,256,balanced,0.07749333480993907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,256,balanced,0.0860533316930135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,1,power_law_1.2,0.8305791854858399
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,256,balanced,0.0897706647713979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,256,balanced,0.10716799894968669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,256,balanced,0.1318826675415039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,16,power_law_1.01,0.12008320093154908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,256,balanced,0.16697067022323608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,256,balanced,0.20030399163564047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,256,balanced,0.26522133747736615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,256,balanced,0.32676267623901367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,256,balanced,0.4598986705144246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,256,balanced,0.5885013341903687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,256,balanced,0.7226666609446207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,256,balanced,1.1035839716593425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,1,power_law_1.2,2.305766487121582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,1,power_law_1.2,0.8446847915649414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,256,balanced,2.174917380015055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,16,power_law_1.01,0.15059839487075805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,1,power_law_1.2,0.8863936424255371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,16,power_law_1.01,0.07116159796714783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,1,power_law_1.2,2.9709823608398436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,16,power_law_1.01,0.08856319785118102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,1,power_law_1.2,0.904422378540039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,16,power_law_1.01,0.10455039739608765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,1,power_law_1.2,3.6527359008789064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,1,power_law_1.2,0.9298815727233887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,16,power_law_1.01,0.1352255940437317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,1,power_law_1.2,1.0400704383850097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,16,power_law_1.01,0.14046080112457277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,1,power_law_1.2,1.1249792098999023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,1,power_law_1.2,5.839628982543945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,16,power_law_1.01,0.1358016014099121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,1,power_law_1.2,1.3034751892089844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,8,power_law_1.2,0.03478400111198425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,8,power_law_1.2,0.02781440019607544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,16,power_law_1.01,0.14224640130996705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,8,power_law_1.2,0.0293503999710083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,8,power_law_1.2,0.029811200499534608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,1,power_law_1.2,1.3843903541564941
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,8,power_law_1.2,0.032671999931335446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,16,power_law_1.01,0.14238719940185546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,8,power_law_1.2,0.03676159977912903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,8,power_law_1.2,0.035718399286270144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,8,power_law_1.2,0.03720960021018982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,1,power_law_1.2,11.043558502197266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,16,power_law_1.01,0.1514240026473999
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,1,power_law_1.2,1.717433547973633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,8,power_law_1.2,0.03749760091304779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,8,power_law_1.2,0.03790079951286316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,8,power_law_1.2,0.04137600064277649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,16,power_law_1.01,0.14858239889144897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,8,power_law_1.2,0.043270400166511534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,1,power_law_1.2,1.8950719833374023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,8,power_law_1.2,0.04535039961338043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,8,power_law_1.2,0.050425601005554196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,16,power_law_1.01,0.15134719610214234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,8,power_law_1.2,0.05489919781684875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,8,power_law_1.2,0.06135680079460144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,8,power_law_1.2,0.0565887987613678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,16,power_law_1.01,0.15975680351257324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,1,power_law_1.2,2.4431232452392577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,8,power_law_1.2,0.06534399986267089
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,8,power_law_1.2,0.07247359752655029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,8,power_law_1.2,0.0822655975818634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,16,power_law_1.01,0.16707199811935425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,8,power_law_1.2,0.09483519792556763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,8,power_law_1.2,0.12110079526901245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,1,power_law_1.2,3.1344768524169924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,16,power_law_1.01,0.17212159633636476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,8,power_law_1.2,0.16561919450759888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,8,power_law_1.2,0.23211519718170165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,8,power_law_1.2,0.28331520557403567
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,16,power_law_1.01,0.175654399394989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,8,power_law_1.2,0.39937920570373536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,1,power_law_1.2,4.449395370483399
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,64,balanced,0.06038400034109751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,64,balanced,0.0602453351020813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,16,power_law_1.01,0.20157439708709718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,64,balanced,0.05978133281071981
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,8,power_law_1.2,0.6059455871582031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,64,balanced,0.06051200131575266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,64,balanced,0.06329066554705302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,64,balanced,0.0717493345340093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,64,balanced,0.07073600093523662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,64,balanced,0.0703359991312027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,64,balanced,0.07300266623497009
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,64,balanced,0.07044800122578938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,64,balanced,0.07228800157705943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,64,balanced,0.07137066622575124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,64,balanced,0.07302399973074596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,64,balanced,0.0714026689529419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,8,power_law_1.2,0.6344704151153564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,64,balanced,0.07779199878374736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,64,balanced,0.07994666695594788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,64,balanced,0.08060266574223836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,16,power_law_1.01,0.22047359943389894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,64,balanced,0.08452266454696655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,64,balanced,0.09179199735323589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,64,balanced,0.10666666428248088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,64,balanced,0.12079466382662456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,64,balanced,0.14958932995796204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,64,balanced,0.1709280014038086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,64,balanced,0.21918400128682455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,64,balanced,0.2675039966901143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,64,balanced,0.37240533034006756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,64,balanced,0.4692213137944539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,8,power_law_1.2,0.9896960258483887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,64,balanced,0.5774293343226115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,64,balanced,0.8730506896972656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,16,power_law_1.01,0.2596415996551514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,1,power_law_1.2,5.311328125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,64,balanced,1.700533390045166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,16,power_law_1.01,0.3043008089065552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,8,power_law_1.2,2.7701183319091798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,16,power_law_1.01,0.37708799839019774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,2,1,power_law_1.2,6.4303741455078125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,16,power_law_1.01,0.38594560623168944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,16,power_law_1.01,0.4910143852233887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,16,power_law_1.01,0.6376832008361817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,2,1,power_law_1.2,9.744153594970703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,1,power_law_1.2,0.035743999481201175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,16,power_law_1.01,0.8785728454589844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,2,power_law_1.01,0.045100799202919005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,1,power_law_1.2,0.048895999789237976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,2,power_law_1.01,0.04666880071163178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,1,power_law_1.2,0.057817602157592775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,2,power_law_1.01,0.0514303982257843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,1,power_law_1.2,0.0775871992111206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,2,power_law_1.01,0.058303999900817874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,16,power_law_1.01,1.0146047592163085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,2,power_law_1.01,0.06871039867401123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,1,power_law_1.2,0.09553920030593872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,2,power_law_1.01,0.08991360068321227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,1,power_law_1.2,0.16412160396575928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,2,power_law_1.01,0.09326720237731934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,2,power_law_1.01,0.09822720289230347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,1,power_law_1.2,0.1849727988243103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,2,power_law_1.01,0.10038399696350098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,1,power_law_1.2,0.19026559591293335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,1,16,power_law_1.01,1.2164095878601073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,2,power_law_1.01,0.10414079427719117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,2,power_law_1.01,0.10721280574798583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,1,power_law_1.2,0.19733760356903077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,2,1,power_law_1.2,18.70942077636719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,2,power_law_1.01,0.11103999614715576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,1,power_law_1.2,0.19455360174179076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,2,power_law_1.01,0.1139456033706665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,1,power_law_1.2,0.20618879795074463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,2,power_law_1.01,0.12117760181427002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,2,power_law_1.01,0.12942080497741698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,1,power_law_1.2,0.21101438999176025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,2,power_law_1.01,0.13626240491867064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,1,power_law_1.2,0.21651840209960938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,2,power_law_1.01,0.14461439847946167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,1,16,power_law_1.01,1.8143808364868164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,1,power_law_1.2,0.227020788192749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,2,power_law_1.01,0.17121280431747438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,2,power_law_1.01,0.19351680278778077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,1,power_law_1.2,0.23158400058746337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,2,power_law_1.01,0.24632959365844725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,1,power_law_1.2,0.24281599521636962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,1,power_law_1.2,0.24744319915771484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,2,power_law_1.01,0.31274240016937255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,2,power_law_1.01,0.4057472229003906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,1,power_law_1.2,0.2798336029052734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,2,power_law_1.01,0.49125118255615235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,1,power_law_1.2,0.27664000988006593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,2,power_law_1.01,0.7221375942230225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,1,power_law_1.2,0.3191551923751831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,1,power_law_1.2,0.35516159534454345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,2,power_law_1.01,0.9257920265197754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,1,power_law_1.2,0.42108798027038574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,1,16,power_law_1.01,3.8397377014160154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,2,power_law_1.01,1.3481280326843261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,1,power_law_1.2,0.4620800018310547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,1,power_law_1.2,0.07359359860420227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,2,power_law_1.01,1.6048128128051757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,1,power_law_1.2,0.586303997039795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,1,power_law_1.2,0.0775871992111206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,1,power_law_1.2,0.7161600112915039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,1,power_law_1.2,0.09018880128860474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,2,power_law_1.01,1.9960384368896484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,1,power_law_1.2,0.9686592102050782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,1,power_law_1.2,0.11418240070343018
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,1,power_law_1.2,0.14261759519577027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,1,power_law_1.2,1.2010111808776855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,2,power_law_1.01,3.0217151641845703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,1,power_law_1.2,0.170905601978302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,1,1,power_law_1.2,1.4455679893493651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,1,power_law_1.2,0.23077759742736817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,1,power_law_1.2,0.24598400592803954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,1,1,power_law_1.2,2.182534408569336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,1,power_law_1.2,0.25189120769500734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,2,power_law_1.01,5.82606086730957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,1,power_law_1.2,0.2578752040863037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,1,power_law_1.2,0.27107839584350585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,32,4,power_law_1.01,0.03993600010871887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,32,4,power_law_1.01,0.036447998881340024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,1,1,power_law_1.2,4.154297637939453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,1,power_law_1.2,0.2828160047531128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,32,4,power_law_1.01,0.03511039912700653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,32,4,power_law_1.01,0.035046398639678955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,32,4,power_law_1.01,0.03642880022525787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,32,4,power_law_1.01,0.0365119993686676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,32,4,power_law_1.01,0.037273600697517395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,1,power_law_1.2,0.28618879318237306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,32,4,power_law_1.01,0.036934399604797365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,32,4,power_law_1.01,0.03738240003585815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,32,4,power_law_1.01,0.03798399865627289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,32,4,power_law_1.01,0.03814400136470795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,32,4,power_law_1.01,0.03935360014438629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,1,power_law_1.2,0.2999039888381958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,32,4,power_law_1.01,0.040064001083374025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,32,4,power_law_1.01,0.04176000058650971
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,32,4,power_law_1.01,0.046342399716377256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,32,4,power_law_1.01,0.0486272007226944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,1,power_law_1.2,0.3134912014007568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,32,4,power_law_1.01,0.05047680139541626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,32,4,power_law_1.01,0.05734400153160095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,32,4,power_law_1.01,0.06166399717330932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,32,4,power_law_1.01,0.07190399765968322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,1,power_law_1.2,0.32058238983154297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,32,4,power_law_1.01,0.10310399532318115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,32,4,power_law_1.01,0.141593599319458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,32,4,power_law_1.01,0.14384000301361083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,32,4,power_law_1.01,0.23320960998535156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,1,power_law_1.2,0.3454655885696411
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,32,4,power_law_1.01,0.23685119152069092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,32,4,power_law_1.01,0.35520000457763673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,32,4,power_law_1.01,0.4607935905456543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,1,power_law_1.2,0.38895359039306643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,32,4,power_law_1.01,0.6230144023895263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,32,4,power_law_1.01,0.9925632476806641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,1,power_law_1.2,0.42920961380004885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,32,4,power_law_1.01,1.7871679306030273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,1,power_law_1.2,0.5238592147827148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,16,balanced,0.04854399959246317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,16,balanced,0.049285332361857094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,16,balanced,0.0498933345079422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,16,balanced,0.05198933184146881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,1,power_law_1.2,0.5965888023376464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,16,balanced,0.05393599967161814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,16,balanced,0.05983466903368632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,16,balanced,0.062122667829195656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,16,balanced,0.06211733321348826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,16,balanced,0.06250666578610738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,16,balanced,0.06425599753856659
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,16,balanced,0.06458133459091187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,16,balanced,0.06405866642793019
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,16,balanced,0.06650133430957794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,16,balanced,0.06610666712125142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,16,balanced,0.07021866738796234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,16,balanced,0.07237866520881653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,16,balanced,0.07647466659545898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,16,balanced,0.08273066580295563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,16,balanced,0.08655466636021932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,16,balanced,0.10314133763313293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,16,balanced,0.118559996287028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,16,balanced,0.1498186687628428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,1,power_law_1.2,0.7704512119293213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,16,balanced,0.17508800824483237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,16,balanced,0.23138133684794107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,16,balanced,0.28313066562016803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,16,balanced,0.3922239939371745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,16,balanced,0.4973119894663493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,16,balanced,0.6105546553929647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,2,2,balanced,0.06284266710281372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,16,balanced,0.9315573374430338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,2,2,balanced,0.11877866586049397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,1,power_law_1.2,0.931334400177002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,2,2,balanced,0.1376213332017263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,2,2,balanced,0.13756266236305237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,2,2,balanced,0.1385706663131714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,16,balanced,1.8244959513346355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,2,2,balanced,0.1386186679204305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,2,2,balanced,0.13819733262062073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,2,2,balanced,0.14014400045077005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,2,2,balanced,0.1390720009803772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,2,2,balanced,0.13903466860453287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,2,2,balanced,0.14379200339317322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,2,2,balanced,0.141893337170283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,2,2,balanced,0.14292800426483154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,2,2,balanced,0.1470133364200592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,2,2,balanced,0.15742933750152588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,2,2,balanced,0.1639306644598643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,1,power_law_1.2,1.2659839630126952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,2,2,balanced,0.16673600673675537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,2,2,balanced,0.1914400060971578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,2,2,balanced,0.20002132654190063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,2,2,balanced,0.2408533294995626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,2,2,balanced,0.2738986611366272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,2,2,balanced,0.3603359858194987
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,2,2,balanced,0.44180798530578613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,2,2,balanced,0.6252959966659546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,2,2,balanced,0.7841333548227946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,2,2,balanced,1.166266679763794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,1,power_law_1.2,1.6509824752807618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,2,2,balanced,1.5201706886291504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,2,2,balanced,1.8603787422180176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,2,2,balanced,2.9646132787068686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,1,power_law_1.2,2.36231689453125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,2,2,balanced,5.861082712809245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,1,power_law_1.2,2.9374208450317383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,4,1,balanced,0.05231999854246775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,4,1,balanced,0.050053333242734276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,4,1,balanced,0.05595199763774872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,4,1,balanced,0.06451199948787689
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,4,1,balanced,0.0870293378829956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,4,1,balanced,0.08664533495903015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,4,1,balanced,0.08830400307973225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,4,1,balanced,0.0890880028406779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,4,1,balanced,0.08935999870300293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,4,1,balanced,0.09001599748929341
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,4,1,balanced,0.09291199843088786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,4,1,balanced,0.09698133667310078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,1,power_law_1.2,3.5947711944580076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,4,1,balanced,0.09876799583435059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,4,1,balanced,0.10309333602587382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,4,1,balanced,0.1095199982325236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,4,1,balanced,0.11522133151690166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,4,1,balanced,0.12483200430870056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,4,1,balanced,0.15025066335995993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,4,1,balanced,0.16699200868606567
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,4,1,balanced,0.21050665775934854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,4,1,balanced,0.24841066201527914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,4,1,balanced,0.35789867242177326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,4,1,balanced,0.44758931795756024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,4,1,balanced,0.6498773495356241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,4,1,balanced,0.8528853257497152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,4,1,balanced,1.26638396581014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,4,1,balanced,1.6570879618326824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,4,1,balanced,2.0553919474283853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,4,1,balanced,3.2527147928873696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,1,power_law_1.2,5.674668884277343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,16,power_law_1.01,0.1032256007194519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,4,1,balanced,6.325626373291016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,16,power_law_1.01,0.10106240510940552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,16,power_law_1.01,0.06465920209884643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,16,power_law_1.01,0.07414399981498718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,16,power_law_1.01,0.0847487986087799
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,16,power_law_1.01,0.10817279815673828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,16,power_law_1.01,0.10481280088424683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,1,power_law_1.2,10.777081298828126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,16,power_law_1.01,0.10488320589065551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,16,power_law_1.01,0.10714240074157715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,16,power_law_1.01,0.10995839834213257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,1,power_law_1.2,0.15587199926376344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,16,power_law_1.01,0.11088000535964966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,16,power_law_1.01,0.11322239637374878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,1,power_law_1.2,0.16781439781188964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,16,power_law_1.01,0.11701760292053223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,16,power_law_1.01,0.11548800468444824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,1,power_law_1.2,0.33742079734802244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,16,power_law_1.01,0.1241536021232605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,16,power_law_1.01,0.1290112018585205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,1,power_law_1.2,0.5493375778198242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,16,power_law_1.01,0.13911039829254152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,16,power_law_1.01,0.16138240098953247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,1,power_law_1.2,0.7660672187805175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,16,power_law_1.01,0.17112959623336793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,16,power_law_1.01,0.199891197681427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,1,power_law_1.2,1.5568960189819336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,16,power_law_1.01,0.2240895986557007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,1,power_law_1.01,0.14538880586624145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,16,power_law_1.01,0.27473280429840086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,1,power_law_1.2,1.725062370300293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,16,power_law_1.01,0.36754560470581055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,1,power_law_1.01,0.24261760711669922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,1,power_law_1.2,1.7932159423828125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,16,power_law_1.01,0.4275263786315918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,4,8,power_law_1.01,0.06533759832382202
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,4,8,power_law_1.01,0.05644800066947937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,1,power_law_1.01,0.2906559944152832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,4,8,power_law_1.01,0.053478401899337766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,4,8,power_law_1.01,0.05438079833984375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,4,8,power_law_1.01,0.05363199710845947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,16,power_law_1.01,0.46851201057434083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,1,power_law_1.2,1.8154304504394532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,4,8,power_law_1.01,0.05432959794998169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,4,8,power_law_1.01,0.058182400465011594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,4,8,power_law_1.01,0.06032000184059143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,1,power_law_1.01,0.6202688217163086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,4,8,power_law_1.01,0.06247040033340454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,4,8,power_law_1.01,0.0633791983127594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,16,power_law_1.01,0.6932799816131592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,4,8,power_law_1.01,0.06289280056953431
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,4,8,power_law_1.01,0.06607999801635742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,1,power_law_1.2,1.8755071640014649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,4,8,power_law_1.01,0.06572160124778748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,4,8,power_law_1.01,0.07239680290222168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,1,power_law_1.01,1.0006272315979003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,4,8,power_law_1.01,0.07918720245361328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,4,8,power_law_1.01,0.08087679743766785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,16,power_law_1.01,0.7718912124633789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,4,8,power_law_1.01,0.10375679731369018
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,4,8,power_law_1.01,0.12250239849090576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,4,8,power_law_1.01,0.1380735993385315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,1,power_law_1.2,2.0659967422485352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,4,8,power_law_1.01,0.17828480005264283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,1,power_law_1.01,1.437382411956787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,4,8,power_law_1.01,0.20987520217895508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,4,8,power_law_1.01,0.2779328107833862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,16,power_law_1.01,0.9671296119689942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,4,8,power_law_1.01,0.38766720294952395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,4,8,power_law_1.01,0.5869696140289307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,1,power_law_1.2,2.0416704177856446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,1,power_law_1.01,2.115987205505371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,4,8,power_law_1.01,0.7577536106109619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,4,8,power_law_1.01,1.1101183891296387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,4,8,power_law_1.01,1.4333632469177247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,16,power_law_1.01,1.5853055953979491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,1,power_law_1.2,2.0930944442749024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,1,power_law_1.01,2.1766336441040037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,4,8,power_law_1.01,1.8351743698120118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,1,power_law_1.2,2.143564796447754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,1,power_law_1.01,2.292460823059082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,4,8,power_law_1.01,2.3808128356933596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,1,power_law_1.2,2.212544059753418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,1,power_law_1.01,2.3924928665161134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,16,power_law_1.01,2.8750463485717774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,4,8,power_law_1.01,4.597350311279297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,1,power_law_1.2,2.2407743453979494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,1,power_law_1.01,2.4190271377563475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,1,power_law_1.2,2.3154624938964843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,1,power_law_1.01,2.6132287979125977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,1,power_law_1.2,2.5039615631103516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,1,power_law_1.01,2.630201530456543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,1,power_law_1.2,2.6805440902709963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,1,power_law_1.01,2.6771711349487304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,1,power_law_1.2,3.1374528884887694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,1,power_law_1.01,2.736979293823242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,1,power_law_1.2,3.2421375274658204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,1,power_law_1.01,2.8240192413330076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,1,power_law_1.2,3.861721420288086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,1,power_law_1.01,3.011577606201172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,1,power_law_1.01,3.183014488220215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,1,power_law_1.2,3.911366271972656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,32,8,power_law_1.2,0.05278080105781555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,2,power_law_1.01,0.063155198097229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,1,power_law_1.01,3.2198848724365234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,1,power_law_1.2,4.701222229003906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,32,8,power_law_1.2,0.05898879766464234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,2,power_law_1.01,0.06768640279769897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,32,8,power_law_1.2,0.05124480128288269
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,1,power_law_1.01,3.578540802001953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,2,power_law_1.01,0.07112320065498352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,32,8,power_law_1.2,0.05248000025749207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,2,power_law_1.01,0.08088319897651672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,1,power_law_1.2,6.1982463836669925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,1,power_law_1.01,3.6313407897949217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,32,8,power_law_1.2,0.057017600536346434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,2,power_law_1.01,0.09640960097312927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,32,8,power_law_1.2,0.060089600086212155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,2,power_law_1.01,0.11288319826126099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,1,power_law_1.01,4.006137466430664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,32,8,power_law_1.2,0.06646400094032287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,2,power_law_1.01,0.13506560325622557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,1,power_law_1.2,8.070387268066407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,32,8,power_law_1.2,0.06775040030479432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,2,power_law_1.01,0.1452415943145752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,1,power_law_1.01,4.326508712768555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,32,8,power_law_1.2,0.06828799843788147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,2,power_law_1.01,0.14613759517669678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,32,8,power_law_1.2,0.06887680292129517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,2,power_law_1.01,0.14677120447158815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,32,8,power_law_1.2,0.07066239714622498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,1,power_law_1.2,9.713651275634765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,1,power_law_1.01,5.056172943115234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,2,power_law_1.01,0.15178240537643434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,32,8,power_law_1.2,0.072326397895813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,2,power_law_1.01,0.15713920593261718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,32,8,power_law_1.2,0.07359359860420227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,2,power_law_1.01,0.15660159587860106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,1,power_law_1.01,5.894073486328125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,32,8,power_law_1.2,0.07820159792900086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,2,power_law_1.01,0.16395519971847533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,32,8,power_law_1.2,0.085452800989151
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,2,power_law_1.01,0.17211519479751586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,1,power_law_1.2,11.86138916015625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,32,8,power_law_1.2,0.08609279990196228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,2,power_law_1.01,0.17818880081176758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,32,8,power_law_1.2,0.09912959933280945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,1,power_law_1.01,7.530585479736328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,2,power_law_1.01,0.18450560569763183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,32,8,power_law_1.2,0.10869760513305664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,2,power_law_1.01,0.21086080074310304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,32,8,power_law_1.2,0.12981120347976685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,2,power_law_1.01,0.23491840362548827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,32,8,power_law_1.2,0.16506240367889405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,1,power_law_1.01,8.923423767089844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,2,power_law_1.01,0.28618879318237306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,1,power_law_1.2,15.665055847167968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,32,8,power_law_1.2,0.18204159736633302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,2,power_law_1.01,0.33838720321655275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,32,8,power_law_1.2,0.23630719184875487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,2,power_law_1.01,0.4422783851623535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,32,8,power_law_1.2,0.2682687997817993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,2,power_law_1.01,0.560588788986206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,1,power_law_1.01,10.391506958007813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,32,8,power_law_1.2,0.3955712080001831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,2,power_law_1.01,0.7407551765441894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,32,8,power_law_1.2,0.4679872035980225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,2,power_law_1.01,0.92740478515625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,32,8,power_law_1.2,0.7563968181610108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,2,power_law_1.01,1.3522624015808105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,32,8,power_law_1.2,0.9513088226318359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,1,power_law_1.01,15.167936706542969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,2,power_law_1.01,1.746905517578125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,1,power_law_1.2,27.9628662109375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,32,8,power_law_1.2,1.1694527626037599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,2,power_law_1.01,2.3024576187133787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,32,8,power_law_1.2,1.7138048171997071
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,2,power_law_1.01,3.40750732421875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,32,8,power_law_1.2,3.925836944580078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,1,power_law_1.01,27.471066284179688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,2,power_law_1.01,6.625657653808593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,1,balanced,0.03630933413902918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,1,balanced,0.03852800031503042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,1,balanced,0.04669866462548574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,1,balanced,0.07320000231266022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,1,balanced,0.10652800401051839
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,1,balanced,0.1312373379866282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,1,balanced,0.1355839967727661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,1,balanced,0.13411200046539307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,1,balanced,0.1344480017820994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,1,balanced,0.1341759959856669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,1,balanced,0.13353600104649863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,1,balanced,0.13706666231155396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,1,balanced,0.13851199547449747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,1,balanced,0.14012799660364786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,1,balanced,0.14247999588648477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,1,balanced,0.1462399959564209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,1,balanced,0.15121600031852722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,1,balanced,0.166293332974116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,1,balanced,0.1682186722755432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,1,balanced,0.2262186606725057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,1,balanced,0.21917333205540976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,1,balanced,0.2928746740023295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,1,balanced,0.3218666712443034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,1,balanced,0.45531201362609863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,1,balanced,0.5294186671574911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,1,balanced,0.7782453695933024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,1,balanced,0.9387626647949219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,2,1,balanced,1.1623679796854656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,2,1,balanced,1.7873493830362956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,2,1,balanced,3.4491840998331704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,32,8,power_law_1.01,0.035411199927330016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,32,8,power_law_1.01,0.03639039993286133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,32,8,power_law_1.01,0.03441280126571655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,32,8,power_law_1.01,0.03269760012626648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,32,8,power_law_1.01,0.03410559892654419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,32,8,power_law_1.01,0.03410559892654419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,32,8,power_law_1.01,0.03459199965000152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,32,8,power_law_1.01,0.034815999865531924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,32,8,power_law_1.01,0.035436800122261046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,32,8,power_law_1.01,0.035692799091339114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,64,balanced,0.056287998954455055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,64,balanced,0.05595199763774872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,32,8,power_law_1.01,0.03610239923000336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,128,balanced,0.03764266769091288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,64,balanced,0.056202664971351624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,64,balanced,0.056186666091283165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,128,balanced,0.03608000030120214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,32,8,power_law_1.01,0.03989759981632233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,64,balanced,0.05964800218741099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,128,balanced,0.03561066587766012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,64,balanced,0.07039999961853027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,128,balanced,0.037871999045213066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,64,balanced,0.08181333541870117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,32,8,power_law_1.01,0.03802880048751831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,128,balanced,0.03757333258787791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,64,balanced,0.08108266691366832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,128,balanced,0.037621334195137024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,128,balanced,0.037632000943024956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,64,balanced,0.08417066931724548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,128,balanced,0.037615999579429626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,64,balanced,0.08146133522192638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,32,8,power_law_1.01,0.03909119963645935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,128,balanced,0.039546666045983635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,64,balanced,0.08649067083994548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,128,balanced,0.037776000797748566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,64,balanced,0.08664000034332275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,128,balanced,0.03754133234421412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,128,balanced,0.039642666776975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,64,balanced,0.08753599723180135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,128,balanced,0.0395413339138031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,32,8,power_law_1.01,0.0426367998123169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,64,balanced,0.08628267049789429
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,128,balanced,0.03977599988381068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,64,balanced,0.0958026647567749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,128,balanced,0.04196266829967499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,128,balanced,0.04372266431649526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,64,balanced,0.09658666451772054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,32,8,power_law_1.01,0.04349440038204193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,128,balanced,0.04417066772778829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,64,balanced,0.1032426655292511
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,128,balanced,0.045925334095954895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,64,balanced,0.11432000001271565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,128,balanced,0.04740266501903534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,128,balanced,0.0498933345079422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,64,balanced,0.11926933129628499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,32,8,power_law_1.01,0.04526079893112182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,128,balanced,0.05388799806435903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,64,balanced,0.14294399817784628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,128,balanced,0.05829333265622457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,64,balanced,0.16152000427246094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,128,balanced,0.06680533289909363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,32,8,power_law_1.01,0.0530239999294281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,64,balanced,0.20115200678507486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,128,balanced,0.09237866600354512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,128,balanced,0.11700800061225891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,64,balanced,0.24009066820144653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,128,balanced,0.1502079963684082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,32,8,power_law_1.01,0.05817599892616272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,64,balanced,0.31881600618362427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,128,balanced,0.18661334117253622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,1,128,balanced,0.2241013248761495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,64,balanced,0.4116373459498088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,32,8,power_law_1.01,0.06945279836654664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,1,128,balanced,0.33188800017038983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,64,balanced,0.5662506818771362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,1,128,balanced,0.61736532052358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,32,8,power_law_1.01,0.07948160171508789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,64,balanced,0.7225386301676432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,32,8,power_law_1.01,0.10964479446411132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,64,balanced,0.8964746793111166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,32,8,power_law_1.01,0.13422080278396606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,64,balanced,1.3797705968221028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,32,8,power_law_1.01,0.17538559436798096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,32,8,power_law_1.01,0.21825919151306153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,64,balanced,2.6960693995157876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,32,8,power_law_1.01,0.3196608066558838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,32,8,power_law_1.01,0.4068160057067871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,32,8,power_law_1.01,0.48404479026794434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,32,8,power_law_1.01,0.7462783813476562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,32,8,power_law_1.01,1.4553664207458497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,2,power_law_1.2,0.09026560187339783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,2,power_law_1.2,0.12781440019607543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,2,power_law_1.2,0.10158720016479492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,2,power_law_1.2,0.14596480131149292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,2,power_law_1.2,0.1921280026435852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,2,power_law_1.2,0.2736448049545288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,2,power_law_1.2,0.04692479968070984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,2,power_law_1.2,0.05053439736366272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,2,power_law_1.2,0.3751039981842041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,2,2,power_law_1.2,0.10821759700775146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,2,power_law_1.2,0.055795198678970336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,2,2,power_law_1.2,0.1630784034729004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,2,2,power_law_1.2,0.1155519962310791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,2,2,power_law_1.2,0.2174463987350464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,2,power_law_1.2,0.06493440270423889
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,2,2,power_law_1.2,0.2121920108795166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,2,2,power_law_1.2,0.21013119220733642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,2,power_law_1.2,0.39336960315704345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,2,2,power_law_1.2,0.22291200160980223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,2,power_law_1.2,0.0787392020225525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,2,2,power_law_1.2,0.22021760940551757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,2,2,power_law_1.2,0.2342911958694458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,2,2,power_law_1.2,0.23555200099945067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,2,power_law_1.2,0.11084159612655639
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,2,2,power_law_1.2,0.22421119213104249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,2,2,power_law_1.2,0.2535167932510376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,2,power_law_1.2,0.42537598609924315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,2,power_law_1.2,0.11486719846725464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,2,2,power_law_1.2,0.25153279304504395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,128,power_law_1.2,0.06935679912567139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,2,2,power_law_1.2,0.25883519649505615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,2,2,power_law_1.2,0.26654078960418703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,2,power_law_1.2,0.12049920558929443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,2,2,power_law_1.2,0.29041280746459963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,128,power_law_1.2,0.07472000122070313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,2,2,power_law_1.2,0.296729588508606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,2,2,power_law_1.2,0.3484544038772583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,2,power_law_1.2,0.4267136096954346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,2,power_law_1.2,0.12414079904556274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,2,2,power_law_1.2,0.4240255832672119
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,2,2,power_law_1.2,0.5419904232025147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,128,power_law_1.2,0.06929280161857605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,2,power_law_1.2,0.13219200372695922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,2,2,power_law_1.2,0.6742527961730957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,2,2,power_law_1.2,0.9323007583618164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,2,power_law_1.2,0.46216959953308107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,2,power_law_1.2,0.13733760118484498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,128,power_law_1.2,0.06910719871520996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,2,2,power_law_1.2,1.1200384140014648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,2,power_law_1.2,0.14762879610061647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,2,2,power_law_1.2,1.8069120407104493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,128,power_law_1.2,0.07212160229682922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,2,power_law_1.2,0.4787327766418457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,2,power_law_1.2,0.15031039714813232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,2,2,power_law_1.2,2.207935905456543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,2,power_law_1.2,0.14748799800872803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,128,power_law_1.2,0.07333120107650756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,2,2,power_law_1.2,3.0139839172363283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,2,power_law_1.2,0.46961278915405275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,2,power_law_1.2,0.15838079452514647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,128,power_law_1.2,0.07440000176429748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,2,power_law_1.2,0.1622655987739563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,2,2,power_law_1.2,4.210764694213867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,2,power_law_1.2,0.4979263782501221
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,128,power_law_1.2,0.07333760261535645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,2,power_law_1.2,0.17630079984664918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,2,power_law_1.2,0.20965120792388917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,128,power_law_1.2,0.07445759773254394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,2,2,power_law_1.2,5.295033645629883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,2,power_law_1.2,0.5339263916015625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,2,power_law_1.2,0.23269119262695312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,128,power_law_1.2,0.07289599776268005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,2,power_law_1.2,0.2817919969558716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,2,power_law_1.2,0.5191936016082763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,128,power_law_1.2,0.07336320281028748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,2,power_law_1.2,0.33177599906921384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,2,2,power_law_1.2,9.17356185913086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,128,power_law_1.2,0.07281919717788696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,2,power_law_1.2,0.4494527816772461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,2,power_law_1.2,0.5557568073272705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,128,power_law_1.2,0.07560960054397584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,2,power_law_1.2,0.5635968208312988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,128,power_law_1.2,0.0743552029132843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,2,power_law_1.2,0.5991104125976563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,2,power_law_1.2,0.8013952255249024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,128,power_law_1.2,0.07966079711914062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,2,power_law_1.2,1.0284223556518555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,2,power_law_1.2,0.6525375843048096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,128,power_law_1.2,0.08021119832992554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,2,2,power_law_1.2,15.383424377441406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,2,power_law_1.2,1.4200575828552247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,128,power_law_1.2,0.09134079813957215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,2,power_law_1.2,0.735756778717041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,128,power_law_1.2,0.09688959717750549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,2,power_law_1.2,1.848307228088379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,2,power_law_1.2,0.7917056083679199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,128,power_law_1.2,0.10543359518051147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,128,power_law_1.2,0.12299519777297974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,2,power_law_1.2,2.1942848205566405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,2,power_law_1.2,0.9443327903747558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,128,power_law_1.2,0.13894399404525756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,2,power_law_1.2,1.0731519699096679
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,128,power_law_1.2,0.18087040185928344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,2,power_law_1.2,3.4916671752929687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,128,power_law_1.2,0.21676158905029297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,2,power_law_1.2,1.4043840408325194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,128,power_law_1.2,0.31351680755615235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,2,power_law_1.2,1.744607925415039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,128,power_law_1.2,0.42695040702819825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,2,power_law_1.2,7.3478271484375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,128,power_law_1.2,0.6052735805511474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,2,power_law_1.2,2.327859115600586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,128,power_law_1.2,0.8088255882263183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,2,power_law_1.2,3.030303955078125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,128,power_law_1.2,1.0586560249328614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,1,power_law_1.2,0.11323519945144653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,1,power_law_1.2,0.14451199769973755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,128,power_law_1.2,1.5416831970214844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,2,power_law_1.2,3.57825927734375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,1,power_law_1.2,0.16986240148544313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,1,power_law_1.2,0.2497472047805786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,2,power_law_1.2,5.321971130371094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,128,power_law_1.2,3.671161651611328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,1,power_law_1.2,0.38908801078796384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,1,power_law_1.2,0.5161087989807129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,8,4,power_law_1.2,0.0374208003282547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,8,4,power_law_1.2,0.03573119938373566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,8,4,power_law_1.2,0.03407999873161316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,8,4,power_law_1.2,0.034143999218940735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,8,4,power_law_1.2,0.035104000568389894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,8,4,power_law_1.2,0.03511039912700653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,8,4,power_law_1.2,0.0361407995223999
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,8,4,power_law_1.2,0.0359360009431839
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,1,power_law_1.2,0.6465407848358155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,8,4,power_law_1.2,0.03606399893760681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,8,4,power_law_1.2,0.03701759874820709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,8,4,power_law_1.2,0.0369024008512497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,8,4,power_law_1.2,0.03966720104217529
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,8,4,power_law_1.2,0.03969280123710632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,8,4,power_law_1.2,0.042828801274299624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,8,4,power_law_1.2,0.05103999972343445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,2,power_law_1.2,9.74471664428711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,1,power_law_1.2,0.9230912208557129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,8,4,power_law_1.2,0.05308160185813904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,8,4,power_law_1.2,0.054681599140167236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,8,4,power_law_1.2,0.061504000425338747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,8,4,power_law_1.2,0.07206400036811829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,8,4,power_law_1.2,0.09687039852142335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,8,4,power_law_1.2,0.09895679950714112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,8,4,power_law_1.2,0.14162559509277345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,1,power_law_1.2,0.9736384391784668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,8,4,power_law_1.2,0.17066880464553832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,8,4,power_law_1.2,0.2216320037841797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,8,4,power_law_1.2,0.3176896095275879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,8,4,power_law_1.2,0.4059648036956787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,8,4,power_law_1.2,0.5493055820465088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,1,power_law_1.2,0.9828351974487305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,8,4,power_law_1.2,0.7305151939392089
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,8,4,power_law_1.2,0.9967743873596191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,1,power_law_1.2,1.0634752273559571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,8,4,power_law_1.2,1.8755199432373046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,1,power_law_1.2,1.1048704147338868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,2,16,power_law_1.01,0.025523200631141663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,2,16,power_law_1.01,0.021804800629615782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,1,power_law_1.2,1.1448384284973145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,2,16,power_law_1.01,0.022860799729824067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,2,16,power_law_1.01,0.02305919975042343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,2,16,power_law_1.01,0.02470400035381317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,1,power_law_1.2,1.1973312377929688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,2,16,power_law_1.01,0.024166400730609893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,2,16,power_law_1.01,0.025567999482154845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,2,16,power_law_1.01,0.029491201043128967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,1,power_law_1.2,1.2760767936706543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,2,16,power_law_1.01,0.031481599807739256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,4,8,balanced,0.03801066676775614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,4,8,balanced,0.03603200117746989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,4,8,balanced,0.036144000788529716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,4,8,balanced,0.03615466753641764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,4,8,balanced,0.04001066585381826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,2,16,power_law_1.01,0.032595199346542356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,4,8,balanced,0.04200000067551931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,4,8,balanced,0.042175998290379844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,4,8,balanced,0.04186666508515676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,4,8,balanced,0.04018666595220566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,4,8,balanced,0.04171733558177948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,4,8,balanced,0.04195199906826019
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,4,8,balanced,0.0422026664018631
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,4,8,balanced,0.0443146675825119
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,2,16,power_law_1.01,0.033632001280784606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,4,8,balanced,0.042117332418759666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,4,8,balanced,0.04610133171081543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,1,power_law_1.2,1.2676735877990724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,4,8,balanced,0.04875733455022176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,4,8,balanced,0.04903466502825419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,4,8,balanced,0.052101333936055504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,4,8,balanced,0.05669333537419637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,2,16,power_law_1.01,0.034790399670600894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,4,8,balanced,0.0622026671965917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,4,8,balanced,0.06623999774456024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,4,8,balanced,0.08230400085449219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,4,8,balanced,0.09308266639709473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,4,8,balanced,0.11871467034022014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,2,16,power_law_1.01,0.03570559918880463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,4,8,balanced,0.15356266498565674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,4,8,balanced,0.21226133902867636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,4,8,balanced,0.2606186668078105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,4,8,balanced,0.31747732559839886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,2,16,power_law_1.01,0.03470079898834229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,4,8,balanced,0.47167468070983887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,1,power_law_1.2,1.333612823486328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,4,8,balanced,0.9052373568216959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,2,16,power_law_1.01,0.03863680064678192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,2,16,power_law_1.01,0.03939839899539947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,2,16,power_law_1.01,0.039366400241851805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,1,power_law_1.2,1.4393919944763183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,2,16,power_law_1.01,0.04431999921798706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,4,balanced,0.053674668073654175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,4,balanced,0.08801066875457764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,4,balanced,0.1037546694278717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,16,power_law_1.01,0.047302401065826415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,4,balanced,0.1553386648495992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,4,balanced,0.25805866718292236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,4,balanced,0.31453333298365277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,4,balanced,0.3137386639912923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,4,balanced,0.31436800956726074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,4,balanced,0.31510400772094727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,16,power_law_1.01,0.05196800231933594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,4,balanced,0.3184266686439514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,1,power_law_1.2,1.5362367630004883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,4,balanced,0.3200053373972575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,4,balanced,0.32340266307195026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,4,balanced,0.3263733386993408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,4,balanced,0.33053332567214966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,4,balanced,0.33822933832804364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,16,power_law_1.01,0.058963197469711306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,4,balanced,0.3391253153483073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,4,balanced,0.348037322362264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,4,balanced,0.3728906710942586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,4,balanced,0.37533334891001385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,4,balanced,0.4605546792348226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,4,balanced,0.4360586802164714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,16,power_law_1.01,0.07075200080871583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,4,balanced,0.6905279954274496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,4,balanced,0.5678879817326864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,4,balanced,0.8620320161183676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,1,power_law_1.2,1.7467967987060546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,4,balanced,0.8931252956390381
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,16,power_law_1.01,0.07813760042190551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,4,balanced,1.3997440338134766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,4,balanced,1.5980960528055828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,16,power_law_1.01,0.10261759757995606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,4,balanced,1.785589377085368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,16,power_law_1.01,0.1262719988822937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,4,balanced,2.876005490620931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,1,power_law_1.2,1.839148712158203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,16,power_law_1.01,0.17918720245361328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,4,balanced,5.436815897623698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,1,power_law_1.2,2.2530496597290037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,16,power_law_1.01,0.2106112003326416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,16,power_law_1.01,0.278656005859375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,1,power_law_1.2,2.4003456115722654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,16,power_law_1.01,0.4132224082946777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,32,balanced,0.042165334026018776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,32,balanced,0.03789866715669632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,32,balanced,0.03579733272393545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,32,balanced,0.035973332822322845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,32,balanced,0.038160001238187156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,32,balanced,0.03818133225043615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,32,balanced,0.03758399933576584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,32,balanced,0.037952000896135964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,32,balanced,0.0383840004603068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,32,balanced,0.038506666819254555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,32,balanced,0.037978666524092354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,32,balanced,0.03782399992148081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,32,balanced,0.04014399896065394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,32,balanced,0.039674667020638786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,32,balanced,0.04215999941031138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,32,balanced,0.04317333300908407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,32,balanced,0.04412800073623657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,1,power_law_1.2,2.988153648376465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,32,balanced,0.045269335309664406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,32,balanced,0.04621333380540212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,32,balanced,0.052560001611709595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,32,balanced,0.05219733218352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,32,balanced,0.06494399905204773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,32,balanced,0.06643733382225037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,32,balanced,0.09178666273752849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,32,balanced,0.09738133351008098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,32,balanced,0.12601066629091898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,32,balanced,0.1508799990018209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,2688,1856,6,128,1,32,balanced,0.17520533005396524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,2688,1856,6,128,1,32,balanced,0.252128005027771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,16,power_law_1.01,0.8150400161743164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,2688,1856,6,128,1,32,balanced,0.4428266684214274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,1,power_law_1.2,3.6011009216308594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,1,power_law_1.2,4.711072158813477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,4,16,power_law_1.2,0.026316800713539125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,4,16,power_law_1.2,0.022227199375629426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,4,16,power_law_1.2,0.02346239984035492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,1,power_law_1.2,5.843443298339844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,4,16,power_law_1.2,0.021984000504016877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,4,16,power_law_1.2,0.02309119999408722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,4,16,power_law_1.2,0.024153600633144378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,4,16,power_law_1.2,0.024512000381946564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,4,16,power_law_1.2,0.028428798913955687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,4,16,power_law_1.2,0.029785600304603577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,1,power_law_1.2,7.04681625366211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,4,16,power_law_1.2,0.029817599058151244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,4,16,power_law_1.2,0.030847999453544616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,4,16,power_law_1.2,0.031167998909950256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,4,16,power_law_1.2,0.031385600566864014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,16,power_law_1.01,0.059001600742340087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,32,1,power_law_1.01,0.07541120052337646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,4,16,power_law_1.2,0.031123200058937074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,4,16,power_law_1.2,0.03405439853668213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,32,1,power_law_1.01,0.08184319734573364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,4,16,power_law_1.2,0.03537279963493347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,16,power_law_1.01,0.06320000290870667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,4,16,power_law_1.2,0.03763839900493622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,32,1,power_law_1.01,0.08574720025062561
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,4,16,power_law_1.2,0.041203200817108154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,1,power_law_1.2,9.984352111816406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,32,1,power_law_1.01,0.10895359516143799
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,16,power_law_1.2,0.044275200366973876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,16,power_law_1.2,0.0506816029548645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,32,1,power_law_1.01,0.13515520095825195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,16,power_law_1.01,0.05676800012588501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,16,power_law_1.2,0.05671039819717407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,16,power_law_1.2,0.07776640057563781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,32,1,power_law_1.01,0.16010240316390992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,16,power_law_1.2,0.08280959725379944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,32,1,power_law_1.01,0.19288320541381837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,16,power_law_1.01,0.054636800289154054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,16,power_law_1.2,0.11201920509338378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,32,1,power_law_1.01,0.20631680488586426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,16,power_law_1.2,0.16554880142211914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,16,power_law_1.01,0.055270397663116456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,32,1,power_law_1.01,0.21134719848632813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,16,power_law_1.2,0.23518719673156738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,32,1,power_law_1.01,0.21427199840545655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,16,power_law_1.2,0.3218303918838501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,16,power_law_1.01,0.058822399377822875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,32,1,power_law_1.01,0.22533121109008789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,1,power_law_1.2,18.728108215332032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,16,power_law_1.2,0.37601280212402344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,32,1,power_law_1.01,0.23266561031341554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,16,power_law_1.01,0.060684800148010254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,32,1,power_law_1.01,0.23990399837493898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,16,power_law_1.2,0.571014404296875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,32,1,power_law_1.01,0.2472383975982666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,16,power_law_1.01,0.06188160181045532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,32,1,power_law_1.01,0.2599744081497192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,16,power_law_1.01,0.06173440217971802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,32,1,power_law_1.01,0.2637631893157959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,16,power_law_1.2,1.1381952285766601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,32,1,power_law_1.01,0.2829952001571655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,16,power_law_1.01,0.062003201246261595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,32,1,power_law_1.01,0.32967679500579833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,16,power_law_1.01,0.06298879981040954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,32,1,power_law_1.01,0.36696319580078124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,16,power_law_1.01,0.0648959994316101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,32,1,power_law_1.01,0.45957121849060056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,32,1,power_law_1.01,0.5393663883209229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,16,power_law_1.01,0.06613759994506836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,32,1,power_law_1.01,0.712985610961914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,16,power_law_1.01,0.06700159907341004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,32,1,power_law_1.01,0.88156156539917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,64,balanced,0.07401599983374278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,64,balanced,0.04674666623274485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,64,balanced,0.046096002062161766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,64,balanced,0.04775999983151754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,64,balanced,0.05021866659323374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,64,balanced,0.07270933190981548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,16,power_law_1.01,0.07221119999885559
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,64,balanced,0.07382933298746745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,64,balanced,0.07435200115044911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,64,balanced,0.07470400134722392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,64,balanced,0.07474666833877563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,64,balanced,0.07491733133792877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,64,balanced,0.07635733485221863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,64,balanced,0.07829866806666057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,64,balanced,0.07678399980068207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,64,balanced,0.08243733147780101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,64,balanced,0.0820853312810262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,64,balanced,0.08340799808502197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,64,balanced,0.08861866593360901
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,64,balanced,0.09243733684221904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,32,1,power_law_1.01,1.2218048095703125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,64,balanced,0.09956799944241841
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,64,balanced,0.10868799686431885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,64,balanced,0.1241439978281657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,64,balanced,0.14014400045077005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,64,balanced,0.1766186753908793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,64,balanced,0.20920000473658243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,16,power_law_1.01,0.07541120052337646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,64,balanced,0.28386133909225464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,64,balanced,0.33839468161265057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,64,balanced,0.4158933162689209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,64,balanced,0.5951840082804362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,64,balanced,1.1443146864573162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,32,1,power_law_1.01,1.5683199882507324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,16,power_law_1.01,0.07938560247421264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,16,power_law_1.01,0.09091200232505799
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,32,1,power_law_1.01,2.253100776672363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,4,power_law_1.2,0.12039680480957031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,16,power_law_1.01,0.09939839839935302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,32,1,power_law_1.01,2.996294403076172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,16,power_law_1.01,0.12097280025482178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,4,power_law_1.2,0.11701760292053223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,16,power_law_1.01,0.13728640079498292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,32,1,power_law_1.01,3.66943359375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,4,power_law_1.2,0.10234240293502808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,16,power_law_1.01,0.1814463973045349
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,4,power_law_1.2,0.1523519992828369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,16,power_law_1.01,0.2148736000061035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,32,1,power_law_1.01,5.630080032348633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,4,power_law_1.2,0.1969599962234497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,16,power_law_1.01,0.3111743927001953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,32,4,balanced,0.03788266579310099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,32,4,balanced,0.0354720006386439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,32,4,balanced,0.0354666660229365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,32,4,balanced,0.03572266548871994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,32,4,balanced,0.03541333228349686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,32,4,balanced,0.03588266670703888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,32,4,balanced,0.03536533315976461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,32,4,balanced,0.03765333443880081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,32,4,balanced,0.03745600084463755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,32,4,balanced,0.03579733272393545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,4,power_law_1.2,0.32006399631500243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,32,4,balanced,0.037808001041412354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,32,4,balanced,0.0378560001651446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,16,power_law_1.01,0.3515392065048218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,32,4,balanced,0.03958400090535482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,32,4,balanced,0.037776000797748566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,32,4,balanced,0.04166933397452036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,32,4,balanced,0.043738668163617454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,32,4,balanced,0.04389866689840952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,32,4,balanced,0.04577599962552389
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,32,4,balanced,0.04971733192602793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,32,4,balanced,0.05610666672388712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,32,4,balanced,0.06026133398214976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,32,4,balanced,0.07052800059318542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,32,4,balanced,0.08668800195058186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,32,4,balanced,0.1178559958934784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,32,4,balanced,0.1402133305867513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,32,4,balanced,0.18933866421381632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,32,4,balanced,0.23043199380238852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,32,4,balanced,0.2799839973449707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,32,4,balanced,0.4134986797968547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,32,4,balanced,0.7756266593933105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,4,power_law_1.2,0.36058878898620605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,16,power_law_1.01,0.5197247982025146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,32,1,power_law_1.01,11.038841247558594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,4,power_law_1.2,0.3821887969970703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,16,power_law_1.01,0.6990464210510254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,4,power_law_1.2,0.38502399921417235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,16,power_law_1.01,0.7770815849304199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,4,power_law_1.2,0.40052480697631837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,16,power_law_1.01,1.1708864212036132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,4,power_law_1.2,0.4074431896209717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,4,power_law_1.2,0.42723841667175294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,4,power_law_1.2,0.44147839546203616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,16,power_law_1.01,2.4316032409667967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,4,power_law_1.2,0.4517183780670166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,64,power_law_1.2,0.04723840057849884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,64,balanced,0.039488000174363456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,64,balanced,0.03367999941110611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,64,balanced,0.02924799919128418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,64,balanced,0.02940800040960312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,64,balanced,0.029669334491093952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,64,balanced,0.02942933390537898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,64,power_law_1.2,0.03573760092258453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,64,balanced,0.031504000226656594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,64,balanced,0.03136533250411352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,64,balanced,0.029359998802344005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,64,balanced,0.03127466638882955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,64,balanced,0.031343999008337654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,64,balanced,0.03143466760714849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,64,balanced,0.03332799921433131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,64,balanced,0.03165333221356074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,64,balanced,0.03523733218510946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,64,balanced,0.03573333223660787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,64,balanced,0.03557866563399633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,4,power_law_1.2,0.46834559440612794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,64,balanced,0.03766933331886927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,64,power_law_1.2,0.029190400242805482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,64,balanced,0.0354666660229365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,64,balanced,0.03979199876387914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,64,balanced,0.0395359992980957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,64,balanced,0.04167999823888143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,64,balanced,0.04383466641108195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,64,balanced,0.05083199838797251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,64,balanced,0.055919999877611794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,64,balanced,0.06824000179767609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,64,balanced,0.08063999811808269
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,2,64,balanced,0.08755200107892354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,64,power_law_1.2,0.03203200101852417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,2,64,balanced,0.11526399850845337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,2,64,balanced,0.1807039976119995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,64,power_law_1.2,0.030425599217414855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,64,power_law_1.2,0.031974399089813234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,4,power_law_1.2,0.4733119964599609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,64,power_law_1.2,0.03146879971027374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,64,power_law_1.2,0.031692799925804135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,64,power_law_1.2,0.031513598561286923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,4,power_law_1.2,0.5164415836334229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,64,power_law_1.2,0.03214719891548157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,64,power_law_1.2,0.03227519989013672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,64,power_law_1.2,0.03366400003433227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,4,power_law_1.2,0.6085887908935547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,64,power_law_1.2,0.03425279855728149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,64,power_law_1.2,0.03619199991226196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,64,power_law_1.2,0.04280959963798523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,4,power_law_1.2,0.5872447967529297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,64,power_law_1.2,0.04227840006351471
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,64,power_law_1.2,0.04387199878692627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,64,power_law_1.2,0.05319679975509643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,4,power_law_1.2,0.6879104137420654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,64,power_law_1.2,0.05271040201187134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,1,1,power_law_1.2,0.045228800177574156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,64,power_law_1.2,0.061433601379394534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,64,power_law_1.2,0.06709759831428527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,1,1,power_law_1.2,0.04987519979476929
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,64,power_law_1.2,0.07930880188941955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,4,power_law_1.2,0.7302720069885253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,64,power_law_1.2,0.09710720181465149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,1,1,power_law_1.2,0.07150080204010009
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,4,1,balanced,0.0554666668176651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,4,1,balanced,0.09363733728726704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,4,1,balanced,0.1179200013478597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,4,1,balanced,0.11969600121180217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,4,1,balanced,0.12008532881736755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,4,1,balanced,0.12161067128181458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,64,power_law_1.2,0.13516160249710082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,4,1,balanced,0.1213653286298116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,4,1,balanced,0.12294933199882507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,4,1,balanced,0.12387733658154805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,4,1,balanced,0.12436800201733907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,4,1,balanced,0.1269813378651937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,4,1,balanced,0.12873066465059915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,4,1,balanced,0.13016000390052795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,4,1,balanced,0.13610133528709412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,4,1,balanced,0.15042133132616678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,4,1,balanced,0.15412267049153647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,4,1,balanced,0.15656532843907675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,1,1,power_law_1.2,0.08870400190353393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,4,1,balanced,0.18313600619633993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,4,power_law_1.2,0.8967743873596191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,64,power_law_1.2,0.1852288007736206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,4,1,balanced,0.19555733601252237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,4,1,balanced,0.2471733291943868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,4,1,balanced,0.2837439974149068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,4,1,balanced,0.37970133622487384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,4,1,balanced,0.47607465585072833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,4,1,balanced,0.6756160259246826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,4,1,balanced,0.8675039609273275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,64,power_law_1.2,0.2755327939987183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,1,1,power_law_1.2,0.11399680376052856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,4,1,balanced,1.2714293003082275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,4,1,balanced,1.6634559631347656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,4,1,balanced,2.0538454055786133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,1,1,power_law_1.2,0.18394880294799804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,64,power_law_1.2,0.3110464096069336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,4,power_law_1.2,0.9976703643798828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,4,1,balanced,3.2161973317464194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,1,1,power_law_1.2,0.19449599981307983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,1,64,power_law_1.2,0.40208001136779786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,4,1,balanced,6.306351979573567
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,1,1,power_law_1.2,0.20740480422973634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,1,64,power_law_1.2,0.5814015865325928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,4,power_law_1.2,1.1866559982299805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,1,1,power_law_1.2,0.21125760078430175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,1,1,power_law_1.2,0.21066880226135254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,1,1,power_law_1.2,0.2221247911453247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,1,64,power_law_1.2,1.363974380493164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,4,power_law_1.2,1.5345664024353027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,1,1,power_law_1.2,0.22206079959869385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,1,1,power_law_1.2,0.22296319007873536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,1,1,power_law_1.2,0.23495039939880372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,1,1,power_law_1.2,0.24461441040039061
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,4,power_law_1.2,2.062656021118164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,1,1,power_law_1.2,0.2517184019088745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,1,1,power_law_1.2,0.2594111919403076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,1,1,power_law_1.2,0.2882368087768555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,1,power_law_1.2,0.30678400993347166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,4,power_law_1.2,2.4415231704711915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,1,power_law_1.2,0.3634047985076904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,1,power_law_1.2,0.40343680381774905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,1,power_law_1.2,0.5096767902374267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,2,power_law_1.2,0.09384959936141968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,1,power_law_1.2,0.5956672191619873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,1,4,power_law_1.2,2.735340881347656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,2,power_law_1.2,0.14206720590591432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,1,power_law_1.2,0.789740800857544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,1,power_law_1.2,0.9751872062683106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,2,power_law_1.2,0.1474560022354126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,2,power_law_1.2,0.03779839873313904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,2,power_law_1.2,0.03986560106277466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,1,power_law_1.2,1.3506943702697753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,2,power_law_1.2,0.16750719547271728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,2,power_law_1.2,0.043968001008033754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,2,power_law_1.2,0.06177279949188232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,2,power_law_1.2,0.29863040447235106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,2,power_law_1.2,0.07120640277862549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,1,power_law_1.2,1.7234432220458984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,2,power_law_1.2,0.10686719417572021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,1,4,power_law_1.2,4.744499206542969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,2,power_law_1.2,0.11085439920425415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,2,power_law_1.2,0.42860798835754393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,2,power_law_1.2,0.1165120005607605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,1,power_law_1.2,2.099507141113281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,2,power_law_1.2,0.11930880546569825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,2,power_law_1.2,0.5317311763763428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,2,power_law_1.2,0.12284799814224243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,2,power_law_1.2,0.12052479982376099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,2,power_law_1.2,0.12299519777297974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,2,power_law_1.2,0.7281599998474121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,1,power_law_1.2,3.2620094299316404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,2,power_law_1.2,0.1284991979598999
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,2,power_law_1.2,0.12937599420547485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,2,power_law_1.2,0.7814015865325927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,2,power_law_1.2,0.14220800399780273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,2,power_law_1.2,0.14572800397872926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,2,power_law_1.2,0.8123968124389649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,2,power_law_1.2,0.15481599569320678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,2,power_law_1.2,0.17401599884033203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,1,power_law_1.2,6.377215957641601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,2,power_law_1.2,0.18305920362472533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,2,power_law_1.2,0.8334976196289062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,2,power_law_1.2,0.22187519073486328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,2,power_law_1.2,0.2457792043685913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,2,power_law_1.2,0.8779456138610839
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,1,4,power_law_1.2,8.828704071044921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,2,power_law_1.2,0.32641279697418213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,2,power_law_1.2,0.3739392042160034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,2,power_law_1.2,0.902348804473877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,2,power_law_1.2,0.5230847835540772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,2,power_law_1.2,0.6818304061889648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,2,power_law_1.2,0.9380991935729981
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,2,power_law_1.2,0.890777587890625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,2,power_law_1.2,1.0003904342651366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,2,power_law_1.2,1.292147159576416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,2,power_law_1.2,1.009023952484131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,2,power_law_1.2,1.403116798400879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,2,power_law_1.2,1.096230411529541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,2,power_law_1.2,2.073535919189453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,2,power_law_1.2,1.1467519760131837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,32,balanced,0.05612266560395559
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,32,balanced,0.05922666688760122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,32,balanced,0.0581279993057251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,2,power_law_1.2,1.2065983772277833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,32,balanced,0.06301866471767426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,32,balanced,0.07260266443093617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,32,balanced,0.09132267038027446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,32,balanced,0.11274666587511699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,32,balanced,0.110042671362559
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,32,balanced,0.11050666371981303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,32,balanced,0.11409599582354228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,32,balanced,0.1153600017229716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,32,balanced,0.11591466267903645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,32,balanced,0.11492266257603963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,32,balanced,0.11642133196194966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,32,balanced,0.12785599629084268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,1,2,balanced,0.07590400179227193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,2,power_law_1.2,4.261324691772461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,1,2,balanced,0.10342400272687276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,32,balanced,0.12262933452924092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,1,2,balanced,0.15011733770370483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,1,2,balanced,0.15029333035151163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,2,power_law_1.2,1.3873087882995605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,32,balanced,0.1283253331979116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,1,2,balanced,0.14892266194025675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,1,2,balanced,0.15059733390808105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,32,balanced,0.13910399874051413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,1,2,balanced,0.1532426675160726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,1,2,balanced,0.1532373329003652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,32,balanced,0.1516746679941813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,1,2,balanced,0.15421866377194723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,32,balanced,0.17193599541982016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,1,2,balanced,0.15320533514022827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,1,2,balanced,0.15427199999491373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,32,balanced,0.19057599703470865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,1,2,balanced,0.15596266587575278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,1,2,balanced,0.1575040022532145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,32,balanced,0.23162666956583658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,1,2,balanced,0.16337600350379944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,1,2,balanced,0.1711946725845337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,32,balanced,0.2723199923833211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,1,2,balanced,0.1750133236249288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,1,2,balanced,0.18101867039998373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,32,balanced,0.3540319999059041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,1,2,balanced,0.21824532747268677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,1,2,balanced,0.2290239930152893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,32,balanced,0.4508853356043498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,1,2,balanced,0.2767413258552551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,1,2,balanced,0.33720000584920246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,32,balanced,0.6106986602147421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,1,2,balanced,0.44947199026743573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,2,power_law_1.2,1.448185634613037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,32,balanced,0.7876479625701904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,1,2,balanced,0.5645493268966675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,1,2,balanced,0.784986654917399
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,32,balanced,0.9883733590443929
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,1,2,balanced,1.0192053318023682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,32,balanced,1.5098293622334797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,1,2,balanced,1.460624059041341
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,1,2,balanced,1.9116214116414387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,32,balanced,2.9275360107421875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,2,power_law_1.2,1.7040319442749023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,1,2,balanced,2.3627732594807944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,1,2,balanced,3.6780853271484375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,2,power_law_1.2,1.9041088104248047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,1,2,balanced,7.164165496826172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,2,power_law_1.2,2.360787200927734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,2,power_law_1.2,2.4808832168579102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,8,balanced,0.05195199946562449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,2,power_law_1.2,3.0451072692871093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,8,balanced,0.05197866757710775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,8,balanced,0.05420266588528951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,8,balanced,0.056261335810025535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,8,balanced,0.06198399762312571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,8,balanced,0.076773335536321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,8,balanced,0.0788213312625885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,8,balanced,0.08083733419577281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,8,balanced,0.08072533210118611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,8,balanced,0.08074133098125458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,8,balanced,0.08296533425649007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,8,balanced,0.08451732993125916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,8,balanced,0.08471999565760295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,8,balanced,0.08746666709582011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,8,balanced,0.0939359962940216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,8,balanced,0.09718400239944458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,8,balanced,0.1011786659558614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,8,balanced,0.11123733719189961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,8,balanced,0.12206400434176128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,8,balanced,0.13894933462142944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,8,balanced,0.16361066699028015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,8,balanced,0.19332265853881836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,8,balanced,0.217738668123881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,8,balanced,0.3006880084673564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,8,balanced,0.3773226737976074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,8,balanced,0.5311466852823893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,8,balanced,0.681658665339152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,8,balanced,0.8357653617858887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,2,power_law_1.2,4.092704010009766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,8,balanced,1.279157320658366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,8,balanced,2.4898826281229653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,64,power_law_1.2,0.03484799861907959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,2,power_law_1.2,4.561363220214844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,1,power_law_1.01,0.05683839917182922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,64,power_law_1.2,0.03457919955253601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,64,power_law_1.2,0.030508801341056824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,64,power_law_1.2,0.03203839957714081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,64,power_law_1.2,0.030969598889350893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,1,power_law_1.01,0.07640320062637329
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,64,power_law_1.2,0.03191039860248566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,64,power_law_1.2,0.0321727991104126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,2,power_law_1.2,6.502329254150391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,64,power_law_1.2,0.03240320086479187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,64,power_law_1.2,0.0328000009059906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,64,power_law_1.2,0.03345920145511627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,1,power_law_1.01,0.1059648036956787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,64,power_law_1.2,0.03362559974193573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,64,power_law_1.2,0.03361920118331909
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,64,power_law_1.2,0.03544960021972656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,64,power_law_1.2,0.0389631986618042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,1,power_law_1.01,0.12328319549560547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,64,power_law_1.2,0.0435263991355896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,64,power_law_1.2,0.04661119878292084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,64,power_law_1.2,0.04615679979324341
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,2,power_law_1.2,11.81088638305664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,1,power_law_1.01,0.2046272039413452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,64,power_law_1.2,0.051072001457214355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,64,power_law_1.2,0.05995519757270813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,64,power_law_1.2,0.07761279940605163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,4,power_law_1.01,0.11984000205993653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,64,power_law_1.2,0.07916799783706666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,1,power_law_1.01,0.2790463924407959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,64,power_law_1.2,0.09654399752616882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,64,power_law_1.2,0.14323840141296387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,4,power_law_1.01,0.18696320056915283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,64,power_law_1.2,0.2187648057937622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,1,power_law_1.01,0.34082560539245604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,64,power_law_1.2,0.24895360469818115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,4,power_law_1.01,0.10298880338668823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,64,power_law_1.2,0.3390592098236084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,1,power_law_1.01,0.43021440505981445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,4,power_law_1.01,0.15494400262832642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,64,power_law_1.2,0.560588788986206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,1,power_law_1.01,0.45169920921325685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,64,power_law_1.2,0.7055168151855469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,4,power_law_1.01,0.21843841075897216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,64,power_law_1.2,0.8301247596740723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,1,power_law_1.01,0.4771967887878418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,4,power_law_1.01,0.33255040645599365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,2,balanced,0.04770666857560476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,2,balanced,0.04809600114822388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,2,balanced,0.054901331663131714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,2,balanced,0.07859733204046886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,2,balanced,0.10898666580518086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,2,balanced,0.16925332943598428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,2,balanced,0.1716266671816508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,2,balanced,0.1676373283068339
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,2,balanced,0.16978132724761963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,2,balanced,0.1713599960009257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,2,balanced,0.17328532536824545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,2,balanced,0.17719467480977377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,1,power_law_1.01,0.48990721702575685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,4,power_law_1.01,0.35207679271698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,2,balanced,0.18045334021250406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,2,balanced,0.18589866161346436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,2,balanced,0.1933493415514628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,2,balanced,0.19985600312550864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,2,balanced,0.22145599126815796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,2,balanced,0.24545600016911825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,64,power_law_1.2,1.915603256225586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,2,balanced,0.26295467217763263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,2,balanced,0.3070613344510396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,2,balanced,0.3489706516265869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,2,balanced,0.4811626672744751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,2,balanced,0.5449973344802856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,2,balanced,0.7778560320536295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,2,balanced,0.9530026912689209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,2,balanced,1.3621652921040852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,4,power_law_1.01,0.3790015935897827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,1,power_law_1.01,0.5069952011108398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,2,balanced,1.7431999842325847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,2,balanced,2.207610607147217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,2,balanced,3.3384265899658203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,4,power_law_1.01,0.38540799617767335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,1,power_law_1.01,0.5331071853637696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,2,balanced,6.452512105305989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,4,power_law_1.01,0.38597118854522705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,1,power_law_1.01,0.5568192005157471
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,4,power_law_1.01,0.39714560508728025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,1,power_law_1.01,0.5673408031463623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,4,power_law_1.01,0.40721921920776366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,1,power_law_1.01,0.5898752212524414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,4,power_law_1.01,0.4183743953704834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,1,power_law_1.01,0.617574405670166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,4,power_law_1.01,0.43315839767456055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,2,power_law_1.01,0.03320960104465485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,1,power_law_1.01,0.6837887763977051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,4,power_law_1.01,0.4425151824951172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,2,power_law_1.01,0.030822399258613586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,2,power_law_1.01,0.03164800107479095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,1,power_law_1.01,0.7380352020263672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,2,power_law_1.01,0.03493120074272156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,4,power_law_1.01,0.45328640937805176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,2,power_law_1.01,0.036339199542999266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,4,balanced,0.03365866591533025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,4,balanced,0.03563733398914337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,4,balanced,0.04580800235271454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,4,balanced,0.07374933362007141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,4,balanced,0.1127839982509613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,4,balanced,0.12217066685358684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,4,balanced,0.126202662785848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,4,balanced,0.12389333049456279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,4,balanced,0.12418666481971741
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,2,power_law_1.01,0.04186240136623383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,4,balanced,0.12452266613642375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,4,balanced,0.12397866447766621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,4,balanced,0.1258240044116974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,4,balanced,0.12596266468365988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,4,balanced,0.1265120009581248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,4,balanced,0.13037332892417908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,4,balanced,0.13306132952372232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,4,balanced,0.13501866658528647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,1,power_law_1.01,0.8498432159423828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,4,balanced,0.16315733393033346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,2,power_law_1.01,0.042847999930381776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,4,balanced,0.1397760013739268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,4,balanced,0.1657546659310659
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,4,balanced,0.15934933225313822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,4,power_law_1.01,0.46634879112243655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,4,balanced,0.23174933592478433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,4,balanced,0.20090667406717935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,4,balanced,0.28965334097544354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,2,power_law_1.01,0.04376319944858551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,4,balanced,0.28938666979471844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,4,balanced,0.39610668023427326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,4,balanced,0.4270506699879964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,1,4,balanced,0.5201226472854614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,2,power_law_1.01,0.04522239863872528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,1,4,balanced,0.7554612954457601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,1,4,balanced,1.4332693417867024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,2,power_law_1.01,0.04665600061416626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,1,power_law_1.01,0.9467904090881347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,4,power_law_1.01,0.5135295867919922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,2,power_law_1.01,0.04890879988670349
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,2,power_law_1.01,0.05330560207366943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,2,power_law_1.01,0.05487359762191772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,1,power_law_1.01,1.191763210296631
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,2,power_law_1.01,0.05615360140800476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,4,power_law_1.01,0.5614463806152343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,2,power_law_1.01,0.06386560201644897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,2,power_law_1.01,0.06798719763755798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,2,power_law_1.01,0.07051519751548767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,4,power_law_1.01,0.6431615829467774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,1,power_law_1.01,1.4836288452148438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,2,power_law_1.01,0.0825215995311737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,2,power_law_1.01,0.09488000273704529
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,2,power_law_1.01,0.12208000421524048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,4,power_law_1.01,0.7340095996856689
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,2,power_law_1.01,0.13912960290908813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,1,power_law_1.01,1.9167104721069337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,2,power_law_1.01,0.1834239959716797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,2,power_law_1.01,0.22855679988861083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,4,power_law_1.01,0.9217215538024902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,2,power_law_1.01,0.332044792175293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,1,power_law_1.01,2.206867218017578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,2,power_law_1.01,0.47285118103027346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,2,power_law_1.01,0.6229055881500244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,4,power_law_1.01,0.9162431716918945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,2,power_law_1.01,0.8502400398254395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,1,power_law_1.01,2.955187225341797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,256,power_law_1.2,0.07266560196876526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,2,power_law_1.01,1.0476096153259278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,4,power_law_1.01,1.1227519989013672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,256,power_law_1.2,0.07521920204162598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,2,power_law_1.01,1.704275131225586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,1,power_law_1.01,3.8352256774902345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,4,power_law_1.01,1.3796223640441894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,256,power_law_1.2,0.0710591971874237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,2,power_law_1.01,3.453548812866211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,256,power_law_1.2,0.07938560247421264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,1,power_law_1.01,4.673657608032227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,4,power_law_1.01,1.9776447296142579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,256,power_law_1.2,0.07813119888305664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,256,power_law_1.2,0.08089600205421447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,1,power_law_1.01,6.846157073974609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,4,power_law_1.01,2.3325056076049804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,256,power_law_1.2,0.07782400250434876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,256,power_law_1.2,0.07599999904632568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,256,power_law_1.2,0.07708160281181335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,2,4,power_law_1.01,2.6347328186035157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,256,power_law_1.2,0.07448959946632386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,1,power_law_1.01,12.838278198242188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,256,power_law_1.2,0.07728000283241272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,1,power_law_1.2,0.08652799725532531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,256,power_law_1.2,0.07949439883232116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,2,4,power_law_1.01,4.264550399780274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,1,power_law_1.2,0.09097599983215332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,256,power_law_1.2,0.07868160009384155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,1,power_law_1.2,0.10399359464645386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,256,power_law_1.2,0.07559679746627808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,1,power_law_1.2,0.12275840044021606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,256,power_law_1.2,0.08164479732513427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,1,power_law_1.2,0.16860159635543823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,1,power_law_1.2,0.20458879470825195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,256,power_law_1.2,0.08586879968643188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,1,power_law_1.2,0.2312256097793579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,256,power_law_1.2,0.09449599981307984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,2,8,balanced,0.04237333436806997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,1,power_law_1.2,0.2921663999557495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,2,8,balanced,0.040896000961462654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,2,4,power_law_1.01,7.897331237792969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,2,8,balanced,0.03882133215665817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,2,8,balanced,0.03974399964014689
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,2,8,balanced,0.04114133367935816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,2,8,balanced,0.04127999891837438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,2,8,balanced,0.03969600051641464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,256,power_law_1.2,0.10812159776687622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,2,8,balanced,0.04094400008519491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,2,8,balanced,0.0428959975639979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,2,8,balanced,0.04154133299986521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,2,8,balanced,0.04334933559099833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,2,8,balanced,0.04368533194065094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,1,power_law_1.2,0.309004807472229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,2,8,balanced,0.04505600035190582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,2,8,balanced,0.047024001677831016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,2,8,balanced,0.05640000104904175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,2,8,balanced,0.05689600110054016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,2,8,balanced,0.05894400179386139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,2,8,balanced,0.07081066568692525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,2,8,balanced,0.07678399980068207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,2,8,balanced,0.09969600041707356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,2,8,balanced,0.09733333190282185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,2,8,balanced,0.12687466541926065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,2,8,balanced,0.14301333824793497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,2,8,balanced,0.1962666710217794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,2,8,balanced,0.23650133609771729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,256,power_law_1.2,0.12243200540542602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,2,8,balanced,0.33293867111206055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,1,power_law_1.2,0.3201215982437134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,2,8,balanced,0.4240479866663615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,2,8,balanced,0.5220106840133667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,2,8,balanced,0.787834644317627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,2,8,balanced,1.5179252624511719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,1,power_law_1.2,0.47562880516052247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,256,power_law_1.2,0.1561535954475403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,1,power_law_1.2,0.5098048210144043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,256,power_law_1.2,0.17940479516983032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,1,power_law_1.2,0.5173888206481934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,256,power_law_1.2,0.24379520416259765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,1,power_law_1.2,0.5448575973510742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,1,power_law_1.2,0.5694784164428711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,256,power_law_1.2,0.28305280208587646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,1,power_law_1.2,0.5769408226013184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,256,power_law_1.2,0.37342081069946287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,1,power_law_1.2,0.6034560203552246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,16,power_law_1.01,0.06968320012092591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,1,power_law_1.2,0.6661888122558594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,16,power_law_1.01,0.12087039947509766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,256,power_law_1.2,0.5051072120666504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,1,power_law_1.2,0.7189184188842773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,16,power_law_1.01,0.12286720275878907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,1,power_law_1.2,0.8511487960815429
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,16,power_law_1.01,0.06183680295944214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,256,power_law_1.2,0.7273727893829346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,8,power_law_1.01,0.07505919933319091
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,1,power_law_1.2,0.9442239761352539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,16,power_law_1.01,0.0737600028514862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,16,power_law_1.01,0.085343998670578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,1,power_law_1.2,1.1586879730224608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,256,power_law_1.2,1.0076607704162597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,8,power_law_1.01,0.09804800152778625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,16,power_law_1.01,0.09219200015068055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,1,power_law_1.2,1.3400128364562989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,16,power_law_1.01,0.10151040554046631
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,1,power_law_1.2,1.7184831619262695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,8,power_law_1.01,0.14659199714660645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,16,power_law_1.01,0.10078719854354859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,1,256,power_law_1.2,1.2400320053100586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,16,power_law_1.01,0.10598399639129638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,1,power_law_1.2,2.0981184005737306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,8,power_law_1.01,0.06844800114631652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,16,power_law_1.01,0.11276799440383911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,1,power_law_1.2,2.857344055175781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,16,power_law_1.01,0.11790720224380494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,8,power_law_1.01,0.09548799991607666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,1,256,power_law_1.2,2.172844886779785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,16,power_law_1.01,0.11397119760513305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,1,power_law_1.2,3.6215232849121093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,16,power_law_1.01,0.11880320310592651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,8,power_law_1.01,0.11927679777145386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,16,power_law_1.01,0.12880640029907225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,8,power_law_1.01,0.13045120239257812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,16,power_law_1.01,0.12845439910888673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,1,power_law_1.2,4.386438369750977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,16,power_law_1.01,0.12833280563354493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,8,power_law_1.01,0.1466431975364685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,16,power_law_1.01,0.14725120067596437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,1,power_law_1.2,6.706137847900391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,8,power_law_1.01,0.15427199602127076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,1,256,power_law_1.2,5.0172992706298825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,16,power_law_1.01,0.16257280111312866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,16,power_law_1.01,0.18355200290679932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,8,power_law_1.01,0.15429120063781737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,16,power_law_1.01,0.1969472050666809
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,8,power_law_1.01,0.15399680137634278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,16,power_law_1.01,0.25084800720214845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,8,power_law_1.01,0.1711743950843811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,1,power_law_1.2,13.056831359863281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,16,power_law_1.01,0.2859328031539917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,16,power_law_1.01,0.3682240009307861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,8,power_law_1.01,0.17225600481033326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,16,power_law_1.01,0.4471104145050049
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,8,power_law_1.01,0.1880959987640381
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,16,power_law_1.01,0.6000127792358398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,8,power_law_1.01,0.18551039695739746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,16,power_law_1.01,0.7683519840240478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,8,power_law_1.01,0.18763519525527955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,2,power_law_1.2,0.11079039573669433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,8,power_law_1.01,0.19548799991607665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,16,power_law_1.01,0.9065792083740234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,2,power_law_1.2,0.10857599973678589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,8,power_law_1.01,0.22220799922943116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,2,power_law_1.2,0.10725120306015015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,128,balanced,0.03766933331886927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,128,balanced,0.037647999823093414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,128,balanced,0.03748266647259394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,128,balanced,0.03754133234421412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,16,power_law_1.01,1.3970303535461426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,128,balanced,0.03754133234421412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,128,balanced,0.039493332306543984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,128,balanced,0.03948266555865606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,128,balanced,0.03973866750796636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,128,balanced,0.03949866692225138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,128,balanced,0.039605334401130676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,128,balanced,0.04170133173465729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,128,balanced,0.04159999887148539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,128,balanced,0.042037333051363625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,128,balanced,0.04160533348719279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,128,balanced,0.04379733403523763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,128,balanced,0.04747200012207031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,128,balanced,0.046015997727712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,128,balanced,0.04796266555786133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,128,balanced,0.051589335004488625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,128,balanced,0.056346664826075234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,128,balanced,0.06021333237489065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,128,balanced,0.0722453345855077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,128,balanced,0.0881866713364919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,128,balanced,0.12918933232625326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,128,balanced,0.15446399648984274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,8,power_law_1.01,0.2351680040359497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,128,balanced,0.20943999290466309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,128,balanced,0.26251200834910077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,128,balanced,0.31856000423431396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,2,power_law_1.2,0.1400320053100586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,128,balanced,0.47916801770528156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,128,balanced,0.904965321222941
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,8,power_law_1.01,0.28475520610809324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,2,power_law_1.2,0.1773695945739746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,16,power_law_1.01,2.639865684509277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,8,power_law_1.01,0.2893631935119629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,2,power_law_1.2,0.2813760042190552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,8,power_law_1.01,0.347161602973938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,2,power_law_1.2,0.29367680549621583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,2,power_law_1.2,0.3099008083343506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,8,power_law_1.01,0.3890752077102661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,2,power_law_1.2,0.325548791885376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,8,power_law_1.01,0.4996479988098145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,2,power_law_1.2,0.32819199562072754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,8,power_law_1.01,0.5783103942871094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,2,power_law_1.2,0.34004480838775636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,1,2,power_law_1.01,0.036524799466133115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,8,power_law_1.01,0.8298111915588379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,1,2,power_law_1.01,0.03500800132751465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,2,power_law_1.2,0.35374720096588136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,1,2,power_law_1.01,0.043750399351119997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,1,2,power_law_1.01,0.062463998794555664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,1,2,power_law_1.01,0.07286400198936463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,2,power_law_1.2,0.36487040519714353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,8,power_law_1.2,0.09521920084953309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,1,2,power_law_1.01,0.10554879903793335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,8,power_law_1.01,1.0580608367919921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,1,2,power_law_1.01,0.10594559907913208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,8,power_law_1.2,0.16771199703216552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,1,2,power_law_1.01,0.10965759754180908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,2,power_law_1.2,0.3867199897766113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,1,2,power_law_1.01,0.11006720066070556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,8,power_law_1.2,0.07461760044097901
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,1,2,power_law_1.01,0.11221760511398315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,1,2,power_law_1.01,0.11353600025177002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,8,power_law_1.2,0.09902719855308532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,8,power_law_1.01,1.2431167602539062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,2,power_law_1.2,0.4101888179779053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,1,2,power_law_1.01,0.11553280353546143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,1,2,power_law_1.01,0.12037119865417481
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,8,power_law_1.2,0.1168768048286438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,1,2,power_law_1.01,0.12407679557800293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,8,power_law_1.2,0.1387712001800537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,1,2,power_law_1.01,0.13013119697570802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,2,power_law_1.2,0.42152957916259765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,1,2,power_law_1.01,0.13332480192184448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,8,power_law_1.2,0.19589760303497314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,8,power_law_1.01,1.7977664947509766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,1,2,power_law_1.01,0.13376640081405639
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,1,2,power_law_1.01,0.14573440551757813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,8,power_law_1.2,0.20588159561157227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,2,power_law_1.2,0.4455296039581299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,2,power_law_1.01,0.1618175983428955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,2,power_law_1.01,0.19338239431381227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,8,power_law_1.2,0.19151359796524048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,2,power_law_1.01,0.20785920619964598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,2,power_law_1.2,0.5227327823638916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,8,power_law_1.2,0.20657920837402344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,2,power_law_1.01,0.25096960067749025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,8,power_law_1.2,0.217574405670166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,2,power_law_1.01,0.2949120044708252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,8,power_law_1.01,3.4787456512451174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,2,power_law_1.2,0.5878208160400391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,2,power_law_1.01,0.39047040939331057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,8,power_law_1.2,0.22008960247039794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,2,power_law_1.01,0.4977280139923096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,8,power_law_1.2,0.22756481170654297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,2,power_law_1.2,0.7340288162231445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,2,power_law_1.01,0.6474559783935547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,8,power_law_1.2,0.23839359283447265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,2,power_law_1.01,0.7976960182189942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,8,power_law_1.2,0.25246078968048097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,2,power_law_1.2,0.836627197265625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,8,power_law_1.2,0.26338560581207277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,2,power_law_1.01,1.055014419555664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,8,power_law_1.2,0.2779968023300171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,2,power_law_1.2,1.1175423622131349
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,8,power_law_1.2,0.3230720043182373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,2,power_law_1.01,1.5399744033813476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,8,power_law_1.2,0.32027521133422854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,2,power_law_1.2,1.30381441116333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,8,power_law_1.2,0.3823040008544922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,2,power_law_1.01,2.9984128952026365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,8,power_law_1.2,0.3996864080429077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,8,power_law_1.2,0.48853120803833006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,2,power_law_1.2,1.7875072479248046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,8,power_law_1.2,0.5773568153381348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,16,power_law_1.2,0.07646080255508422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,8,power_law_1.2,0.7280576229095459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,16,power_law_1.2,0.05842559933662415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,2,power_law_1.2,2.2084287643432616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,16,power_law_1.2,0.05304319858551025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,8,power_law_1.2,0.9281344413757324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,16,power_law_1.2,0.06188160181045532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,8,power_law_1.2,1.2511936187744142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,16,power_law_1.2,0.0627135992050171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,2,power_law_1.2,3.198316764831543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,16,power_law_1.2,0.07927680015563965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,8,power_law_1.2,1.534598445892334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,16,power_law_1.2,0.07657600045204163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,16,power_law_1.2,0.07893760204315185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,8,power_law_1.2,1.874028778076172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,16,power_law_1.2,0.07682560086250305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,2,power_law_1.2,3.818502426147461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,16,power_law_1.2,0.07973120212554932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,32,1,power_law_1.2,0.047193598747253415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,16,power_law_1.2,0.081004798412323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,32,1,power_law_1.2,0.05240319967269898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,8,power_law_1.2,2.9593984603881838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,16,power_law_1.2,0.08073599934577942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,32,1,power_law_1.2,0.061964797973632815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,32,1,power_law_1.2,0.07340160012245178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,16,power_law_1.2,0.08528640270233154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,32,1,power_law_1.2,0.089273601770401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,16,power_law_1.2,0.08807680010795593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,32,1,power_law_1.2,0.12634240388870238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,2,2,power_law_1.2,4.9037120819091795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,32,1,power_law_1.2,0.13434239625930786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,16,power_law_1.2,0.09549440145492553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,32,1,power_law_1.2,0.1387712001800537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,8,power_law_1.2,5.818252944946289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,16,power_law_1.2,0.09998080134391785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,32,1,power_law_1.2,0.14442239999771117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,32,1,power_law_1.2,0.14598400592803956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,16,power_law_1.2,0.10432000160217285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,32,1,power_law_1.2,0.15892479419708253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,16,power_law_1.2,0.12769919633865356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,32,1,power_law_1.2,0.1627519965171814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,16,power_law_1.2,0.1173248052597046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,32,1,power_law_1.2,0.16565760374069213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,32,1,power_law_1.2,0.17369600534439086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,16,power_law_1.2,0.14332799911499022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,32,1,power_law_1.2,0.18252160549163818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,2,2,power_law_1.2,7.646611022949219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,16,power_law_1.2,0.16351360082626343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,32,1,power_law_1.2,0.19559040069580078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,32,1,power_law_1.2,0.20635519027709961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,16,power_law_1.2,0.2093951940536499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,32,1,power_law_1.2,0.25185279846191405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,32,1,power_law_1.2,0.28742399215698244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,32,1,power_law_1.2,0.3699903964996338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,16,power_law_1.2,0.2517375946044922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,32,1,power_law_1.2,0.4419839859008789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,32,1,power_law_1.2,0.5846208095550537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,16,power_law_1.2,0.3374272108078003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,32,1,power_law_1.2,0.7232831954956055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,16,power_law_1.2,0.4424704074859619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,32,1,power_law_1.2,0.9734080314636231
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,2,balanced,0.07377066711584727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,2,balanced,0.1197760005791982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,2,balanced,0.14850667119026184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,2,balanced,0.2472533384958903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,2,balanced,0.43891199429829914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,2,balanced,0.8327679634094238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,32,1,power_law_1.2,1.3071871757507325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,2,balanced,0.8317600091298422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,16,power_law_1.2,0.6955776214599609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,2,balanced,0.8374719619750977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,2,balanced,0.8407999674479166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,2,balanced,0.8428693612416586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,2,balanced,0.8457919756571451
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,2,balanced,0.8526399930318197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,2,balanced,0.8561493555704752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,32,1,power_law_1.2,1.7787967681884767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,2,balanced,0.8632533550262451
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,2,balanced,0.8678507010142008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,2,balanced,0.8798133532206217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,16,power_law_1.2,0.797049617767334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,2,balanced,0.8882666428883871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,2,balanced,0.9238613446553549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,2,2,power_law_1.2,15.875648498535156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,2,balanced,0.9460906982421875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,2,balanced,0.9948320388793945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,2,balanced,1.058362642923991
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,32,1,power_law_1.2,2.268441581726074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,2,balanced,1.163424015045166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,2,balanced,1.23909330368042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,16,power_law_1.2,1.1168576240539552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,2,balanced,1.5875412623087566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,2,balanced,1.6557812690734863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,2,balanced,2.488074620564779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,32,1,power_law_1.2,2.786342430114746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,2,balanced,2.737914721171061
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,2,balanced,3.75434144337972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,16,power_law_1.2,1.7809471130371093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,2,balanced,5.141786575317383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,32,1,power_law_1.2,4.156147384643555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,2,balanced,9.974954605102539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,16,power_law_1.2,3.5486080169677736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,32,1,power_law_1.2,8.134662628173828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,16,power_law_1.2,0.07374719977378845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,16,power_law_1.2,0.08605440258979798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,16,power_law_1.2,0.07919999957084656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,16,power_law_1.2,0.05013759732246399
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,16,2,balanced,0.031770666440327965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,16,2,balanced,0.03166933357715607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,16,power_law_1.2,0.05612159967422485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,16,2,balanced,0.035418666899204254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,16,2,balanced,0.033887999753157295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,16,2,balanced,0.03341866781314214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,1,balanced,0.05526400109132131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,16,2,balanced,0.03566933423280716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,16,2,balanced,0.03470933437347412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,16,2,balanced,0.03606933355331421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,1,balanced,0.0913759966691335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,16,2,balanced,0.03550933301448822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,16,2,balanced,0.035642666121323906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,1,balanced,0.10571199655532837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,16,2,balanced,0.03766933331886927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,16,2,balanced,0.03794133414824804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,1,balanced,0.1585653324921926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,16,2,balanced,0.03782933453718821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,16,2,balanced,0.04087999959786733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,16,2,balanced,0.05081599950790405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,1,balanced,0.26868265867233276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,16,2,balanced,0.05204799771308899
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,16,2,balanced,0.05566933254400889
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,1,balanced,0.2693279981613159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,16,2,balanced,0.0640533318122228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,16,2,balanced,0.06613333523273468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,1,balanced,0.2757599949836731
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,16,2,balanced,0.08337066570917766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,16,2,balanced,0.0906986693541209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,1,balanced,0.2731413245201111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,16,2,balanced,0.12220799922943115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,16,2,balanced,0.1339306632677714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,1,balanced,0.27404266595840454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,16,2,balanced,0.17531200249989828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,16,power_law_1.2,0.061478400230407716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,16,2,balanced,0.20963199933369955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,1,balanced,0.2778720060984294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,16,2,balanced,0.28836266199747723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,1,balanced,0.2788800001144409
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,16,2,balanced,0.35131200154622394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,1,balanced,0.2836479942003886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,16,2,balanced,0.43160001436869305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,1,balanced,0.2892319957415263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,16,2,balanced,0.6395306587219238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,1,balanced,0.29341866572697956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,1,balanced,0.3023573358853658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,1,balanced,0.31067200501759845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,16,2,balanced,1.2196853160858154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,1,balanced,0.3277333378791809
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,1,balanced,0.3596800168355306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,1,balanced,0.4035946528116862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,16,power_law_1.2,0.06606079936027527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,1,balanced,0.46117866039276123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,1,balanced,0.5065333445866903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,1,balanced,0.7085119883219401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,1,balanced,0.7842400074005127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,1,balanced,0.10321066776911418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,1,balanced,0.19141334295272827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,1,balanced,1.132981300354004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,1,balanced,0.2585600018501282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,1,balanced,0.4686986605326335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,1,balanced,1.3818346659342449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,1,balanced,0.8870879809061686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,16,power_law_1.2,0.07478399872779846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,1,balanced,1.099962631861369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,1,balanced,1.9786933263142903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,1,balanced,1.1046453317006428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,1,balanced,1.1054133574167888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,1,balanced,1.1124426523844402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,1,balanced,2.6042826970418296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,1,balanced,1.1134613355000813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,1,balanced,1.1229973634084065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,16,power_law_1.2,0.07582079768180847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,1,balanced,3.217616081237793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,1,balanced,1.1285706361134846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,1,balanced,1.134778658548991
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,1,balanced,1.1485919952392578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,1,balanced,1.165455977121989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,1,balanced,5.060298601786296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,1,balanced,1.174015998840332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,16,power_law_1.2,0.07993599772453308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,1,balanced,1.1937493483225505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,1,balanced,1.2365386486053467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,1,balanced,1.2722079753875732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,1,balanced,1.5415679613749187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,1,balanced,1.4486719767252605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,16,power_law_1.2,0.07969279885292054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,1,balanced,9.70034154256185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,1,balanced,2.1877973874409995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,1,balanced,1.8042613665262859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,1,balanced,2.8342933654785156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,16,power_law_1.2,0.08178560137748718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,1,balanced,2.80460262298584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,1,balanced,4.559743881225586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,16,power_law_1.2,0.08402559757232667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,1,balanced,5.165210723876953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,16,power_law_1.2,0.08994560241699219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,1,balanced,5.675050735473633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,16,power_law_1.2,0.09575039744377137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,1,balanced,9.319903691609701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,16,power_law_1.2,0.09790080189704894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,64,power_law_1.2,0.07310079932212829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,16,power_law_1.2,0.1049280047416687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,16,power_law_1.2,0.12284159660339355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,64,power_law_1.2,0.06937599778175355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,1,balanced,17.601951599121094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,16,power_law_1.2,0.132806396484375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,64,power_law_1.2,0.05728639960289002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,16,power_law_1.2,0.15119999647140503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,64,power_law_1.01,0.03608959913253784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,64,power_law_1.2,0.057580798864364624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,16,power_law_1.2,0.17233279943466187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,64,power_law_1.01,0.03152639865875244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,64,power_law_1.01,0.029523199796676634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,64,power_law_1.2,0.0637503981590271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,16,power_law_1.2,0.21235198974609376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,64,power_law_1.01,0.02953599989414215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,64,power_law_1.01,0.03056640028953552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,16,power_law_1.2,0.2587519884109497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,64,power_law_1.2,0.06766080260276794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,64,power_law_1.01,0.030726400017738343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,64,power_law_1.01,0.03036159873008728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,16,power_law_1.2,0.3510207891464233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,64,power_law_1.01,0.031302401423454286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,64,power_law_1.2,0.06793599724769592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,64,power_law_1.01,0.03061760067939758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,16,power_law_1.2,0.4374527931213379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,64,power_law_1.01,0.030931198596954347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,64,power_law_1.2,0.06651520133018493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,64,power_law_1.01,0.03152639865875244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,16,power_law_1.2,0.5848320007324219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,64,power_law_1.01,0.03208320140838623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,32,balanced,0.05462933580080668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,32,balanced,0.054671997825304665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,32,balanced,0.05560533205668131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,32,balanced,0.05677866439024607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,32,balanced,0.0611413319905599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,32,balanced,0.07050666709740956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,64,power_law_1.2,0.06875519752502442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,32,balanced,0.08480000495910645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,32,balanced,0.0846026639143626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,32,balanced,0.08509332935015361
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,32,balanced,0.08564266562461853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,32,balanced,0.08518933256467183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,32,balanced,0.08705066641171773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,64,power_law_1.01,0.034143999218940735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,32,balanced,0.08803199728329976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,32,balanced,0.09123200178146362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,32,balanced,0.09697066744168599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,32,balanced,0.0958720048268636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,32,balanced,0.10569600264231364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,32,balanced,0.11198932925860088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,32,balanced,0.12270933389663696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,32,balanced,0.14577066898345947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,32,balanced,0.16366400321324667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,32,balanced,0.20675732692082724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,32,balanced,0.2459999918937683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,32,balanced,0.3266773422559102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,32,balanced,0.4236533244450887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,16,power_law_1.2,0.7769536018371582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,32,balanced,0.5841386715571085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,64,power_law_1.2,0.06906239986419678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,32,balanced,0.7502559820810953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,64,power_law_1.01,0.03521279990673065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,32,balanced,0.9255306720733643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,32,balanced,1.4312052726745605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,32,balanced,2.8014933268229165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,64,power_law_1.01,0.03828479945659637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,64,power_law_1.2,0.06892160177230836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,16,power_law_1.2,0.9526911735534668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,64,power_law_1.01,0.039103999733924866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,64,power_law_1.2,0.07209600210189819
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,64,power_law_1.01,0.03976320028305054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,16,power_law_1.2,1.4570879936218262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,64,power_law_1.01,0.04127359986305237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,64,power_law_1.2,0.07623040080070495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,64,power_law_1.01,0.04420480132102966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,64,power_law_1.01,0.04885759949684143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,64,power_law_1.2,0.07841280102729797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,64,power_law_1.01,0.0527616024017334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,16,power_law_1.2,2.6752639770507813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,64,power_law_1.01,0.06320000290870667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,64,power_law_1.2,0.08659840226173401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,64,power_law_1.01,0.0759872019290924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,64,power_law_1.01,0.10632319450378418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,64,power_law_1.2,0.0874176025390625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,64,power_law_1.01,0.13300479650497438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,64,power_law_1.01,0.17200000286102296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,64,power_law_1.2,0.09488639831542969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,64,power_law_1.01,0.22418560981750488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,64,power_law_1.2,0.11034879684448243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,16,power_law_1.2,0.1887935996055603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,2,64,power_law_1.01,0.27640318870544434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,16,power_law_1.2,0.14910720586776732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,64,power_law_1.2,0.12275840044021606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,2,64,power_law_1.01,0.4040383815765381
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,16,power_law_1.2,0.09430400133132935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,64,power_law_1.2,0.15182080268859863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,2,64,power_law_1.01,0.7272384166717529
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,16,power_law_1.2,0.10762879848480225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,64,power_law_1.2,0.1829568028450012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,2,power_law_1.2,0.09338240027427673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,16,power_law_1.2,0.13806719779968263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,64,power_law_1.2,0.23709440231323242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,2,power_law_1.2,0.18835840225219727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,16,power_law_1.2,0.19987839460372925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,2,power_law_1.2,0.1130560040473938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,64,power_law_1.2,0.25932159423828127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,2,power_law_1.2,0.1843392014503479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,16,power_law_1.2,0.21466879844665526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,2,power_law_1.2,0.2616512060165405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,64,power_law_1.2,0.36083838939666746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,16,power_law_1.2,0.22403841018676757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,2,power_law_1.2,0.3678015947341919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,2,power_law_1.2,0.5660160064697266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,16,power_law_1.2,0.22461440563201904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,64,power_law_1.2,0.4211008071899414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,2,power_law_1.2,0.577561616897583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,2,power_law_1.2,0.6136576175689697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,16,power_law_1.2,0.22654080390930176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,2,power_law_1.2,0.6260032176971435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,64,power_law_1.2,0.6461696147918701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,16,power_law_1.2,0.2338560104370117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,8,power_law_1.01,0.05256320238113403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,2,power_law_1.2,0.6603007793426514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,2,power_law_1.2,0.6851391792297363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,8,power_law_1.01,0.06341760158538819
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,16,power_law_1.2,0.23820159435272217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,2,power_law_1.2,0.6967423915863037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,8,power_law_1.01,0.04129279851913452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,64,power_law_1.2,0.9158143997192383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,16,power_law_1.2,0.24812159538269044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,2,power_law_1.2,0.7451839923858643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,8,power_law_1.01,0.04883840084075928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,2,power_law_1.2,0.7531455993652344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,8,power_law_1.01,0.05264639854431152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,16,power_law_1.2,0.2533247947692871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,2,power_law_1.2,0.8013055801391602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,8,power_law_1.01,0.0720575988292694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,8,power_law_1.01,0.07527679800987244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,16,power_law_1.2,0.26917119026184083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,2,power_law_1.2,0.8382335662841797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,1,64,power_law_1.2,1.001420783996582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,8,power_law_1.01,0.07029759883880615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,2,power_law_1.2,0.8852928161621094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,8,power_law_1.01,0.06975359916687011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,16,power_law_1.2,0.27835519313812257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,2,power_law_1.2,0.9395839691162109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,8,power_law_1.01,0.0720575988292694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,8,power_law_1.01,0.07326719760894776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,2,power_law_1.2,1.0683839797973633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,16,power_law_1.2,0.2853503942489624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,8,power_law_1.01,0.07232000231742859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,2,power_law_1.2,1.0661696434020995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,1,64,power_law_1.2,1.631180763244629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,8,power_law_1.01,0.07300480008125305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,16,power_law_1.2,0.3385024070739746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,2,power_law_1.2,1.2159680366516112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,8,power_law_1.01,0.07747200131416321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,8,power_law_1.01,0.0815999984741211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,2,power_law_1.2,1.3898688316345216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,16,power_law_1.2,0.33802878856658936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,8,power_law_1.01,0.0927616000175476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,8,power_law_1.01,0.09345279932022095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,2,power_law_1.2,1.7633407592773438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,16,power_law_1.2,0.4327807903289795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,8,power_law_1.01,0.09678080081939697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,8,power_law_1.01,0.11745920181274414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,2,power_law_1.2,2.0030784606933594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,16,power_law_1.2,0.4416895866394043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,8,power_law_1.01,0.1270591974258423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,8,power_law_1.01,0.12059520483016968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,2,power_law_1.2,2.499225616455078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,1,64,power_law_1.2,3.497760009765625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,16,power_law_1.2,0.5764416217803955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,8,power_law_1.01,0.1447487950325012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,8,power_law_1.01,0.16188160181045533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,2,power_law_1.2,3.1154111862182616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,8,power_law_1.01,0.21802239418029784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,16,power_law_1.2,0.6478400230407715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,8,power_law_1.01,0.2380671977996826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,2,power_law_1.2,3.9531326293945312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,8,power_law_1.01,0.3199039936065674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,16,power_law_1.2,0.7684864044189453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,8,power_law_1.01,0.38959360122680664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,1,8,power_law_1.01,0.464140796661377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,2,power_law_1.2,5.361254501342773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,16,power_law_1.2,1.0359423637390137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,1,8,power_law_1.01,0.7500800132751465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,16,power_law_1.2,1.4064895629882812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,1,8,power_law_1.01,1.4255231857299804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,2,power_law_1.2,9.778508758544922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,4,8,power_law_1.2,0.03397760093212128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,4,8,power_law_1.2,0.03492479920387268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,4,8,power_law_1.2,0.033932799100875856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,4,8,power_law_1.2,0.03386240005493164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,16,power_law_1.2,1.6456127166748047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,4,8,power_law_1.2,0.034643200039863584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,4,8,power_law_1.2,0.03537279963493347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,4,8,power_law_1.2,0.03804160058498383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,4,8,power_law_1.2,0.04291839897632599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,4,8,power_law_1.2,0.04602240025997162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,4,8,power_law_1.2,0.04689919948577881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,4,8,power_law_1.2,0.045926401019096376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,4,8,power_law_1.2,0.05367040038108826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,4,8,power_law_1.2,0.05422080159187317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,4,8,power_law_1.2,0.05849599838256836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,4,8,power_law_1.2,0.06920959949493408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,4,8,power_law_1.2,0.07151359915733338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,4,8,power_law_1.2,0.08241279721260071
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,4,8,power_law_1.2,0.10238720178604126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,4,8,power_law_1.2,0.10395519733428955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,4,8,power_law_1.2,0.1334272027015686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,4,8,power_law_1.2,0.1654144048690796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,1,16,power_law_1.2,2.01146240234375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,4,8,power_law_1.2,0.23384320735931396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,4,8,power_law_1.2,0.2898175954818726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,4,8,power_law_1.2,0.379750394821167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,4,8,power_law_1.2,0.48932480812072754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,4,8,power_law_1.2,0.7425024032592773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,4,8,power_law_1.2,0.9565312385559082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,4,8,power_law_1.2,1.3401023864746093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,4,8,power_law_1.2,1.827788734436035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,4,8,power_law_1.2,3.519424057006836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,1,16,power_law_1.2,3.1822080612182617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,4,power_law_1.01,0.03588480055332184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,8,power_law_1.2,0.08798720240592957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,4,power_law_1.01,0.03315840065479279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,8,power_law_1.2,0.08478080034255982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,8,power_law_1.2,0.0706496000289917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,4,power_law_1.01,0.03164800107479095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,8,power_law_1.2,0.0789247989654541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,4,power_law_1.01,0.036345601081848145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,8,power_law_1.2,0.08655999898910523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,8,power_law_1.2,0.114028799533844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,4,power_law_1.01,0.03765760064125061
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,8,power_law_1.2,0.12482559680938721
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,4,power_law_1.01,0.051046401262283325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,8,power_law_1.2,0.12363519668579101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,8,power_law_1.2,0.12460160255432129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,4,power_law_1.01,0.0514303982257843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,8,power_law_1.2,0.13063679933547973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,1,16,power_law_1.2,6.55889892578125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,4,power_law_1.01,0.05082240104675293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,8,power_law_1.2,0.1302783966064453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,4,power_law_1.01,0.05252479910850525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,8,power_law_1.2,0.1335039973258972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,4,power_law_1.01,0.05355520248413086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,8,power_law_1.2,0.13501440286636351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,8,power_law_1.2,0.14179199934005737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,4,power_law_1.01,0.05516800284385681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,1,balanced,0.03756266583998998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,1,balanced,0.04980266590913137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,1,balanced,0.06985599795977275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,1,balanced,0.10444266597429912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,1,balanced,0.17086400588353476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,8,power_law_1.2,0.14972800016403198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,1,balanced,0.21836266915003458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,4,power_law_1.01,0.05658239722251892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,1,balanced,0.22276800870895386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,1,balanced,0.22150399287541708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,1,balanced,0.22405334313710532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,1,balanced,0.22317866484324136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,1,balanced,0.22678399085998535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,1,balanced,0.22876266638437906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,1,balanced,0.22985066970189413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,1,balanced,0.23358400662740073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,8,power_law_1.2,0.1564352035522461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,1,balanced,0.23690134286880493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,1,balanced,0.24820800622304282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,1,balanced,0.24715199073155722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,4,power_law_1.01,0.057055997848510745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,1,balanced,0.2887893319129944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,1,balanced,0.2678133249282837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,1,balanced,0.3193333347638448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,1,balanced,0.3092106580734253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,1,balanced,0.42496001720428467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,8,power_law_1.2,0.16128640174865722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,1,balanced,0.41819198926289874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,1,balanced,0.5864319801330566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,4,power_law_1.01,0.05839359760284424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,1,balanced,0.6534773508707682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,1,balanced,0.9524373213450114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,8,power_law_1.2,0.18171520233154298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,1,balanced,1.103434642155965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,4,power_law_1.01,0.06731520295143127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,1,1,balanced,1.3862293561299641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,8,power_law_1.2,0.20161280632019044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,1,1,balanced,2.1102986335754395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,4,power_law_1.01,0.06688640117645264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,8,power_law_1.2,0.25280640125274656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,1,1,balanced,4.037173271179199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,4,power_law_1.01,0.06914560198783874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,8,power_law_1.2,0.25973119735717776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,4,power_law_1.01,0.07203199863433837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,8,power_law_1.2,0.3231040000915527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,4,power_law_1.01,0.07648000121116638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,8,power_law_1.2,0.4124351978302002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,4,power_law_1.01,0.08202239871025085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,8,power_law_1.2,0.5736447811126709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,4,power_law_1.01,0.09657599925994872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,8,power_law_1.2,0.6670591831207275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,4,power_law_1.01,0.11337599754333497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,2,power_law_1.01,0.07575039863586426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,4,power_law_1.01,0.14347519874572753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,8,power_law_1.2,1.0139712333679198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,2,power_law_1.01,0.10782719850540161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,4,power_law_1.01,0.16907520294189454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,2,power_law_1.01,0.0868224024772644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,8,power_law_1.2,1.3385791778564453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,2,power_law_1.01,0.12401280403137208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,4,power_law_1.01,0.21086719036102294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,2,power_law_1.01,0.1550336003303528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,4,power_law_1.01,0.27204480171203616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,2,power_law_1.01,0.24686079025268554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,8,power_law_1.2,1.7049152374267578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,2,power_law_1.01,0.26481919288635253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,4,power_law_1.01,0.347871994972229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,2,power_law_1.01,0.271724796295166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,2,power_law_1.01,0.27100160121917727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,8,power_law_1.2,2.288934326171875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,4,4,power_law_1.01,0.4261375904083252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,2,power_law_1.01,0.28071041107177735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,2,power_law_1.01,0.2857280015945435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,4,4,power_law_1.01,0.6185984134674072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,2,power_law_1.01,0.2949376106262207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,2,power_law_1.01,0.3026560068130493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,2,power_law_1.01,0.12183680534362792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,2,power_law_1.01,0.31763200759887694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,8,power_law_1.2,5.972588729858399
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,2,power_law_1.01,0.3247936010360718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,4,4,power_law_1.01,1.247865581512451
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,2,power_law_1.01,0.33644800186157225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,2,power_law_1.01,0.35038080215454104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,2,power_law_1.01,0.18991999626159667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,2,power_law_1.01,0.3858304023742676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,2,power_law_1.01,0.4072127819061279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,2,power_law_1.01,0.14380799531936644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,2,power_law_1.01,0.472211217880249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,2,power_law_1.01,0.5289408206939697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,2,power_law_1.01,0.6511360168457031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,2,power_law_1.01,0.2590912103652954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,2,power_law_1.01,0.758847999572754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,2,power_law_1.01,0.3664576053619385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,2,power_law_1.01,0.9493887901306153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,2,power_law_1.01,1.0714624404907227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,2,power_law_1.01,0.5988224029541016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,2,power_law_1.01,1.5746432304382325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,2,power_law_1.01,0.6474368095397949
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,2,power_law_1.01,1.9315519332885742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,2,balanced,0.03536533315976461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,2,balanced,0.035599999129772186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,2,balanced,0.054933334390322365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,2,balanced,0.07619200150171916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,2,balanced,0.12566933035850525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,2,balanced,0.12211733063062032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,2,balanced,0.12418666481971741
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,2,balanced,0.12313066919644673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,2,balanced,0.12307199835777283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,2,balanced,0.1263146698474884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,2,balanced,0.1277653376261393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,2,balanced,0.12959999839464822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,2,balanced,0.13090133666992188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,2,balanced,0.13285866379737854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,2,balanced,0.13912000258763632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,2,balanced,0.14084800084431967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,2,balanced,0.15147733688354492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,2,balanced,0.16756800810496011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,2,power_law_1.01,0.6860479831695556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,2,balanced,0.17266666889190674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,2,balanced,0.20244799057642618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,2,balanced,0.22341867287953696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,2,balanced,0.2822986642519633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,2,power_law_1.01,2.4421951293945314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,2,balanced,0.3227733373641968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,16,power_law_1.2,0.03746559917926788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,2,balanced,0.4441973368326823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,2,balanced,0.5520639816919962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,2,balanced,0.7634133497873942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,2,balanced,0.9912799994150797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,2,balanced,1.2026453018188477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,2,balanced,1.7951092720031738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,2,power_law_1.01,0.6980288028717041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,16,power_law_1.2,0.032467201352119446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,2,balanced,3.462992032368978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,2,power_law_1.01,3.602764892578125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,16,power_law_1.2,0.030662399530410767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,2,power_law_1.01,0.7259456157684326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,16,power_law_1.2,0.03214080035686493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,16,power_law_1.2,0.03380480110645294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,2,power_law_1.01,0.7402048110961914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,16,power_law_1.2,0.03594880104064942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,2,power_law_1.01,6.91923828125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,16,power_law_1.2,0.03596799969673157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,2,power_law_1.01,0.7665791988372803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,16,power_law_1.2,0.03521920144557953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,16,power_law_1.2,0.034969601035118106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,2,power_law_1.01,0.7831679821014405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,16,power_law_1.2,0.03522560000419617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,16,power_law_1.2,0.039628800749778745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,2,power_law_1.01,0.7994688034057618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,16,power_law_1.2,0.04023039937019348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,16,power_law_1.2,0.041203200817108154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,2,power_law_1.01,0.8330495834350586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,16,power_law_1.2,0.04464640021324158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,16,power_law_1.2,0.050425601005554196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,2,power_law_1.01,0.8367744445800781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,16,power_law_1.2,0.05592319965362549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,256,power_law_1.2,0.056601601839065555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,16,power_law_1.2,0.05456640124320984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,16,power_law_1.2,0.060915201902389526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,256,power_law_1.2,0.058156800270080564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,2,power_law_1.01,0.9138303756713867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,16,power_law_1.2,0.06949120163917541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,256,power_law_1.2,0.05739520192146301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,16,power_law_1.2,0.07948799729347229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,256,power_law_1.2,0.056985598802566526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,1,balanced,0.037647999823093414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,1,balanced,0.03967999915281931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,2,power_law_1.01,1.0266431808471679
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,1,balanced,0.045781334241231285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,1,balanced,0.06834666430950165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,1,balanced,0.09602133433024089
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,1,balanced,0.10064533352851868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,1,balanced,0.1020853320757548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,1,balanced,0.101583997408549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,1,balanced,0.10454932848612468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,16,power_law_1.2,0.09104639887809754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,1,balanced,0.10517866412798564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,1,balanced,0.1092693308989207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,1,balanced,0.1128053367137909
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,1,balanced,0.11737066507339478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,1,balanced,0.12361066540082295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,1,balanced,0.13199466466903687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,1,balanced,0.13756266236305237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,256,power_law_1.2,0.05639680027961731
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,1,balanced,0.14498133460680643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,1,balanced,0.17520000537236533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,1,balanced,0.1902986764907837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,1,balanced,0.27033599217732746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,1,balanced,0.2918933431307475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,1,balanced,0.4337013165156047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,16,power_law_1.2,0.1205183982849121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,1,balanced,0.5209493239720663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,1,balanced,0.7636746565500895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,256,power_law_1.2,0.05878400206565857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,1,balanced,0.9606239795684814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,2,power_law_1.01,1.0080063819885254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,1,balanced,1.4172479311625164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,16,power_law_1.2,0.14328960180282593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,1,balanced,1.862837314605713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,256,power_law_1.2,0.057529598474502563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,1,balanced,2.311728000640869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,16,power_law_1.2,0.19756799936294556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,1,balanced,3.591343879699707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,256,power_law_1.2,0.05858560204505921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,2,power_law_1.01,1.1831040382385254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,16,power_law_1.2,0.26264960765838624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,256,power_law_1.2,0.05847679972648621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,1,balanced,7.003231684366862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,16,power_law_1.2,0.34621438980102537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,256,power_law_1.2,0.058931201696395874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,2,power_law_1.01,1.2665727615356446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,256,power_law_1.2,0.06045439839363098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,16,power_law_1.2,0.4443391799926758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,256,power_law_1.2,0.060659199953079224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,256,power_law_1.2,0.06115840077400207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,16,power_law_1.2,0.639353609085083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,2,power_law_1.01,1.513817596435547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,256,power_law_1.2,0.061292797327041626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,256,power_law_1.2,0.0661952018737793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,16,power_law_1.2,1.1329407691955566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,2,power_law_1.01,1.5997504234313964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,256,power_law_1.2,0.06595839858055115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,256,power_law_1.2,0.07205119729042053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,256,power_law_1.2,0.07714560031890869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,16,power_law_1.2,2.311417579650879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,2,power_law_1.01,1.9499391555786132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,256,power_law_1.2,0.09010559916496277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,256,power_law_1.2,0.11455999612808228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,256,power_law_1.2,0.13692159652709962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,16,8,power_law_1.2,0.03165439963340759
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,2,power_law_1.01,2.4325632095336913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,16,8,power_law_1.2,0.03128960132598877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,16,8,power_law_1.2,0.031014400720596313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,16,8,power_law_1.2,0.030265599489212036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,256,power_law_1.2,0.17679359912872314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,16,8,power_law_1.2,0.030847999453544616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,16,8,power_law_1.2,0.031839999556541446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,16,8,power_law_1.2,0.03240320086479187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,16,8,power_law_1.2,0.033497598767280576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,16,8,power_law_1.2,0.03425279855728149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,256,power_law_1.2,0.22157440185546876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,16,8,power_law_1.2,0.03473280072212219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,16,8,power_law_1.2,0.03550719916820526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,16,8,power_law_1.2,0.036582401394844054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,16,8,power_law_1.2,0.03709439933300018
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,16,8,power_law_1.2,0.03856000006198883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,16,8,power_law_1.2,0.049644801020622256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,256,power_law_1.2,0.2977855920791626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,16,8,power_law_1.2,0.04976640045642853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,16,8,power_law_1.2,0.05295360088348389
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,2,power_law_1.01,3.2974334716796876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,16,8,power_law_1.2,0.06284160017967225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,16,8,power_law_1.2,0.06636800169944763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,256,power_law_1.2,0.38279678821563723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,16,8,power_law_1.2,0.08837760090827942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,16,8,power_law_1.2,0.11253119707107544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,16,8,power_law_1.2,0.16023679971694946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,16,8,power_law_1.2,0.18761600255966188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,16,8,power_law_1.2,0.26294400691986086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,256,power_law_1.2,0.609548807144165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,16,8,power_law_1.2,0.3429631948471069
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,16,8,power_law_1.2,0.5029248237609864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,16,8,power_law_1.2,0.6203392028808594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,16,8,power_law_1.2,0.7668543815612793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,256,power_law_1.2,0.7561984062194824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,2,power_law_1.01,3.802598571777344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,16,8,power_law_1.2,1.202291202545166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,2,4,balanced,0.04952000081539154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,2,4,balanced,0.0430026650428772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,256,power_law_1.2,0.9662015914916993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,16,8,power_law_1.2,2.1753664016723633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,2,4,balanced,0.0513973335425059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,2,4,balanced,0.04948799808820089
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,2,4,balanced,0.05109866460164388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,2,4,balanced,0.0513866643110911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,2,4,balanced,0.05167999863624573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,2,4,balanced,0.05156266689300537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,2,4,balanced,0.053247998158137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,2,4,balanced,0.05264000097910563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,2,4,balanced,0.053685332338015236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,2,4,balanced,0.05478399991989136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,2,4,balanced,0.055439998706181846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,2,4,balanced,0.05996266504128774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,2,4,balanced,0.06512533128261566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,2,4,balanced,0.0682826687892278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,2,4,balanced,0.07311999797821045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,2,4,balanced,0.07932266592979431
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,2,4,balanced,0.08238933483759563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,2,4,balanced,0.0993386705716451
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,2,4,balanced,0.11181333661079407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,2,4,balanced,0.13885333140691122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,2,4,balanced,0.15836800138155618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,2,4,balanced,0.21346133947372437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,256,power_law_1.2,1.7054079055786133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,2,4,balanced,0.2526986598968506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,2,4,balanced,0.3508053223292033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,2,4,balanced,0.45554133256276447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,2,4,balanced,0.5590506792068481
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,2,4,balanced,0.8523253599802653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,1,2,power_law_1.01,4.625132751464844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,2,4,balanced,1.6339252789815266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,256,power_law_1.2,3.3883712768554686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,1,2,power_law_1.01,7.0292610168457035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,1,power_law_1.2,0.13772799968719482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,1,power_law_1.2,0.1390079975128174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,16,power_law_1.2,0.05516160130500793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,1,power_law_1.2,0.16926079988479614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,2,1,balanced,0.0672106643517812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,2,1,balanced,0.12220266461372375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,2,1,balanced,0.1423733333746592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,2,1,balanced,0.1400266687075297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,2,1,balanced,0.14442666371663412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,2,1,balanced,0.14461867014567056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,2,1,balanced,0.14441066980361938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,2,1,balanced,0.14620799819628397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,2,1,balanced,0.14497599999109903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,2,1,balanced,0.1476426621278127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,2,1,balanced,0.1495626668135325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,2,1,balanced,0.15060266852378845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,2,1,balanced,0.15312000115712485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,2,1,balanced,0.1588159998257955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,2,1,balanced,0.17337600390116373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,2,1,balanced,0.17520000537236533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,2,1,balanced,0.17891732851664224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,2,1,balanced,0.22217599550882974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,2,1,balanced,0.2332693338394165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,2,1,balanced,0.2905386686325073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,1,power_law_1.2,0.22445440292358398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,2,1,balanced,0.35281066099802655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,2,1,balanced,0.4723626772562663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,1,2,power_law_1.01,13.475053405761718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,16,power_law_1.2,0.08080000281333924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,2,1,balanced,0.5730186700820923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,2,1,balanced,0.8072426319122314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,2,1,balanced,1.031551996866862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,2,1,balanced,1.4805973370869954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,2,1,balanced,1.9565653800964355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,1,power_law_1.2,0.2885119915008545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,2,1,balanced,2.4053492546081543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,16,power_law_1.2,0.050419199466705325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,2,1,balanced,3.7617066701253257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,1,power_law_1.2,0.4813119888305664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,16,power_law_1.2,0.05123839974403381
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,2,1,balanced,7.567056020100911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,1,power_law_1.2,0.5359488010406495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,16,power_law_1.2,0.0541055977344513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,1,power_law_1.2,0.5623167991638184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,16,power_law_1.2,0.06139519810676575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,16,power_law_1.2,0.06926079988479614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,1,power_law_1.2,0.5883584022521973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,16,power_law_1.2,0.06839680075645446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,1,power_law_1.2,0.5986368179321289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,16,power_law_1.2,0.073471999168396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,1,power_law_1.2,0.631660795211792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,32,4,balanced,0.0481279989083608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,32,4,balanced,0.05035200218359629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,32,4,balanced,0.05227200190226237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,32,4,balanced,0.054474666714668274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,32,4,balanced,0.06286933521429698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,32,4,balanced,0.09205866853396098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,32,4,balanced,0.1155413289864858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,32,4,balanced,0.11829333504041036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,32,4,balanced,0.11788800358772278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,32,4,balanced,0.11938666303952535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,32,4,balanced,0.11844266454378764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,16,power_law_1.2,0.07245439887046815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,32,4,balanced,0.12160533666610718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,32,4,balanced,0.12244799733161926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,32,4,balanced,0.12485333283742268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,32,4,balanced,0.13035200039545694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,32,4,balanced,0.1334719955921173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,32,4,balanced,0.13636266191800436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,32,4,balanced,0.14889066418011984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,32,4,balanced,0.16130666931470236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,32,4,balanced,0.18230400482813516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,32,4,balanced,0.20465066035588583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,32,4,balanced,0.2571626702944438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,1,power_law_1.2,0.6565440177917481
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,32,4,balanced,0.32098132371902466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,32,4,balanced,0.440341313680013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,32,4,balanced,0.544053316116333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,32,4,balanced,0.7944213549296061
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,16,power_law_1.2,0.07609599828720093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,32,4,balanced,1.010159969329834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,32,4,balanced,1.2608160177866619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,32,4,balanced,1.9119359652201335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,1,power_law_1.2,0.6805568218231202
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,16,power_law_1.2,0.07945600152015686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,32,4,balanced,3.6773866017659507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,16,power_law_1.2,0.07992960214614868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,1,power_law_1.2,0.7194240093231201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,256,power_law_1.01,0.0722815990447998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,16,power_law_1.2,0.08409600257873535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,1,power_law_1.2,0.7416768074035645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,16,power_law_1.2,0.09449599981307984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,256,power_law_1.01,0.07681919932365418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,16,power_law_1.2,0.09868159890174866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,1,power_law_1.2,0.7811647891998291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,256,power_law_1.01,0.06951680183410644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,16,power_law_1.2,0.10163840055465698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,1,power_law_1.2,0.8305279731750488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,256,power_law_1.01,0.07411199808120728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,16,power_law_1.2,0.11598720550537109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,16,power_law_1.2,0.12338559627532959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,1,power_law_1.2,0.9575296401977539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,256,power_law_1.01,0.07502719759941101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,16,power_law_1.2,0.15179519653320311
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,256,power_law_1.01,0.07367680072784424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,1,power_law_1.2,1.0843520164489746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,8,balanced,0.05230933427810669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,8,balanced,0.05395199855168661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,8,balanced,0.05418133238951365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,8,balanced,0.06309333443641663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,8,balanced,0.08737599849700928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,16,power_law_1.2,0.196396803855896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,8,balanced,0.11766933401425679
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,8,balanced,0.12100799878438313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,8,balanced,0.12193600336710612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,8,balanced,0.12286399801572163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,8,balanced,0.123471995194753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,8,balanced,0.1230506698290507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,8,balanced,0.12448533376057942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,8,balanced,0.12587733070055643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,8,balanced,0.1269973317782084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,8,balanced,0.13590932885805765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,256,power_law_1.01,0.07498239874839782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,8,balanced,0.13500799735387167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,8,balanced,0.14010666807492575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,8,balanced,0.15212266643842062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,8,balanced,0.16034133235613504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,8,balanced,0.1810879906018575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,8,balanced,0.20125865936279297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,1,power_law_1.2,1.318342399597168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,8,balanced,0.24131200710932413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,8,balanced,0.27348266045252484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,16,power_law_1.2,0.2502016067504883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,8,balanced,0.3624906539916992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,8,balanced,0.42550933361053467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,8,balanced,0.6016373236974081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,8,balanced,0.7486933072408041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,8,balanced,0.9209280014038086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,256,power_law_1.01,0.07597439885139465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,8,balanced,1.395850658416748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,16,power_law_1.2,0.2819008111953735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,8,balanced,2.702389399210612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,1,power_law_1.2,1.4950783729553223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,256,power_law_1.01,0.0769919991493225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,16,power_law_1.2,0.384224009513855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,1,power_law_1.2,1.93768310546875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,256,power_law_1.01,0.07447680234909057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,16,power_law_1.2,0.5162559986114502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,256,power_law_1.01,0.07487360239028931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,16,power_law_1.2,0.748799991607666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,1,power_law_1.2,2.3650623321533204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,256,power_law_1.01,0.07724159955978394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,16,power_law_1.2,0.9502335548400879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,256,power_law_1.01,0.08064640164375306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,1,power_law_1.2,3.2115009307861326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,16,power_law_1.2,1.1334464073181152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,256,power_law_1.01,0.0763647973537445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,256,power_law_1.01,0.08083840012550354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,1,power_law_1.2,4.094732666015625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,16,power_law_1.2,1.8260992050170899
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,2,balanced,0.06817600131034851
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,2,balanced,0.07230400045712788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,2,balanced,0.0766133318344752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,2,balanced,0.09051733215649922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,2,balanced,0.12945600350697836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,2,balanced,0.19185600678126016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,2,balanced,0.24823999404907227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,256,power_law_1.01,0.08130559921264649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,2,balanced,0.24024534225463867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,2,balanced,0.24463466803232828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,2,balanced,0.245194673538208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,2,balanced,0.24708799521128336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,2,balanced,0.2485546668370565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,2,balanced,0.2510613401730855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,2,balanced,0.2539466619491577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,2,balanced,0.26340266068776447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,2,balanced,0.266154666741689
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,2,balanced,0.2779093384742737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,2,balanced,0.2990986704826355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,2,balanced,0.31918932994206745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,2,balanced,0.3620479901631673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,2,balanced,0.4001599947611491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,2,balanced,0.4931360085805257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,2,balanced,0.5814133485158285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,256,power_law_1.01,0.08846719861030579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,2,balanced,0.7725973129272461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,2,balanced,1.0232799847920735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,2,balanced,1.3829973538716633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,1,power_law_1.2,5.858329772949219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,2,balanced,1.8586293856302898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,2,balanced,2.297173341115316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,16,power_law_1.2,4.035744094848633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,256,power_law_1.01,0.10618239641189575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,2,balanced,3.5221227010091147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,2,balanced,6.69758415222168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,256,power_law_1.01,0.11947519779205322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,256,power_law_1.01,0.14226560592651366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,1,power_law_1.2,7.2653053283691404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,256,power_law_1.01,0.17378560304641724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,256,power_law_1.01,0.22403199672698976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,2,1,power_law_1.2,8.875257873535157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,256,power_law_1.01,0.24769279956817628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,256,power_law_1.01,0.2975231885910034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,32,1,power_law_1.01,0.04784640073776245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,16,balanced,0.054154664278030396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,16,balanced,0.05006400247414907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,16,balanced,0.05198933184146881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,16,balanced,0.07062933345635732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,16,balanced,0.09610666831334432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,32,1,power_law_1.01,0.0531328022480011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,16,balanced,0.14948800206184387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,16,balanced,0.14824000000953674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,16,balanced,0.14920000235239664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,16,balanced,0.14920000235239664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,16,balanced,0.14995200435320535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,16,balanced,0.1495413382848104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,16,balanced,0.15289599696795145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,16,balanced,0.15296533703804016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,16,balanced,0.15454933047294617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,256,power_law_1.01,0.40074877738952636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,16,balanced,0.15731199582417807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,16,balanced,0.16012799739837646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,16,balanced,0.16109866897265115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,16,balanced,0.16821332772572836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,16,balanced,0.17524800697962442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,16,balanced,0.18627200524012247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,16,balanced,0.19896533091862997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,16,balanced,0.2223680019378662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,16,balanced,0.2432159980138143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,16,balanced,0.3072906732559204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,16,balanced,0.34330133597056073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,16,balanced,0.48162134488423664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,16,balanced,0.5507093270619711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,32,1,power_law_1.01,0.06330879926681518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,16,balanced,0.7026879787445068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,16,balanced,0.9665280183156332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,16,balanced,1.8732320467631023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,2,1,power_law_1.2,13.976914978027343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,32,1,power_law_1.01,0.07988479733467102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,256,power_law_1.01,0.5528639793395996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,32,1,power_law_1.01,0.0937279999256134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,32,1,power_law_1.01,0.12887040376663209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,256,power_law_1.01,0.7086719989776611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,32,1,power_law_1.01,0.13544319868087767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,32,1,power_law_1.01,0.1408128023147583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,32,1,power_law_1.01,0.14316799640655517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,1,256,power_law_1.01,0.8491071701049805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,32,1,power_law_1.01,0.14497920274734497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,32,1,power_law_1.01,0.15032320022583007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,32,1,power_law_1.01,0.1567744016647339
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,32,1,power_law_1.01,0.16225279569625856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,32,1,power_law_1.01,0.169977605342865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,1,256,power_law_1.01,1.4036543846130372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,4,power_law_1.2,0.05625600218772888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,32,1,power_law_1.01,0.17927680015563965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,2,1,power_law_1.2,26.606509399414062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,4,power_law_1.2,0.05905280113220215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,4,power_law_1.2,0.05472000241279602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,32,1,power_law_1.01,0.18625919818878173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,4,power_law_1.2,0.06740480065345764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,32,1,power_law_1.01,0.20320639610290528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,4,power_law_1.2,0.07630720138549804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,4,power_law_1.2,0.09952639937400817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,32,1,power_law_1.01,0.24535679817199707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,4,power_law_1.2,0.1079103946685791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,32,1,power_law_1.01,0.28409600257873535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,4,power_law_1.2,0.10687359571456909
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,4,power_law_1.2,0.11316479444503784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,32,1,power_law_1.01,0.3619584083557129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,4,power_law_1.2,0.11633280515670777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,32,1,power_law_1.01,0.43566079139709474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,4,power_law_1.2,0.1158911943435669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,1,256,power_law_1.01,2.927359962463379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,4,power_law_1.2,0.11831680536270142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,32,1,power_law_1.01,0.5820543766021729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,4,power_law_1.2,0.12135679721832275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,32,1,power_law_1.01,0.706982421875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,4,power_law_1.2,0.12779519557952881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,4,power_law_1.2,0.13667839765548706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,32,1,power_law_1.01,0.9652095794677734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,4,power_law_1.2,0.1395840048789978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,4,power_law_1.2,0.15092480182647705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,32,1,power_law_1.01,1.2755776405334474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,4,power_law_1.2,0.16898560523986816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,4,power_law_1.2,0.17793920040130615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,2,8,power_law_1.2,0.036601600050926206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,2,8,power_law_1.2,0.03683840036392212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,32,1,power_law_1.01,1.7774976730346679
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,2,8,power_law_1.2,0.036774399876594546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,2,8,power_law_1.2,0.036185601353645326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,4,power_law_1.2,0.21629440784454346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,2,8,power_law_1.2,0.03642880022525787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,2,8,power_law_1.2,0.04359039962291718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,2,8,power_law_1.2,0.04832000136375427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,4,power_law_1.2,0.2510848045349121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,2,8,power_law_1.2,0.04922879934310913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,2,8,power_law_1.2,0.05248640179634094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,2,8,power_law_1.2,0.05381760001182556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,2,8,power_law_1.2,0.053260797262191774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,32,1,power_law_1.01,2.252390480041504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,4,power_law_1.2,0.32868480682373047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,2,8,power_law_1.2,0.061350399255752565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,2,8,power_law_1.2,0.0616320013999939
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,2,8,power_law_1.2,0.06272000074386597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,2,8,power_law_1.2,0.07256960272789001
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,2,8,power_law_1.2,0.08074880242347718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,4,power_law_1.2,0.37511041164398196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,2,8,power_law_1.2,0.0812928020954132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,2,8,power_law_1.2,0.10157439708709717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,2,8,power_law_1.2,0.11581439971923828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,2,8,power_law_1.2,0.1445248007774353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,4,power_law_1.2,0.5742784023284913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,2,8,power_law_1.2,0.18673280477523804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,32,1,power_law_1.01,2.734022331237793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,2,8,power_law_1.2,0.24623360633850097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,2,8,power_law_1.2,0.29018239974975585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,2,8,power_law_1.2,0.4362495899200439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,4,power_law_1.2,0.6874239921569825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,2,8,power_law_1.2,0.5175551891326904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,2,8,power_law_1.2,0.8237440109252929
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,4,power_law_1.2,0.9248000144958496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,2,8,power_law_1.2,1.1567680358886718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,32,1,power_law_1.01,4.143199920654297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,2,8,power_law_1.2,1.4479616165161133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,32,4,power_law_1.01,0.05604479908943176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,4,power_law_1.2,1.1742976188659668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,2,8,power_law_1.2,2.0527679443359377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,32,4,power_law_1.01,0.06219519972801209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,4,power_law_1.2,1.5053440093994142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,2,8,power_law_1.2,4.021273422241211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,32,4,power_law_1.01,0.05851519703865051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,32,4,power_law_1.01,0.06156799793243408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,4,power_law_1.2,2.2954111099243164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,32,4,power_law_1.01,0.0695039987564087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,32,1,power_law_1.01,8.120294189453125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,32,4,power_law_1.01,0.0757311999797821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,32,4,power_law_1.01,0.08455680012702942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,32,4,power_law_1.01,0.0892416000366211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,4,power_law_1.2,4.596646499633789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,32,4,power_law_1.01,0.0896448016166687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,32,4,power_law_1.01,0.09043840169906617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,32,4,power_law_1.01,0.09267200231552124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,32,4,power_law_1.01,0.09537280201911927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,32,4,power_law_1.01,0.09896960258483886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,32,4,power_law_1.01,0.10217599868774414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,32,4,power_law_1.01,0.1114240050315857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,32,4,power_law_1.01,0.11434240341186523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,32,4,power_law_1.01,0.1212224006652832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,4,balanced,0.04722133278846741
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,4,balanced,0.049072002371152244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,4,balanced,0.04956266780694326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,4,balanced,0.055231998364130654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,4,balanced,0.0631039987007777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,4,balanced,0.09118933478991191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,32,4,power_law_1.01,0.13645440340042114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,4,balanced,0.09551999966303508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,4,balanced,0.09473066528638203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,4,balanced,0.09827733039855957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,4,balanced,0.09807999928792317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,4,balanced,0.10125333070755005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,4,balanced,0.10222400228182475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,4,balanced,0.10526933272679646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,4,balanced,0.10579199592272441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,4,balanced,0.11664533615112305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,4,balanced,0.12041067083676656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,32,4,power_law_1.01,0.15648640394210817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,4,balanced,0.12365333239237468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,4,balanced,0.14096533258756003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,4,balanced,0.1516586641470591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,4,balanced,0.17890665928522745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,4,balanced,0.20705066124598184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,4,balanced,0.27621867259343463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,4,balanced,0.3255946636199951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,32,4,power_law_1.01,0.18471039533615113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,4,balanced,0.4698293209075928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,4,balanced,0.5811466773351034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,4,balanced,0.8480853239695231
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,8,power_law_1.2,0.08680319786071777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,4,balanced,1.1034666697184246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,32,4,power_law_1.01,0.21156480312347412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,4,balanced,1.3544960021972656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,1,balanced,0.09795199831326802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,1,balanced,0.19167999426523843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,4,balanced,2.101322650909424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,1,balanced,0.242031991481781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,1,balanced,0.44016532103220624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,1,balanced,0.8251146475474039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,32,4,power_law_1.01,0.2819839954376221
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,1,balanced,1.6016480127970378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,8,power_law_1.2,0.07618560194969178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,4,balanced,4.135295867919922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,1,balanced,1.6069013277689617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,1,balanced,1.6117653846740723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,1,balanced,1.6155519485473633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,32,4,power_law_1.01,0.34108800888061525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,1,balanced,1.6200000445048015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,1,balanced,1.6199520428975422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,8,power_law_1.2,0.06938880085945129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,1,balanced,1.6285600662231445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,32,4,power_law_1.01,0.48609280586242676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,1,balanced,1.6366292635599773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,1,balanced,1.648896058400472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,1,balanced,1.6595093409220378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,1,balanced,1.6704799334208171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,1,balanced,1.6904106140136719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,32,4,power_law_1.01,0.584332799911499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,8,power_law_1.2,0.07835519909858704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,1,balanced,1.73636261622111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,1,balanced,1.7961546579996746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,1,balanced,1.8722186088562012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,1,balanced,1.953445275624593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,32,4,power_law_1.01,0.9021823883056641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,1,balanced,2.0936106046040854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,8,power_law_1.2,0.0959231972694397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,1,balanced,2.3066399892171225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,1,balanced,2.5721279780069985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,32,4,power_law_1.01,1.1408703804016114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,1,balanced,2.8320531845092773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,8,power_law_1.2,0.13068159818649291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,1,balanced,4.35098139444987
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,32,4,power_law_1.01,1.4312640190124513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,1,balanced,4.549434661865234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,8,power_law_1.2,0.13188480138778685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,1,balanced,6.497445424397786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,8,power_law_1.2,0.14122240543365477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,32,4,power_law_1.01,2.1906816482543947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,1,balanced,8.411717096964518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,4,balanced,0.045754666129748024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,4,balanced,0.047194664676984154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,8,power_law_1.2,0.14494080543518068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,4,balanced,0.050810664892196655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,4,balanced,0.05457599957784017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,4,balanced,0.06890666484832764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,4,balanced,0.07723199824492137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,4,balanced,0.08136000235875447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,4,balanced,0.08051733175913493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,4,balanced,0.08194666604201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,4,balanced,0.08208000163237254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,4,balanced,0.08194666604201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,4,balanced,0.0846560001373291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,4,balanced,0.08622399965922038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,4,balanced,0.08819199601809184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,4,balanced,0.09521067142486572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,4,balanced,0.10098133484522502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,8,power_law_1.2,0.1425279974937439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,4,balanced,0.10346666971842448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,4,balanced,0.11948266625404358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,4,balanced,0.1265653371810913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,4,balanced,0.1641439994176229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,1,balanced,16.511082967122395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,4,balanced,0.1880426605542501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,4,balanced,0.25412799914677936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,4,balanced,0.29362134138743085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,32,4,power_law_1.01,4.061043167114258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,4,balanced,0.4243359963099162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,4,balanced,0.5257546504338583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,4,balanced,0.7465653419494629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,8,power_law_1.2,0.15306240320205688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,4,balanced,0.9750080108642578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,4,balanced,1.1879093647003174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,4,balanced,1.8428959846496582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,4,balanced,3.5552587509155273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,8,power_law_1.2,0.158624005317688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,8,power_law_1.2,0.15869439840316774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,8,power_law_1.2,0.160863995552063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,8,power_law_1.2,0.1759104013442993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,8,power_law_1.2,0.17660800218582154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,8,power_law_1.2,0.1863808035850525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,1,4,power_law_1.01,0.20069119930267335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,1,1,power_law_1.01,0.09626880288124084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,8,power_law_1.2,0.2145983934402466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,1,4,power_law_1.01,0.30001919269561766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,1,1,power_law_1.01,0.1578879952430725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,1,4,power_law_1.01,0.18369280099868773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,1,1,power_law_1.01,0.1608896017074585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,1,4,power_law_1.01,0.21699199676513672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,1,1,power_law_1.01,0.2231231927871704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,1,4,power_law_1.01,0.21683840751647948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,1,1,power_law_1.01,0.24291200637817384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,4,balanced,0.06076799829800924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,4,balanced,0.06488533318042755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,1,4,power_law_1.01,0.22699520587921143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,1,1,power_law_1.01,0.2443392038345337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,4,balanced,0.0899733304977417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,4,balanced,0.12116799751917522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,1,4,power_law_1.01,0.22904319763183595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,1,1,power_law_1.01,0.24503679275512696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,4,balanced,0.1804800033569336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,4,balanced,0.30326932668685913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,1,4,power_law_1.01,0.22513279914855958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,1,1,power_law_1.01,0.24725759029388428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,4,balanced,0.42156799634297687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,1,4,power_law_1.01,0.2524415969848633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,1,1,power_law_1.01,0.24980480670928956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,4,balanced,0.41891201337178546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,8,power_law_1.2,0.23266561031341554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,4,balanced,0.4220373233159383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,1,4,power_law_1.01,0.2524800062179565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,1,1,power_law_1.01,0.24551680088043212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,4,balanced,0.4231199820836385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,1,4,power_law_1.01,0.2594815969467163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,1,1,power_law_1.01,0.2593024015426636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,4,balanced,0.4297279914220174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,1,4,power_law_1.01,0.24654719829559327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,4,balanced,0.4331306616465251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,1,1,power_law_1.01,0.26659839153289794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,4,balanced,0.4368373155593872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,1,4,power_law_1.01,0.28356480598449707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,1,1,power_law_1.01,0.2690623998641968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,4,balanced,0.4370986620585124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,1,4,power_law_1.01,0.2883903980255127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,1,1,power_law_1.01,0.28012800216674805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,4,balanced,0.4440266688664754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,1,4,power_law_1.01,0.32848639488220216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,4,balanced,0.4466293255488078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,1,1,power_law_1.01,0.3019455909729004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,4,balanced,0.4559146563212077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,1,4,power_law_1.01,0.4108096122741699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,1,1,power_law_1.01,0.3099008083343506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,4,balanced,0.4711093505223592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,8,power_law_1.2,0.28871679306030273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,1,4,power_law_1.01,0.3720832109451294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,1,1,power_law_1.01,0.3425920009613037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,4,balanced,0.4843786557515462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,1,4,power_law_1.01,0.4722559928894043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,4,balanced,0.51528533299764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,1,1,power_law_1.01,0.3911936044692993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,4,balanced,0.5441173315048218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,1,4,power_law_1.01,0.525875186920166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,1,1,power_law_1.01,0.4443647861480713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,4,balanced,0.6152746677398682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,1,4,power_law_1.01,0.6492608070373536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,1,1,power_law_1.01,0.5650688171386719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,4,balanced,0.6812960306803385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,4,balanced,0.8073493639628092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,1,1,power_law_1.01,0.6716864109039307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,1,4,power_law_1.01,0.9680704116821289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,4,balanced,0.9845279852549235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,1,1,power_law_1.01,0.8934399604797363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,1,4,power_law_1.01,1.0970815658569335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,4,balanced,1.2239466508229573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,8,power_law_1.2,0.3218111991882324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,1,1,power_law_1.01,1.113740825653076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,4,balanced,1.647216002146403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,1,4,power_law_1.01,1.4834112167358398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,4,balanced,2.05729071299235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,1,1,power_law_1.01,1.556608009338379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,1,4,power_law_1.01,2.067340850830078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,4,balanced,2.9284534454345703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,1,1,power_law_1.01,1.9945024490356444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,8,power_law_1.2,0.4139455795288086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,1,4,power_law_1.01,2.9578304290771484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,1,1,power_law_1.01,2.95164794921875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,4,balanced,5.541706720987956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,1,4,power_law_1.01,4.574342346191406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,1,1,power_law_1.01,3.752755355834961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,8,power_law_1.2,0.43128318786621095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,1,4,power_law_1.01,5.409151840209961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,1,1,power_law_1.01,4.633919906616211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,8,power_law_1.2,0.6154687881469727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,1,4,power_law_1.01,6.958866882324219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,1,1,power_law_1.01,7.018790435791016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,8,power_law_1.2,0.8199487686157226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,1,4,power_law_1.01,13.367398071289063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,1,1,power_law_1.01,13.6259521484375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,8,power_law_1.2,1.1036352157592773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,4,4,power_law_1.01,0.04644480049610138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,8,power_law_1.2,1.3377216339111329
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,4,4,power_law_1.01,0.04232960045337677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,4,4,power_law_1.01,0.04126079976558685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,4,4,power_law_1.01,0.041484799981117246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,4,4,power_law_1.01,0.04196479916572571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,4,4,power_law_1.01,0.04242559969425201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,1,4,power_law_1.01,26.188491821289062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,4,4,power_law_1.01,0.04304639995098114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,4,4,power_law_1.01,0.04352000057697296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,4,4,power_law_1.01,0.05028480291366577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,4,4,power_law_1.01,0.048416000604629514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,4,4,power_law_1.01,0.050918400287628174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,4,4,power_law_1.01,0.05553920269012451
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,4,4,power_law_1.01,0.05678079724311828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,2,8,power_law_1.2,1.5652480125427246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,4,4,power_law_1.01,0.06199679970741272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,4,4,power_law_1.01,0.07061120271682739
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,4,4,power_law_1.01,0.07359359860420227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,4,4,power_law_1.01,0.08673920035362244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,4,4,power_law_1.01,0.09834240078926086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,4,4,power_law_1.01,0.11422079801559448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,4,4,power_law_1.01,0.12773760557174682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,4,4,power_law_1.01,0.16162559986114503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,4,4,power_law_1.01,0.2172096014022827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,4,4,power_law_1.01,0.24165759086608887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,4,4,power_law_1.01,0.3348479986190796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,4,4,power_law_1.01,0.4274752140045166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,4,4,power_law_1.01,0.696940803527832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,2,8,power_law_1.2,2.438822364807129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,4,4,power_law_1.01,0.9128895759582519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,64,power_law_1.01,0.044172799587249754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,4,4,power_law_1.01,0.8336640357971191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,64,power_law_1.01,0.05073919892311096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,64,power_law_1.01,0.03178240060806274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,4,4,power_law_1.01,1.592204761505127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,64,power_law_1.01,0.03213439881801605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,64,power_law_1.01,0.03315199911594391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,64,power_law_1.01,0.03294720053672791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,4,4,power_law_1.01,2.6574207305908204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,64,power_law_1.01,0.033478400111198424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,64,power_law_1.01,0.0331712007522583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,64,power_law_1.01,0.03294720053672791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,64,power_law_1.01,0.03346560001373291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,128,power_law_1.2,0.09016960263252258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,64,power_law_1.01,0.035488000512123107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,64,power_law_1.01,0.03823359906673431
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,128,power_law_1.2,0.0892799973487854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,64,power_law_1.01,0.041331198811531064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,64,power_law_1.01,0.043680000305175784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,128,power_law_1.2,0.08524799942970276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,64,power_law_1.01,0.04958719909191132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,2,8,power_law_1.2,5.2222846984863285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,64,power_law_1.01,0.05370240211486817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,128,power_law_1.2,0.09121919870376587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,64,power_law_1.01,0.05214080214500427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,128,power_law_1.2,0.09321600198745728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,64,power_law_1.01,0.06110079884529114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,64,power_law_1.01,0.060844802856445314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,128,power_law_1.2,0.09937919974327088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,64,power_law_1.01,0.08042880296707153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,64,power_law_1.01,0.08343039751052857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,128,power_law_1.2,0.09341440200805665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,64,power_law_1.01,0.12389119863510131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,128,power_law_1.2,0.09763200283050537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,64,power_law_1.01,0.128601598739624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,64,power_law_1.01,0.15871360301971435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,128,power_law_1.2,0.09781759977340698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,64,power_law_1.01,0.20135679244995117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,128,power_law_1.2,0.09816960096359253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,32,8,power_law_1.2,0.03914879858493805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,64,power_law_1.01,0.27870080471038816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,128,power_law_1.2,0.09660159945487976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,32,8,power_law_1.2,0.03454720079898834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,32,8,power_law_1.2,0.03275519907474518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,64,power_law_1.01,0.3468735933303833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,128,power_law_1.2,0.09696639776229858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,32,8,power_law_1.2,0.03243519961833954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,32,8,power_law_1.2,0.031795200705528257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,32,8,power_law_1.2,0.03278079926967621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,128,power_law_1.2,0.09818239808082581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,64,power_law_1.01,0.39084160327911377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,32,8,power_law_1.2,0.03327359855175018
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,32,8,power_law_1.2,0.03399679958820343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,128,power_law_1.2,0.0960319995880127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,32,8,power_law_1.2,0.03434880077838898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,32,8,power_law_1.2,0.03394559919834137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,32,8,power_law_1.2,0.03519999980926514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,64,power_law_1.01,0.595577621459961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,128,power_law_1.2,0.09884799718856811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,32,8,power_law_1.2,0.03696640133857727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,32,8,power_law_1.2,0.036575999855995175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,32,8,power_law_1.2,0.03697920143604279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,128,power_law_1.2,0.09669119715690613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,4,power_law_1.2,0.03506560027599335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,32,8,power_law_1.2,0.04287360012531281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,4,power_law_1.2,0.03774079978466034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,32,8,power_law_1.2,0.044352000951766966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,128,power_law_1.2,0.10856959819793702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,32,8,power_law_1.2,0.04550400078296661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,4,power_law_1.2,0.03738240003585815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,32,8,power_law_1.2,0.055801600217819214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,64,power_law_1.01,1.3667455673217774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,4,power_law_1.2,0.04300160109996796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,32,8,power_law_1.2,0.060550397634506224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,128,power_law_1.2,0.1190335988998413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,32,8,power_law_1.2,0.07552639842033386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,4,power_law_1.2,0.04856959879398346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,32,8,power_law_1.2,0.08895360231399536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,32,8,power_law_1.2,0.12438399791717529
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,4,power_law_1.2,0.0701312005519867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,128,power_law_1.2,0.13647359609603882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,32,8,power_law_1.2,0.13489279747009278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,4,power_law_1.2,0.07517439723014832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,32,8,power_law_1.2,0.18794239759445192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,4,power_law_1.2,0.07448959946632386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,128,power_law_1.2,0.16414079666137696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,32,8,power_law_1.2,0.23037440776824952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,32,8,power_law_1.2,0.32794239521026614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,4,power_law_1.2,0.07807360291481018
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,32,8,power_law_1.2,0.3897919893264771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,128,power_law_1.2,0.19461120367050172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,4,power_law_1.2,0.08104320168495179
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,32,8,power_law_1.2,0.48275198936462405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,4,power_law_1.2,0.08117120265960694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,128,power_law_1.2,0.24328320026397704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,32,8,power_law_1.2,0.7296768188476562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,4,power_law_1.2,0.0791808009147644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,4,power_law_1.2,0.08734719753265381
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,128,power_law_1.2,0.3182463884353638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,32,8,power_law_1.2,1.3830911636352539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,4,power_law_1.2,0.08722559809684753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,4,power_law_1.2,0.09223679900169372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,128,power_law_1.2,0.45116801261901857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,4,power_law_1.2,0.09838079810142517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,4,power_law_1.2,0.101363205909729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,4,power_law_1.2,0.11123199462890625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,128,power_law_1.2,0.5260608196258545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,4,power_law_1.2,0.11921919584274292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,4,power_law_1.2,0.14005119800567628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,128,power_law_1.2,0.8547519683837891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,4,power_law_1.2,0.17233920097351074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,4,power_law_1.2,0.19634560346603394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,128,power_law_1.2,1.1540096282958985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,4,power_law_1.2,0.25842559337615967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,4,power_law_1.2,0.3359231948852539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,4,power_law_1.2,0.41089282035827634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,128,power_law_1.2,1.547878360748291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,128,power_law_1.2,0.063155198097229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,4,power_law_1.2,0.6155456066131592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,4,power_law_1.2,0.8203904151916503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,128,power_law_1.2,2.4349632263183594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,128,power_law_1.2,0.06264320015907288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,4,power_law_1.2,1.1027839660644532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,32,8,balanced,0.03162133445342382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,32,8,balanced,0.030906667311986286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,32,8,balanced,0.0317493329445521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,32,8,balanced,0.03284800052642822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,32,8,balanced,0.03349333256483078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,32,8,balanced,0.035061334570248924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,128,power_law_1.2,0.05361279845237732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,32,8,balanced,0.037050666908423104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,32,8,balanced,0.03495999922355016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,32,8,balanced,0.03522133330504099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,32,8,balanced,0.03526933242877325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,32,8,balanced,0.035216001172860466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,32,8,balanced,0.0371573343873024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,32,8,balanced,0.03695466617743174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,4,power_law_1.2,1.7702016830444336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,32,8,balanced,0.03722133239110311
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,32,8,balanced,0.04053866614898046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,32,8,balanced,0.04031466692686081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,32,8,balanced,0.04509866734345754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,32,8,balanced,0.04780266682306925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,32,8,balanced,0.049141332507133484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,32,8,balanced,0.055311997731526695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,32,8,balanced,0.06196266909440359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,32,8,balanced,0.07804266611735027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,32,8,balanced,0.09869333108266194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,32,8,balanced,0.12735467155774435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,128,power_law_1.2,4.755078506469727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,32,8,balanced,0.14686399698257446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,128,power_law_1.2,0.055936002731323244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,32,8,balanced,0.20907199382781982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,32,8,balanced,0.31493866443634033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,32,8,balanced,0.41524799664815265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,32,8,balanced,0.5845226844151815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,32,8,balanced,1.0838186740875244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,128,power_law_1.2,0.05541120171546936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,4,power_law_1.2,3.172972869873047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,128,power_law_1.2,0.05624960064888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,128,power_law_1.2,0.05618559718132019
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,128,power_law_1.2,0.05681279897689819
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,128,power_law_1.2,0.05605760216712952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,128,power_law_1.2,0.05661439895629883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,128,power_law_1.2,0.058303999900817874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,8,8,power_law_1.01,0.033766400814056394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,128,power_law_1.2,0.061990398168563846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,8,8,power_law_1.01,0.032179200649261476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,8,8,power_law_1.01,0.031641599535942075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,4,power_law_1.01,0.09694719910621644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,8,8,power_law_1.01,0.03191039860248566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,8,8,power_law_1.01,0.031692799925804135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,128,power_law_1.2,0.061375999450683595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,8,8,power_law_1.01,0.03221119940280914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,4,power_law_1.01,0.18686720132827758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,8,8,power_law_1.01,0.03402880132198334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,4,power_law_1.01,0.07214080095291138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,8,8,power_law_1.01,0.03394559919834137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,8,8,power_law_1.01,0.03467519879341126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,128,power_law_1.2,0.0642687976360321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,8,8,power_law_1.01,0.036748799681663516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,4,power_law_1.01,0.24686079025268554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,8,8,power_law_1.01,0.034764799475669864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,4,power_law_1.01,0.10261119604110717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,8,8,power_law_1.01,0.03678719997406006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,8,8,power_law_1.01,0.039212799072265624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,128,power_law_1.2,0.07410560250282287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,8,8,power_law_1.01,0.03825919926166534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,4,power_law_1.01,0.12842240333557128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,8,8,power_law_1.01,0.044889599084854126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,8,8,power_law_1.01,0.04865919947624207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,4,power_law_1.01,0.1217919945716858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,8,8,power_law_1.01,0.0506879985332489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,4,power_law_1.01,0.197817599773407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,8,8,power_law_1.01,0.05685120224952698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,128,power_law_1.2,0.07490559816360473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,8,8,power_law_1.01,0.06595839858055115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,4,power_law_1.01,0.0874176025390625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,8,8,power_law_1.01,0.08055040240287781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,8,8,power_law_1.01,0.09288319945335388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,4,power_law_1.01,0.28558080196380614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,8,8,power_law_1.01,0.13696000576019288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,128,power_law_1.2,0.07985919713973999
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,8,8,power_law_1.01,0.16579840183258057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,4,power_law_1.01,0.10405119657516479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,8,8,power_law_1.01,0.21440000534057618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,4,power_law_1.01,0.30737919807434083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,8,8,power_law_1.01,0.27870080471038816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,4,power_law_1.01,0.13612799644470214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,8,8,power_law_1.01,0.4214144229888916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,128,power_law_1.2,0.09434880018234253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,8,8,power_law_1.01,0.5289087772369385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,4,power_law_1.01,0.41616001129150393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,8,8,power_law_1.01,0.6425856113433838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,4,power_law_1.01,0.15016959905624389
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,128,power_law_1.2,0.1100160002708435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,8,8,power_law_1.01,0.9599679946899414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,4,power_law_1.01,0.4290112018585205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,4,power_law_1.01,0.18377599716186524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,8,8,power_law_1.01,1.8600576400756836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,4,power_law_1.01,0.44846081733703613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,128,power_law_1.2,0.1420032024383545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,4,power_law_1.01,0.18588160276412963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,4,power_law_1.01,0.4705984115600586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,4,power_law_1.01,0.1967743992805481
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,128,power_law_1.2,0.1573248028755188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,4,power_law_1.01,0.4822080135345459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,4,power_law_1.01,0.20597119331359864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,128,power_law_1.2,0.20806400775909423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,4,power_law_1.01,0.49532160758972166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,4,power_law_1.01,0.20838398933410646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,4,power_law_1.01,0.5128640174865723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,4,power_law_1.01,0.2082304000854492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,128,power_law_1.2,0.22350080013275148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,4,power_law_1.01,0.5351935863494873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,4,power_law_1.01,0.2170367956161499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,128,power_law_1.2,0.2919935941696167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,4,power_law_1.01,0.23497600555419923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,4,power_law_1.01,0.5355519771575927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,4,power_law_1.01,0.24296960830688477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,4,power_law_1.01,0.576639986038208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,128,power_law_1.2,0.36740479469299314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,4,power_law_1.01,0.24887681007385254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,4,power_law_1.01,0.6044223785400391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,4,power_law_1.01,0.27945599555969236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,4,power_law_1.01,0.6449984073638916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,128,power_law_1.2,0.554252815246582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,4,power_law_1.01,0.309555196762085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,128,power_law_1.01,0.09578880071640014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,4,power_law_1.01,0.7413248062133789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,4,power_law_1.01,0.3562623977661133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,4,power_law_1.01,0.7502528190612793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,128,power_law_1.01,0.09662719964981079
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,128,power_law_1.2,0.8208512306213379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,4,power_law_1.01,0.39706881046295167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,4,power_law_1.01,0.8642239570617676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,128,power_law_1.01,0.0915391981601715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,4,power_law_1.01,0.4930111885070801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,4,power_law_1.01,0.984556770324707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,4,power_law_1.01,0.5508480072021484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,1,128,power_law_1.2,0.8448703765869141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,128,power_law_1.01,0.09200639724731445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,4,power_law_1.01,1.2124223709106445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,4,power_law_1.01,0.7075071811676026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,128,power_law_1.01,0.09638400077819824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,4,power_law_1.01,1.2861696243286134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,4,power_law_1.01,0.9124608039855957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,128,power_law_1.01,0.0963648021221161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,4,power_law_1.01,1.6739776611328125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,1,128,power_law_1.2,1.423731231689453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,4,power_law_1.01,1.210591983795166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,128,power_law_1.01,0.09729920029640197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,128,power_law_1.01,0.09557120203971863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,4,power_law_1.01,2.0072256088256837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,4,power_law_1.01,1.5714624404907227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,128,power_law_1.01,0.09477120041847228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,4,power_law_1.01,1.848089599609375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,4,power_law_1.01,2.558470344543457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,128,power_law_1.01,0.09938560128211975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,128,power_law_1.01,0.09389439821243287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,4,power_law_1.01,2.918060874938965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,4,power_law_1.01,3.486489486694336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,128,power_law_1.01,0.09543039798736572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,1,128,power_law_1.2,3.638470458984375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,128,power_law_1.01,0.10034559965133667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,128,power_law_1.01,0.09994239807128906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,4,power_law_1.01,5.348979187011719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,128,power_law_1.01,0.09874560236930847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,4,power_law_1.01,6.9596611022949215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,128,power_law_1.01,0.0974016010761261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,128,power_law_1.01,0.09934080243110657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,128,power_law_1.01,0.11409280300140381
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,128,power_law_1.01,0.11523200273513794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,128,power_law_1.01,0.15156480073928832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,128,power_law_1.01,0.17018879652023317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,128,power_law_1.01,0.20382080078125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,1,balanced,0.1286133329073588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,1,balanced,0.190938671429952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,128,power_law_1.01,0.22498559951782227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,1,balanced,0.3197919925053914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,1,balanced,0.5987039804458618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,1,balanced,1.1202613512674968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,1,balanced,1.62609068552653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,1,balanced,1.6352640787760417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,128,power_law_1.01,0.3449919939041138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,1,balanced,1.6397333145141602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,1,balanced,1.6427626609802246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,1,balanced,1.6488587061564128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,1,balanced,1.6636533737182617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,1,balanced,1.6766773859659831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,128,power_law_1.01,0.42320637702941893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,1,balanced,1.6848586400349934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,1,balanced,1.7030879656473796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,32,1,power_law_1.01,0.03022719919681549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,2,1,balanced,0.03386666625738144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,1,balanced,1.7241439819335938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,2,1,balanced,0.03573866685231527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,32,1,power_law_1.01,0.030502399802207945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,2,1,balanced,0.05363733569780985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,2,1,balanced,0.07841599980990092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,2,1,balanced,0.12187199791272481
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,2,1,balanced,0.12449600299199422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,32,1,power_law_1.01,0.030643200874328612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,2,1,balanced,0.12729066610336304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,1,balanced,1.7411999702453613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,2,1,balanced,0.12380266189575195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,2,1,balanced,0.1272586683432261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,32,1,power_law_1.01,0.03312000036239624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,2,1,balanced,0.13037866353988647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,2,1,balanced,0.12787733475367227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,2,1,balanced,0.1316106617450714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,32,1,power_law_1.01,0.03353599905967712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,1,balanced,1.7771520614624023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,2,1,balanced,0.1317759950955709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,2,1,balanced,0.13614400227864584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,128,power_law_1.01,0.610649585723877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,2,1,balanced,0.14298133055369058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,32,1,power_law_1.01,0.035596799850463864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,2,1,balanced,0.14511467019716898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,2,1,balanced,0.15337066849072775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,2,1,balanced,0.16736533244450888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,1,balanced,0.18094933032989502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,1,balanced,1.9805760383605957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,1,balanced,0.22103466590245566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,32,1,power_law_1.01,0.03615359961986542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,1,balanced,0.2374346653620402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,1,balanced,0.3131573398907979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,32,1,power_law_1.01,0.03635840117931366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,1,balanced,0.36354132493336994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,1,balanced,1.9000693957010906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,1,balanced,0.49950400988260907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,32,1,power_law_1.01,0.03715839982032776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,1,balanced,0.6147679885228475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,32,1,power_law_1.01,0.03765760064125061
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,1,balanced,0.8750186761220297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,1,balanced,2.33349339167277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,32,1,power_law_1.01,0.03797760009765625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,1,balanced,1.1540586948394775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,32,1,power_law_1.01,0.042028799653053284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,1,balanced,2.1661760012308755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,32,1,power_law_1.01,0.04213759899139404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,1,balanced,1.438634713490804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,32,1,power_law_1.01,0.0460671991109848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,128,power_law_1.01,0.8253631591796875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,32,1,power_law_1.01,0.054502397775650024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,1,balanced,2.1623679796854653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,32,1,power_law_1.01,0.05727360248565674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,1,balanced,3.515690803527832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,32,1,power_law_1.01,0.05581439733505249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,32,1,power_law_1.01,0.0689791977405548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,1,balanced,2.6436959902445474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,32,1,power_law_1.01,0.07908480167388916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,1,balanced,4.185231844584147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,32,1,power_law_1.01,0.10104320049285889
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,32,1,power_law_1.01,0.11645439863204957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,32,1,power_law_1.01,0.15735679864883423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,1,balanced,3.9035679499308267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,32,1,power_law_1.01,0.1857856035232544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,128,power_law_1.01,0.9485312461853027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,32,1,power_law_1.01,0.2644351959228516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,1,balanced,4.035120010375977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,32,1,power_law_1.01,0.338918399810791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,32,1,power_law_1.01,0.4930431842803955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,32,1,power_law_1.01,0.6445951938629151
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,1,balanced,5.875514984130859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,32,1,power_law_1.01,0.8003711700439453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,32,1,power_law_1.01,1.2678400039672852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,128,power_law_1.01,1.7263616561889648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,1,balanced,6.8606828053792315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,32,1,power_law_1.01,2.4790719985961913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,2,1,balanced,8.521573384602865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,128,power_law_1.01,3.07956485748291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,2,1,balanced,12.38491694132487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,2,1,balanced,24.38720957438151
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,1,power_law_1.01,0.1493824005126953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,1,8,power_law_1.2,0.0573311984539032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,1,8,power_law_1.2,0.06530560255050659
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,1,8,power_law_1.2,0.05248640179634094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,1,power_law_1.01,0.1921023964881897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,1,8,power_law_1.2,0.052121597528457644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,1,8,power_law_1.2,0.053420799970626834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,1,8,power_law_1.2,0.05359359979629517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,1,8,power_law_1.2,0.05454720258712768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,1,8,power_law_1.2,0.05573760271072388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,1,8,power_law_1.2,0.06213759779930115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,1,8,power_law_1.2,0.06352000236511231
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,1,8,power_law_1.2,0.06272640228271484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,1,8,power_law_1.2,0.06650239825248719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,1,8,power_law_1.2,0.0692031979560852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,1,power_law_1.01,0.21901440620422363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,1,8,power_law_1.2,0.07326080203056336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,1,8,power_law_1.2,0.09162880182266235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,1,8,power_law_1.2,0.09521920084953309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,1,8,power_law_1.2,0.09718400239944458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,128,power_law_1.01,0.039628800749778745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,1,8,power_law_1.2,0.11534080505371094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,1,8,power_law_1.2,0.14228479862213134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,1,8,power_law_1.2,0.18587520122528076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,1,8,power_law_1.2,0.19765759706497193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,1,power_law_1.01,0.3504256010055542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,1,8,power_law_1.2,0.2889024019241333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,128,power_law_1.01,0.03935999870300293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,1,8,power_law_1.2,0.3777856111526489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,1,8,power_law_1.2,0.5566783905029297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,1,8,power_law_1.2,0.7467520236968994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,128,power_law_1.01,0.03866879940032959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,1,8,power_law_1.2,1.0367039680480956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,1,power_law_1.01,0.46796159744262694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,1,8,power_law_1.2,1.3604672431945801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,128,power_law_1.01,0.04017280042171478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,2,128,power_law_1.01,0.03132160007953644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,1,8,power_law_1.2,1.8897855758666993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,128,power_law_1.01,0.03963519930839539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,1,power_law_1.01,0.8330495834350586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,2,128,power_law_1.01,0.03043839931488037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,1,8,power_law_1.2,2.5663616180419924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,2,128,power_law_1.01,0.030988800525665283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,128,power_law_1.01,0.04036479890346527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,2,128,power_law_1.01,0.030988800525665283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,128,power_law_1.01,0.04019840061664581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,2,128,power_law_1.01,0.03141759932041168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,1,power_law_1.01,0.8757184028625489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,1,8,power_law_1.2,5.032198333740235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,2,128,power_law_1.01,0.03185279965400696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,128,power_law_1.01,0.03985919952392578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,2,128,power_law_1.01,0.03212159872055054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,2,128,power_law_1.01,0.032595199346542356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,128,power_law_1.01,0.03956480026245117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,1,power_law_1.01,0.9407360076904296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,2,128,power_law_1.01,0.032793599367141726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,128,power_law_1.01,0.04071039855480194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,2,128,power_law_1.01,0.03248000144958496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,2,128,power_law_1.01,0.03318400084972382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,128,power_law_1.01,0.04048640131950378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,2,128,power_law_1.01,0.033555200695991515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,1,power_law_1.01,0.974886417388916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,128,power_law_1.01,0.04017919898033142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,2,128,power_law_1.01,0.03357439935207367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,2,128,power_law_1.01,0.03440000116825104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,128,power_law_1.01,0.04048640131950378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,2,128,power_law_1.01,0.0367935985326767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,1,power_law_1.01,0.9665792465209961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,128,power_law_1.01,0.041145598888397215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,2,128,power_law_1.01,0.03711999952793121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,128,power_law_1.01,0.04485760033130646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,2,128,power_law_1.01,0.0374208003282547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,2,128,power_law_1.01,0.03944959938526153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,128,power_law_1.01,0.044627198576927186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,1,power_law_1.01,1.019603157043457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,128,power_law_1.01,0.039552000164985654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,128,power_law_1.01,0.04519039988517761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,128,power_law_1.01,0.043398401141166686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,128,power_law_1.01,0.04574080109596253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,128,power_law_1.01,0.047443199157714847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,1,power_law_1.01,1.0476672172546386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,128,power_law_1.01,0.05138559937477112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,128,power_law_1.01,0.04943360090255737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,128,power_law_1.01,0.05675519704818725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,128,power_law_1.01,0.055232000350952146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,128,power_law_1.01,0.07608960270881653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,1,power_law_1.01,1.0520319938659668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,128,power_law_1.01,0.05868160128593445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,128,power_law_1.01,0.09167360067367554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,128,power_law_1.01,0.06826239824295044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,1,power_law_1.2,0.034764799475669864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,128,power_law_1.01,0.12117760181427002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,1,power_law_1.2,0.03372800052165985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,1,power_law_1.01,1.09814395904541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,128,power_law_1.01,0.08401280045509338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,1,power_law_1.2,0.03803519904613495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,128,power_law_1.01,0.1533951997756958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,1,power_law_1.2,0.041043201088905336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,128,power_law_1.01,0.10836479663848878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,1,power_law_1.2,0.04599039852619171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,128,power_law_1.01,0.19156479835510254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,1,power_law_1.01,1.148697566986084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,1,power_law_1.2,0.05971840023994446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,128,power_law_1.01,0.13038719892501832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,1,power_law_1.2,0.0637503981590271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,1,power_law_1.2,0.06565120220184326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,128,power_law_1.01,0.17072639465332032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,128,power_law_1.01,0.30368640422821047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,1,power_law_1.2,0.06838399767875672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,1,power_law_1.01,1.1744895935058595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,1,power_law_1.2,0.06963840126991272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,128,power_law_1.01,0.22588160037994384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,1,power_law_1.2,0.07366399765014649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,1,power_law_1.2,0.07864959836006165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,1,power_law_1.01,1.2482879638671875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,2688,1856,6,128,1,128,power_law_1.01,0.26686720848083495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,128,power_law_1.01,0.5857215881347656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,1,power_law_1.2,0.082259202003479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,1,power_law_1.2,0.08333439826965332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,1,power_law_1.2,0.09142400026321411
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,2688,1856,6,128,1,128,power_law_1.01,0.43726720809936526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,1,power_law_1.01,1.3761216163635255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,1,power_law_1.2,0.09400320053100586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,1,power_law_1.2,0.10362240076065063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,1,power_law_1.2,0.12395520210266113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,1,power_law_1.2,0.14558080434799195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,2688,1856,6,128,1,128,power_law_1.01,0.9186623573303223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,1,power_law_1.01,1.5162816047668457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,1,power_law_1.2,0.1843840003013611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,1,power_law_1.2,0.22095999717712403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,1,power_law_1.2,0.3028543949127197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,1,power_law_1.01,1.8373823165893555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,1,power_law_1.2,0.3848896026611328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,1,power_law_1.2,0.5476480007171631
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,1,power_law_1.2,0.711513614654541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,1,power_law_1.01,1.961555290222168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,1,power_law_1.2,1.0427007675170898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,1,power_law_1.2,1.3652480125427247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,1,power_law_1.01,2.4526975631713865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,8,8,balanced,0.0359199990828832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,8,8,balanced,0.03604800005753835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,8,8,balanced,0.033813332517941795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,8,8,balanced,0.03219199925661087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,8,8,balanced,0.03182400017976761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,8,8,balanced,0.03186666717131933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,8,8,balanced,0.03366400053103765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,8,8,balanced,0.031727999448776245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,1,power_law_1.2,1.6879680633544922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,8,8,balanced,0.03347733368476232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,8,8,balanced,0.0317546675602595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,8,8,balanced,0.032618666688601174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,8,8,balanced,0.033743999898433685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,8,8,balanced,0.03377600014209747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,8,8,balanced,0.033813332517941795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,8,8,balanced,0.03993066648642222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,8,8,balanced,0.039642666776975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,8,8,balanced,0.041738669077555336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,8,8,balanced,0.04382933179537455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,8,8,balanced,0.043920000394185386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,8,8,balanced,0.05727999905745188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,8,8,balanced,0.05624533196290334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,8,8,balanced,0.07493333518505096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,8,8,balanced,0.0772159993648529
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,8,8,balanced,0.10077866911888123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,8,8,balanced,0.11429333686828613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,8,8,balanced,0.15094400445620218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,8,8,balanced,0.17709867159525552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,8,8,balanced,0.21606934070587158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,8,8,balanced,0.29205334186553955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,8,8,balanced,0.5336266756057739
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,1,power_law_1.01,2.9553152084350587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,1,power_law_1.2,2.6530431747436523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,2,8,power_law_1.01,0.04118399918079376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,2,8,power_law_1.01,0.042483198642730716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,2,8,power_law_1.01,0.039987200498580934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,1,power_law_1.01,3.9309886932373046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,2,8,power_law_1.01,0.039603200554847715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,2,8,power_law_1.01,0.04087679982185364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,2,8,power_law_1.01,0.04140160083770752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,2,8,power_law_1.01,0.04196479916572571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,1,power_law_1.2,5.301612854003906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,2,8,power_law_1.01,0.041254401206970215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,2,8,power_law_1.01,0.043289598822593686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,2,8,power_law_1.01,0.0447488009929657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,2,8,power_law_1.01,0.045126399397850035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,2,8,power_law_1.01,0.04917120039463043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,2,8,power_law_1.01,0.049465599656105044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,2,8,power_law_1.01,0.05187839865684509
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,2,8,power_law_1.01,0.058323198556900026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,1,power_law_1.01,5.150809478759766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,2,8,power_law_1.01,0.06969599723815918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,2,8,power_law_1.01,0.06739839911460876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,2,8,power_law_1.01,0.07807999849319458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,2,8,power_law_1.01,0.09941759705543518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,2,8,power_law_1.01,0.12721920013427734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,2,8,power_law_1.01,0.1336127996444702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,2,8,power_law_1.01,0.17946879863739013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,2,8,power_law_1.01,0.2310015916824341
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,2,8,power_law_1.01,0.2840127944946289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,2,8,power_law_1.01,0.3660928010940552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,2,8,power_law_1.01,0.5724544048309326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,1,power_law_1.01,7.099327850341797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,2,8,power_law_1.01,0.726688003540039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,2,8,power_law_1.01,0.8422783851623535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,2,8,power_law_1.01,1.4362431526184083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,2,8,power_law_1.01,2.7019584655761717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,1,power_law_1.01,8.917632293701171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,16,1,power_law_1.01,0.04072319865226746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,16,1,power_law_1.01,0.05028480291366577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,16,1,power_law_1.01,0.05484799742698669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,16,1,power_law_1.01,0.0720255970954895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,16,1,power_law_1.01,0.07461760044097901
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,16,1,power_law_1.01,0.07505919933319091
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,16,1,power_law_1.01,0.07507839798927307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,16,1,power_law_1.01,0.07596799731254578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,16,1,power_law_1.01,0.08085759878158569
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,16,1,power_law_1.01,0.07991039752960205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,16,1,power_law_1.01,0.07853440046310425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,16,1,power_law_1.01,0.08172799944877625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,16,1,power_law_1.01,0.08370559811592101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,16,1,power_law_1.01,0.09399679899215699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,16,1,power_law_1.01,0.10229760408401489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,16,1,power_law_1.01,0.10410239696502685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,16,1,power_law_1.01,0.11684479713439941
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,16,1,power_law_1.01,0.1348031997680664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,16,1,power_law_1.01,0.13812479972839356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,16,1,power_law_1.01,0.1744704008102417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,16,1,power_law_1.01,0.19298559427261353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,16,1,power_law_1.01,0.25271039009094237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,16,1,power_law_1.01,0.2897984027862549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,16,1,power_law_1.01,0.4013696193695068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,16,1,power_law_1.01,0.5127103805541993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,1,1,power_law_1.01,10.944818878173828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,16,1,power_law_1.01,0.7131392002105713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,16,1,power_law_1.01,0.9166655540466309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,16,1,power_law_1.01,1.1308159828186035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,16,1,power_law_1.01,1.7518655776977539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,16,1,power_law_1.01,3.426252746582031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,64,balanced,0.062165334820747375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,64,balanced,0.05946666498978933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,64,balanced,0.062021334966023765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,64,balanced,0.06135466694831848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,64,balanced,0.06506666541099548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,64,balanced,0.06868800024191539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,64,balanced,0.07776533563931783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,64,balanced,0.07792533437410991
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,64,balanced,0.0775786687930425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,64,balanced,0.0792746643225352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,64,balanced,0.07653866708278656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,64,balanced,0.07745600243409474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,64,balanced,0.0801333338022232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,64,balanced,0.07949333389600118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,64,balanced,0.08772266904513042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,64,balanced,0.0863200028737386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,64,balanced,0.09297066926956177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,64,balanced,0.10190932949384053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,64,balanced,0.11085333426793416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,64,balanced,0.12277332941691081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,64,balanced,0.13984533150990805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,64,balanced,0.17167999347050986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,64,balanced,0.19896533091862997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,64,balanced,0.2595786650975545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,64,balanced,0.3319573402404785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,64,balanced,0.4480160077412923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,64,balanced,0.5695093472798666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,64,balanced,0.7039360205332438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,1,1,power_law_1.01,16.98802490234375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,64,balanced,1.075610637664795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,64,balanced,2.101306597391764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,16,power_law_1.01,0.0573311984539032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,16,power_law_1.01,0.05559679865837097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,16,power_law_1.01,0.04513919949531555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,16,power_law_1.01,0.045721599459648134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,16,power_law_1.01,0.049932798743247984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,4,1,power_law_1.01,0.04467839896678925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,4,1,power_law_1.01,0.04809600114822388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,16,power_law_1.01,0.05002239942550659
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,4,1,power_law_1.01,0.05057920217514038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,1,1,power_law_1.01,32.27095031738281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,4,1,power_law_1.01,0.06967679858207702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,16,power_law_1.01,0.055366402864456175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,4,1,power_law_1.01,0.06806399822235107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,4,1,power_law_1.01,0.0689791977405548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,4,1,power_law_1.01,0.06933760046958923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,16,power_law_1.01,0.05666559934616089
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,4,1,power_law_1.01,0.07023360133171082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,4,1,power_law_1.01,0.07046399712562561
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,16,power_law_1.01,0.05615360140800476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,4,1,power_law_1.01,0.07153919935226441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,4,1,power_law_1.01,0.07256320118904114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,4,1,power_law_1.01,0.07511039972305297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,16,power_law_1.01,0.05815039873123169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,4,1,power_law_1.01,0.07793920040130616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,4,1,power_law_1.01,0.08456959724426269
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,4,1,power_law_1.01,0.09550079703330994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,16,power_law_1.01,0.05952640175819397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,4,1,power_law_1.01,0.0969215989112854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,4,1,power_law_1.01,0.09832320213317872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,4,1,power_law_1.01,0.11470719575881957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,16,power_law_1.01,0.0604095995426178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,4,1,power_law_1.01,0.12904319763183594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,4,1,power_law_1.01,0.15854079723358155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,4,1,power_law_1.01,0.1964416027069092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,16,power_law_1.01,0.06331520080566407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,4,1,power_law_1.01,0.24787840843200684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,4,1,power_law_1.01,0.29809279441833497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,16,power_law_1.01,0.06796799898147583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,4,1,power_law_1.01,0.4102975845336914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,16,power_law_1.01,0.07406719923019409
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,4,1,power_law_1.01,0.5112895965576172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,4,1,power_law_1.01,0.7216959953308105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,16,power_law_1.01,0.07983999848365783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,4,1,power_law_1.01,0.9366975784301758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,4,1,power_law_1.01,1.1539456367492675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,16,power_law_1.01,0.08625919818878174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,4,1,power_law_1.01,1.773721694946289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,16,power_law_1.01,0.09889280200004577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,16,power_law_1.01,0.11401599645614624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,1,power_law_1.2,0.05440639853477478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,4,1,power_law_1.01,3.5074047088623046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,1,power_law_1.2,0.05974400043487549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,16,power_law_1.01,0.13548799753189086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,1,power_law_1.2,0.0874239981174469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,16,power_law_1.01,0.16289279460906983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,1,power_law_1.2,0.12490880489349365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,1,power_law_1.2,0.1687999963760376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,16,power_law_1.01,0.21301119327545165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,1,power_law_1.2,0.314899206161499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,1,power_law_1.2,0.32984960079193115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,16,power_law_1.01,0.25523200035095217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,1,power_law_1.2,0.3473407983779907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,1,power_law_1.2,0.35203840732574465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,16,power_law_1.01,0.34494719505310056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,1,power_law_1.2,0.3549247980117798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,1,power_law_1.2,0.37141120433807373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,16,power_law_1.01,0.43669757843017576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,1,power_law_1.2,0.36966400146484374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,1,power_law_1.2,0.37985279560089114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,16,power_law_1.01,0.6525504112243652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,1,power_law_1.2,0.39248640537261964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,1,power_law_1.2,0.40402560234069823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,1,power_law_1.2,0.4207744121551514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,16,power_law_1.01,0.8248767852783203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,1,power_law_1.2,0.42821760177612306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,1,power_law_1.2,0.47055997848510744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,16,power_law_1.01,0.9907967567443847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,1,power_law_1.2,0.48821120262145995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,1,power_law_1.2,0.5612671852111817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,1,power_law_1.2,0.6082560062408447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,16,power_law_1.01,1.489363193511963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,64,power_law_1.01,0.03971199989318848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,1,power_law_1.2,0.738259220123291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,1,power_law_1.2,0.8725055694580078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,64,power_law_1.01,0.039059200882911684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,1,power_law_1.2,1.1379072189331054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,64,power_law_1.01,0.03578880131244659
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,1,power_law_1.2,1.411564826965332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,16,power_law_1.01,3.148703956604004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,64,power_law_1.01,0.03575679957866669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,1,power_law_1.2,1.9223615646362304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,64,power_law_1.01,0.03615359961986542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,64,power_law_1.01,0.03702400028705597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,1,power_law_1.2,2.4663616180419923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,64,power_law_1.01,0.03591679930686951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,64,power_law_1.01,0.03648639917373657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,1,power_law_1.2,3.0021760940551756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,64,power_law_1.01,0.037324801087379456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,64,power_law_1.01,0.036985599994659425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,1,power_law_1.2,4.496281433105469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,64,power_law_1.01,0.0377344012260437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,64,power_law_1.01,0.03893119990825653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,64,power_law_1.01,0.039443200826644896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,64,power_law_1.01,0.04036479890346527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,64,power_law_1.01,0.042259201407432556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,1,power_law_1.2,8.504434967041016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,64,power_law_1.01,0.04366079866886139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,8,4,power_law_1.01,0.042156800627708435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,64,power_law_1.01,0.04535039961338043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,8,4,power_law_1.01,0.04565759897232056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,8,4,power_law_1.01,0.048518401384353635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,8,4,power_law_1.01,0.05204480290412903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,8,4,power_law_1.01,0.05220479965209961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,64,power_law_1.01,0.05010560154914856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,8,4,power_law_1.01,0.05262079834938049
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,8,4,power_law_1.01,0.055270397663116456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,8,4,power_law_1.01,0.05811840295791626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,8,4,power_law_1.01,0.05795199871063232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,8,4,power_law_1.01,0.058950400352478026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,64,power_law_1.01,0.051558399200439455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,8,4,power_law_1.01,0.06077439785003662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,8,4,power_law_1.01,0.06630399823188782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,8,4,power_law_1.01,0.06661760210990905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,8,4,power_law_1.01,0.06849279999732971
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,64,power_law_1.01,0.058931201696395874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,8,4,power_law_1.01,0.07388799786567687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,8,4,power_law_1.01,0.07453439831733703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,8,4,power_law_1.01,0.08156160116195679
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,8,4,power_law_1.01,0.09908480048179627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,64,power_law_1.01,0.06635519862174988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,8,4,power_law_1.01,0.11396479606628418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,8,4,power_law_1.01,0.14648959636688233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,8,4,power_law_1.01,0.18316800594329835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,8,4,power_law_1.01,0.22833919525146484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,64,power_law_1.01,0.0930624008178711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,8,4,power_law_1.01,0.26525440216064455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,8,4,power_law_1.01,0.41447038650512696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,8,4,power_law_1.01,0.48508801460266116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,64,power_law_1.01,0.0946175992488861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,8,4,power_law_1.01,0.6891583919525146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,8,4,power_law_1.01,1.0079423904418945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,64,power_law_1.01,0.1351807951927185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,8,4,power_law_1.01,1.2504639625549316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,8,4,power_law_1.01,1.7380352020263672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,64,power_law_1.01,0.15962239503860473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,64,power_law_1.01,0.23113598823547363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,8,4,power_law_1.01,3.2790912628173827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,64,power_law_1.01,0.2833791971206665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,32,balanced,0.05420266588528951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,32,balanced,0.050069332122802734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,32,balanced,0.05026133358478546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,32,balanced,0.051813334226608276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,32,balanced,0.05402666827042898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,32,balanced,0.0617439995209376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,32,balanced,0.060831998785336815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,32,balanced,0.06221333146095276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,32,balanced,0.062261333068211876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,32,balanced,0.06436266501744588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,32,balanced,0.06410666803518932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,32,balanced,0.06660800178845723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,64,power_law_1.01,0.35060479640960696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,32,balanced,0.06613866488138835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,32,balanced,0.06778133412202199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,32,balanced,0.07522133489449818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,32,balanced,0.07843199868996938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,32,balanced,0.08144533137480418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,32,balanced,0.08945066730181377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,32,balanced,0.09444800019264221
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,32,balanced,0.10813333590825398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,32,balanced,0.11943999926249187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,32,balanced,0.14010666807492575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,32,balanced,0.167738676071167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,32,balanced,0.21493866046269736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,32,balanced,0.25514666239420575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,32,balanced,0.3500746488571167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,32,balanced,0.4386399984359741
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,32,balanced,0.540229320526123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,32,balanced,0.798138697942098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,64,power_law_1.01,0.5334015846252441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,32,balanced,1.5592427253723145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,64,power_law_1.01,1.0320256233215332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,1,8,power_law_1.2,0.12117760181427002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,1,8,power_law_1.2,0.12711679935455322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,1,8,power_law_1.2,0.1235967993736267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,1,8,power_law_1.2,0.12189439535140992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,1,8,power_law_1.2,0.1220479965209961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,1,8,power_law_1.2,0.12829439640045165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,1,8,power_law_1.2,0.18050559759140014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,1,8,power_law_1.2,0.1474496006965637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,1,8,power_law_1.2,0.16035200357437135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,1,8,power_law_1.2,0.2035968065261841
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,1,8,power_law_1.2,0.15054080486297608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,1,8,power_law_1.2,0.16603519916534423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,1,8,power_law_1.2,0.1756351947784424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,1,8,power_law_1.2,0.228767991065979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,1,8,power_law_1.2,0.28520960807800294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,1,8,power_law_1.2,0.3029887914657593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,1,8,power_law_1.2,0.2641792058944702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,1,8,power_law_1.2,0.34963839054107665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,32,4,power_law_1.01,0.03127039968967438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,1,8,power_law_1.2,0.41834239959716796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,32,4,power_law_1.01,0.030771198868751525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,1,8,power_law_1.2,0.5961919784545898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,32,4,power_law_1.01,0.03076480031013489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,1,8,power_law_1.2,0.6940671920776367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,32,4,power_law_1.01,0.03000960052013397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,32,4,power_law_1.01,0.030028799176216127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,1,8,power_law_1.2,0.9882368087768555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,32,4,power_law_1.01,0.03139199912548065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,32,4,power_law_1.01,0.031590399146080014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,32,4,power_law_1.01,0.03170560002326965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,1,8,power_law_1.2,1.2672575950622558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,32,4,power_law_1.01,0.03255040049552917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,32,4,power_law_1.01,0.03245440125465393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,32,4,power_law_1.01,0.03344640135765076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,1,8,power_law_1.2,2.0028608322143553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,32,4,power_law_1.01,0.03425279855728149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,32,4,power_law_1.01,0.035308799147605895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,32,4,power_law_1.01,0.036499199271202085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,32,4,power_law_1.01,0.04074879884719849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,1,8,power_law_1.2,2.3231552124023436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,32,4,power_law_1.01,0.04289920032024384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,32,4,power_law_1.01,0.04497919976711273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,32,4,power_law_1.01,0.05175039768218994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,1,power_law_1.2,0.05066879987716675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,32,4,power_law_1.01,0.05433599948883057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,32,4,power_law_1.01,0.06908159852027893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,1,8,power_law_1.2,3.885388946533203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,32,4,power_law_1.01,0.07452160120010376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,32,4,power_law_1.01,0.12163200378417968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,32,4,power_law_1.01,0.14328320026397706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,32,4,power_law_1.01,0.202726411819458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,1,power_law_1.2,0.05541759729385376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,32,4,power_law_1.01,0.2682111978530884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,1,8,power_law_1.2,4.895743942260742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,32,4,power_law_1.01,0.3514496088027954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,32,4,power_law_1.01,0.4342976093292236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,1,power_law_1.2,0.06236159801483154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,32,4,power_law_1.01,0.6792064189910889
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,32,4,power_law_1.01,0.9423168182373047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,1,8,power_law_1.2,6.222143936157226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,1,power_law_1.2,0.07410560250282287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,32,4,power_law_1.01,1.8772672653198241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,1,power_law_1.2,0.0887167990207672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,1,power_law_1.2,0.1256127953529358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,1,8,power_law_1.2,9.676748657226563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,1,power_law_1.2,0.13351680040359498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,1,power_law_1.2,0.1344704031944275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,1,power_law_1.2,0.13848960399627686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,1,power_law_1.2,0.14087680578231812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,1,power_law_1.2,0.14350080490112305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,1,power_law_1.2,0.14799360036849976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,1,8,power_law_1.2,19.887667846679687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,1,power_law_1.2,0.14869120121002197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,1,power_law_1.2,0.15762560367584227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,1,power_law_1.2,0.16915199756622315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,1,power_law_1.2,0.17405439615249635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,1,power_law_1.2,0.18966399431228637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,1,power_law_1.2,0.2236543893814087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,1,power_law_1.2,0.25671679973602296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,1,power_law_1.2,0.32238719463348386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,32,power_law_1.2,0.09744639992713929
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,1,power_law_1.2,0.38181118965148925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,32,power_law_1.2,0.10096640586853027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,1,power_law_1.2,0.5190207958221436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,32,power_law_1.2,0.0632960021495819
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,1,power_law_1.2,0.6485568046569824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,32,power_law_1.2,0.06510720252990723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,1,power_law_1.2,0.9075072288513184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,32,power_law_1.2,0.07351679801940918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,1,power_law_1.2,1.1556415557861328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,32,power_law_1.2,0.08241919875144958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,4,balanced,0.052426666021347046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,4,balanced,0.08302933474381764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,4,balanced,0.09726400176684062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,4,balanced,0.1467359960079193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,4,balanced,0.24079465866088867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,4,balanced,0.43702932198842365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,4,balanced,0.43678398927052814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,4,balanced,0.44207465648651123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,32,power_law_1.2,0.08695679903030396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,4,balanced,0.4408213297526042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,1,power_law_1.2,1.675846481323242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,4,balanced,0.44417067368825275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,4,balanced,0.44297067324320477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,4,balanced,0.44675731658935547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,4,balanced,0.4490400155385335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,4,balanced,0.4549386501312256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,4,balanced,0.4585760037104289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,4,balanced,0.46276267369588214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,4,balanced,0.4714133342107137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,32,power_law_1.2,0.09283840060234069
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,4,balanced,0.48448534806569415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,4,balanced,0.4981173276901245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,4,balanced,0.5226453145345052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,4,balanced,0.5545226732889811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,4,balanced,0.596501350402832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,4,balanced,0.63810133934021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,1,power_law_1.2,2.179667282104492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,32,power_law_1.2,0.0882752001285553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,4,balanced,0.8061280250549316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,4,balanced,0.8420800367991129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,4,balanced,1.22434663772583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,4,balanced,1.343290646870931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,32,power_law_1.2,0.09611520171165466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,4,balanced,1.8280266125996907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,4,balanced,2.4545920689900718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,1,power_law_1.2,2.6975807189941405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,32,power_law_1.2,0.09649279713630676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,4,balanced,4.782032012939453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,32,power_law_1.2,0.09904639720916748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,32,power_law_1.2,0.10135040283203126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,1,power_law_1.2,4.165311813354492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,32,power_law_1.2,0.10646400451660157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,32,power_law_1.2,0.11437439918518066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,32,power_law_1.2,0.1229632019996643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,32,power_law_1.2,0.1285375952720642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,1,power_law_1.2,8.207494354248047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,32,power_law_1.2,0.14449280500411987
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,32,power_law_1.2,0.1527232050895691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,32,power_law_1.2,0.1900607943534851
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,32,power_law_1.2,0.20836479663848878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,32,power_law_1.2,0.27543039321899415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,4,4,balanced,0.03762666632731756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,4,4,balanced,0.037861332297325134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,4,4,balanced,0.05624533196290334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,4,4,balanced,0.05426133175690969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,4,4,balanced,0.056608001391092934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,32,power_law_1.2,0.3390336036682129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,4,4,balanced,0.05598400036493937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,4,4,balanced,0.0581279993057251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,4,4,balanced,0.059690664211908974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,4,4,balanced,0.05807999769846598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,4,4,balanced,0.05856533348560333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,4,4,balanced,0.06145599981149038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,4,4,balanced,0.06242666641871134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,4,4,balanced,0.06355733176072438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,4,4,balanced,0.06539733211199443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,4,4,balanced,0.07419200241565704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,4,4,balanced,0.07565333445866902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,4,4,balanced,0.08075200021266937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,4,4,balanced,0.08758399883906047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,4,4,balanced,0.09645332892735799
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,4,4,balanced,0.12566933035850525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,4,4,balanced,0.1316106617450714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,4,4,balanced,0.16856533288955688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,4,4,balanced,0.20313066244125366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,4,4,balanced,0.2811253269513448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,32,power_law_1.2,0.47968640327453616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,4,4,balanced,0.34831468264261883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,4,4,balanced,0.49517865975697833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,4,4,balanced,0.6371786594390869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,4,4,balanced,0.7726453145345052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,4,4,balanced,1.1884960333506267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,4,4,balanced,2.2948800722757974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,32,power_law_1.2,0.5868415832519531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,32,power_law_1.2,0.8153535842895507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,32,power_law_1.2,1.0490559577941894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,1,2,balanced,0.03403733422358831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,1,2,balanced,0.037018666664759316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,1,2,balanced,0.052229334910710655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,1,2,balanced,0.07753600180149078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,1,2,balanced,0.1202186644077301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,1,2,balanced,0.11987732847531636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,1,2,balanced,0.12096533179283142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,1,2,balanced,0.11906133095423381
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,1,2,balanced,0.1200320025285085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,1,2,balanced,0.12034666538238525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,1,2,balanced,0.1218933363755544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,1,2,balanced,0.12461333473523457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,1,2,balanced,0.12404800454775493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,1,2,balanced,0.12563199798266092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,1,2,balanced,0.1323306659857432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,1,2,balanced,0.132341335217158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,1,2,balanced,0.1368160049120585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,1,2,balanced,0.14435733358065286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,2,balanced,0.15330132842063904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,32,power_law_1.2,1.3512639999389648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,2,balanced,0.17773866653442383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,2,balanced,0.18595200777053833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,2,balanced,0.2316640019416809
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,2,balanced,0.2606559991836548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,2,balanced,0.3546026547749837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,2,balanced,0.4127360184987386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,2,balanced,0.5708959897359213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,2,balanced,0.7237599690755209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,2,balanced,0.8805546760559082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,2,balanced,1.3751413027445476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,2,balanced,2.6642026901245117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,32,power_law_1.2,2.079635238647461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,32,power_law_1.2,4.4135486602783205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,1,32,power_law_1.2,0.026156800985336303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,1,32,power_law_1.2,0.024223999679088594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,1,32,power_law_1.2,0.02748799920082092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,1,32,power_law_1.2,0.026150399446487428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,1,32,power_law_1.2,0.02855679988861084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,1,32,power_law_1.2,0.02602880001068115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,1,32,power_law_1.2,0.026438400149345398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,1,32,power_law_1.2,0.030272001028060914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,1,32,power_law_1.2,0.03111039996147156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,1,32,power_law_1.2,0.030847999453544616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,1,32,power_law_1.2,0.031769600510597226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,1,32,power_law_1.2,0.032793599367141726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,1,32,power_law_1.2,0.032652801275253295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,1,32,power_law_1.2,0.03761920034885406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,1,32,power_law_1.2,0.038815999031066896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,1,32,power_law_1.2,0.04094719886779785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,2,power_law_1.2,0.09086719751358033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,1,32,power_law_1.2,0.0426367998123169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,1,32,power_law_1.2,0.04718720018863678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,2,power_law_1.2,0.08318079710006714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,32,power_law_1.2,0.05097600221633911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,32,power_law_1.2,0.05764480233192444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,32,power_law_1.2,0.062080001831054686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,2,power_law_1.2,0.09827839732170104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,32,power_law_1.2,0.08229119777679443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,2,power_law_1.2,0.13763200044631957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,32,power_law_1.2,0.10428800582885742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,16,power_law_1.01,0.050380802154541014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,32,power_law_1.2,0.12977919578552247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,2,power_law_1.2,0.1811776041984558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,16,power_law_1.01,0.09482240080833435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,32,power_law_1.2,0.14722559452056885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,2,power_law_1.2,0.30319359302520754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,32,power_law_1.2,0.2663935899734497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,16,power_law_1.01,0.08780159950256347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,2,power_law_1.2,0.32718720436096194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,32,power_law_1.2,0.35548160076141355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,16,power_law_1.01,0.05356159806251526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,2,power_law_1.2,0.34158079624176024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,32,power_law_1.2,0.44618878364562986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,16,power_law_1.01,0.05937920212745666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,2,power_law_1.2,0.3443711996078491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,32,power_law_1.2,0.7416639804840088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,2,power_law_1.2,0.3432255983352661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,16,power_law_1.01,0.06976640224456787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,2,power_law_1.2,0.36662399768829346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,16,power_law_1.01,0.07195519804954528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,32,power_law_1.2,1.3075648307800294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,2,power_law_1.2,0.3753599882125854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,16,power_law_1.01,0.08195199966430664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,2,power_law_1.2,0.3757375955581665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,2,power_law_1.2,0.3945472002029419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,16,power_law_1.01,0.0803264021873474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,2,power_law_1.2,0.4014080047607422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,16,power_law_1.01,0.0830784022808075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,2,power_law_1.2,0.41224961280822753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,16,power_law_1.01,0.08547840118408204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,2,power_law_1.2,0.42201600074768064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,16,power_law_1.01,0.08581759929656982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,2,power_law_1.2,0.4796480178833008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,2,power_law_1.2,0.5201216220855713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,16,power_law_1.01,0.0880895972251892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,2,power_law_1.2,0.6394559860229492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,16,power_law_1.01,0.0945792019367218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,2,power_law_1.2,0.6669695854187012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,16,power_law_1.01,0.10005760192871094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,2,power_law_1.2,0.8297599792480469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,2,8,power_law_1.2,0.05777919888496399
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,16,power_law_1.01,0.1050879955291748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,2,8,power_law_1.2,0.05917440056800842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,2,8,power_law_1.2,0.05655679702758789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,2,power_law_1.2,0.9879296302795411
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,2,8,power_law_1.2,0.05742080211639404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,2,8,power_law_1.2,0.05803520083427429
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,2,8,power_law_1.2,0.061990398168563846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,2,8,power_law_1.2,0.06280959844589233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,2,8,power_law_1.2,0.07292159795761108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,16,power_law_1.01,0.1151360034942627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,2,8,power_law_1.2,0.2614527940750122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,2,power_law_1.2,1.2354496002197266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,2,8,power_law_1.2,0.3240256071090698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,2,8,power_law_1.2,0.06513280272483826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,2,8,power_law_1.2,0.08394880294799804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,2,8,power_law_1.2,0.08078719973564148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,2,8,power_law_1.2,0.07979519963264466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,16,power_law_1.01,0.12728320360183715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,2,8,power_law_1.2,0.09742079973220825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,2,8,power_law_1.2,0.10856319665908813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,2,power_law_1.2,1.5724991798400878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,2,8,power_law_1.2,0.09757440090179444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,2,8,power_law_1.2,0.12454400062561036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,2,8,power_law_1.2,0.155840003490448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,2,8,power_law_1.2,0.21665279865264891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,16,power_law_1.01,0.13582079410552977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,2,8,power_law_1.2,0.24567039012908937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,2,8,power_law_1.2,0.2929919958114624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,2,8,power_law_1.2,0.42698240280151367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,2,power_law_1.2,2.3267648696899412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,2,8,power_law_1.2,0.6258944034576416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,2,8,power_law_1.2,0.7249855995178223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,16,power_law_1.01,0.16850559711456298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,2,8,power_law_1.2,1.2297280311584473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,2,8,power_law_1.2,1.5545408248901367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,16,power_law_1.01,0.19018880128860474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,2,power_law_1.2,3.0131263732910156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,2,8,power_law_1.2,1.9677631378173828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,16,power_law_1.01,0.25369598865509035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,2,8,power_law_1.2,3.0966720581054688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,2,power_law_1.2,3.4256641387939455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,16,power_law_1.01,0.30390400886535646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,2,8,power_law_1.2,5.823231887817383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,16,power_law_1.01,0.4066624164581299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,2,power_law_1.2,5.216083145141601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,16,power_law_1.01,0.5151423931121826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,16,power_law_1.01,0.7360832214355468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,16,power_law_1.01,0.9164671897888184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,2,power_law_1.2,10.710348510742188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,16,power_law_1.01,1.1133760452270507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,16,power_law_1.01,1.7525568008422852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,2,power_law_1.01,0.03383679986000061
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,2,power_law_1.01,0.03113600015640259
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,2,power_law_1.01,0.03498879969120026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,2,power_law_1.01,0.0385343998670578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,2,power_law_1.01,0.040659201145172116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,16,power_law_1.01,3.455859375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,2,power_law_1.01,0.05445759892463684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,2,power_law_1.01,0.056531202793121335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,2,power_law_1.01,0.058182400465011594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,2,power_law_1.01,0.059571200609207155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,2,power_law_1.01,0.060915201902389526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,2,power_law_1.01,0.06410239934921265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,2,power_law_1.01,0.0676479995250702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,2,power_law_1.01,0.06922879815101624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,16,power_law_1.01,0.05503360033035278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,2,power_law_1.01,0.06965759992599488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,2,power_law_1.01,0.07712000012397766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,2,power_law_1.01,0.07980160117149353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,16,power_law_1.01,0.0497408002614975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,2,power_law_1.01,0.08591359853744507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,2,power_law_1.01,0.10085760354995728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,16,power_law_1.01,0.04552960097789764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,2,power_law_1.01,0.11122560501098633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,2,power_law_1.01,0.13866879940032958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,16,power_law_1.01,0.04636160135269165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,2,power_law_1.01,0.1523519992828369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,2,power_law_1.01,0.19563519954681396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,16,power_law_1.01,0.04766719937324524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,2,power_law_1.01,0.24726400375366211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,16,power_law_1.01,0.05002880096435547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,2,power_law_1.01,0.3743743896484375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,1,32,power_law_1.01,0.026169601082801818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,16,power_law_1.01,0.05057920217514038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,2,power_law_1.01,0.4584320068359375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,1,32,power_law_1.01,0.02587519884109497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,2,power_law_1.01,0.667302417755127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,16,power_law_1.01,0.05151360034942627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,1,32,power_law_1.01,0.02717440128326416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,1,32,power_law_1.01,0.02720000147819519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,16,power_law_1.01,0.05121279954910278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,2,power_law_1.01,0.8726400375366211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,1,32,power_law_1.01,0.028313601016998292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,1,32,power_law_1.01,0.025567999482154845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,16,power_law_1.01,0.05189759731292724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,1,32,power_law_1.01,0.026150399446487428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,2,power_law_1.01,1.1245247840881347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,1,32,power_law_1.01,0.028812798857688903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,16,power_law_1.01,0.05245440006256104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,1,32,power_law_1.01,0.030003198981285097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,1,32,power_law_1.01,0.030079999566078187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,16,power_law_1.01,0.05418879985809326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,2,power_law_1.01,1.6770048141479492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,1,32,power_law_1.01,0.031865599751472476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,1,32,power_law_1.01,0.0324864000082016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,16,power_law_1.01,0.05508480072021484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,1,32,power_law_1.01,0.032691198587417605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,16,power_law_1.01,0.060012799501419065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,1,32,power_law_1.01,0.03537920117378235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,1,32,power_law_1.01,0.03969280123710632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,16,power_law_1.01,0.0664192020893097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,1,32,power_law_1.01,0.03961600065231323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,2,power_law_1.01,3.084294319152832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,1,32,power_law_1.01,0.041382399201393125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,16,power_law_1.01,0.06924800276756286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,1,32,power_law_1.01,0.04311679899692535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,16,power_law_1.01,0.06949759721755981
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,32,power_law_1.01,0.04738560020923614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,32,power_law_1.01,0.05140479803085327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,16,power_law_1.01,0.0791487991809845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,32,power_law_1.01,0.06051200032234192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,32,power_law_1.01,0.06839039921760559
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,16,power_law_1.01,0.08497920036315917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,32,power_law_1.01,0.07975040078163147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,16,power_law_1.01,0.10503679513931274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,32,power_law_1.01,0.10810240507125854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,16,power_law_1.01,0.12764159440994263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,32,power_law_1.01,0.12993279695510865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,16,power_law_1.01,0.15191680192947388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,32,power_law_1.01,0.16050560474395753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,16,power_law_1.01,0.19408639669418334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,32,power_law_1.01,0.22633600234985352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,16,power_law_1.01,0.23096320629119874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,32,power_law_1.01,0.2436352014541626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,16,power_law_1.01,0.285318398475647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,32,power_law_1.01,0.35824639797210694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,2,1,balanced,0.052186667919158936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,16,power_law_1.01,0.39829120635986326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,2,1,balanced,0.07008533179759979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,2,1,balanced,0.10017066200574239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,2,1,balanced,0.09994133313496907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,2,1,balanced,0.1018986701965332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,2,1,balanced,0.10129066308339436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,2,1,balanced,0.1018399993578593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,2,1,balanced,0.10333333412806193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,2,1,balanced,0.10363200306892395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,2,1,balanced,0.10269332925478618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,2,1,balanced,0.10507200161616008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,2,1,balanced,0.10504532853762309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,2,1,balanced,0.10982933640480042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,2,1,balanced,0.11322666207949321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,2,1,balanced,0.12286933263142903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,2,1,balanced,0.12637333075205484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,2,1,balanced,0.13404267032941183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,2,1,balanced,0.1632213294506073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,2,1,balanced,0.17289066314697266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,2,1,balanced,0.22123199701309204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,2,1,balanced,0.2669493357340495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,2,1,balanced,0.35949333508809406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,32,power_law_1.01,0.670963191986084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,2,1,balanced,0.4466826518376668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,16,power_law_1.01,0.5955008029937744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,2,1,balanced,0.6242080132166544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,2,1,balanced,0.7655626932779948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,2,1,balanced,1.1094613075256348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,2,1,balanced,1.454485257466634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,2,1,balanced,1.8011040687561035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,16,power_law_1.01,0.6431039810180664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,2,1,balanced,2.802149454752604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,2,1,balanced,5.528234481811523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,16,power_law_1.01,0.9668928146362304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,16,power_law_1.01,1.9074432373046875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,1,power_law_1.2,0.07816960215568543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,1,power_law_1.2,0.11725440025329589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,1,power_law_1.2,0.11282559633255004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,4,power_law_1.2,0.08674560189247131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,1,power_law_1.2,0.1566848039627075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,4,power_law_1.2,0.08536319732666016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,1,power_law_1.2,0.20991361141204834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,1,power_law_1.2,0.27093119621276857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,1,power_law_1.2,0.03526400029659271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,4,power_law_1.2,0.08142719864845276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,1,power_law_1.2,0.40087041854858396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,1,power_law_1.2,0.041126400232315063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,1,power_law_1.2,0.42273921966552735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,4,power_law_1.2,0.1012992024421692
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,1,power_law_1.2,0.04979200065135956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,1,power_law_1.2,0.43393921852111816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,4,power_law_1.2,0.12199039459228515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,1,power_law_1.2,0.06236799955368042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,1,power_law_1.2,0.46360321044921876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,1,power_law_1.2,0.07783679962158203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,1,power_law_1.2,0.4795839786529541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,4,power_law_1.2,0.16704000234603883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,1,power_law_1.2,0.1158784031867981
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,1,power_law_1.2,0.5114431858062745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,1,power_law_1.2,0.12557439804077147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,1,power_law_1.2,0.5178559780120849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,4,power_law_1.2,0.1775871992111206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,1,power_law_1.2,0.13165440559387206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,1,power_law_1.2,0.5438591957092285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,1,power_law_1.2,0.13812479972839356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,4,power_law_1.2,0.18854399919509887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,1,power_law_1.2,0.573632001876831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,1,power_law_1.2,0.13985919952392578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,1,power_law_1.2,0.589734411239624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,4,power_law_1.2,0.19055360555648804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,1,power_law_1.2,0.14355839490890504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,1,power_law_1.2,0.630521583557129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,1,power_law_1.2,0.14972800016403198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,4,power_law_1.2,0.19584640264511108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,1,power_law_1.2,0.15318399667739868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,1,power_law_1.2,0.7062079906463623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,1,power_law_1.2,0.16150399446487426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,1,power_law_1.2,0.7820096015930176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,4,power_law_1.2,0.20250239372253417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,1,power_law_1.2,0.17112319469451903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,1,power_law_1.2,0.9211711883544922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,1,power_law_1.2,0.17934080362319946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,4,power_law_1.2,0.21293439865112304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,1,power_law_1.2,0.1943168044090271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,1,power_law_1.2,0.9529984474182129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,1,power_law_1.2,0.23084158897399903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,4,power_law_1.2,0.22544000148773194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,1,power_law_1.2,1.1550975799560548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,1,power_law_1.2,0.2549504041671753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,1,power_law_1.2,0.324780797958374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,1,power_law_1.2,1.383148765563965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,4,power_law_1.2,0.23116800785064698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,1,power_law_1.2,0.366048002243042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,1,power_law_1.2,1.7775487899780273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,1,power_law_1.2,0.48604798316955566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,4,power_law_1.2,0.2434623956680298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,1,power_law_1.2,0.601804780960083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,1,power_law_1.2,2.195155143737793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,4,power_law_1.2,0.24737279415130614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,1,power_law_1.2,0.8260224342346192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,1,power_law_1.2,1.0483648300170898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,1,power_law_1.2,2.9874879837036135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,4,power_law_1.2,0.2707520008087158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,1,power_law_1.2,1.4893888473510741
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,4,power_law_1.2,0.311955189704895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,1,power_law_1.2,3.892959976196289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,1,power_law_1.2,1.9779327392578125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,4,power_law_1.2,0.3543231964111328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,1,power_law_1.2,4.701651382446289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,1,power_law_1.2,2.4297407150268553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,4,power_law_1.2,0.44261760711669923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,4,power_law_1.2,0.4999807834625244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,1,power_law_1.2,3.709056091308594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,1,power_law_1.2,7.023193359375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,4,power_law_1.2,0.6576576232910156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,4,power_law_1.2,0.7675392150878906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,1,power_law_1.2,7.295667266845703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,1,power_law_1.2,13.644473266601562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,4,power_law_1.2,1.0264191627502441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,4,power_law_1.2,1.3425984382629395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,4,power_law_1.2,1.9349056243896485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,4,power_law_1.2,2.49816951751709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,4,power_law_1.01,0.04256640076637268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,4,power_law_1.01,0.03694719970226288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,1,power_law_1.2,0.03743999898433685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,2,4,power_law_1.2,2.821049690246582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,1,power_law_1.2,0.04291839897632599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,4,power_law_1.01,0.0365119993686676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,1,power_law_1.2,0.04493440091609955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,1,power_law_1.2,0.058387202024459836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,4,power_law_1.01,0.03950079977512359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,1,power_law_1.2,0.07124480009078979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,4,power_law_1.01,0.040320000052452086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,1,power_law_1.2,0.11136640310287475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,1,power_law_1.2,0.11543040275573731
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,4,power_law_1.01,0.04711039960384369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,1,power_law_1.2,0.11806720495223999
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,4,power_law_1.01,0.04674560129642487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,1,power_law_1.2,0.12247680425643921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,1,power_law_1.2,0.12776960134506227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,2,4,power_law_1.2,5.028972625732422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,4,power_law_1.01,0.0482367992401123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,1,power_law_1.2,0.13244800567626952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,4,power_law_1.01,0.04795520007610321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,1,power_law_1.2,0.13802239894866944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,4,power_law_1.01,0.04831359982490539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,1,power_law_1.2,0.1391487956047058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,1,power_law_1.2,0.14614399671554565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,4,power_law_1.01,0.049030399322509764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,1,power_law_1.2,0.15688320398330688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,1,power_law_1.2,0.16240639686584474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,4,power_law_1.01,0.05120000243186951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,1,power_law_1.2,0.17319040298461913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,4,power_law_1.01,0.052716797590255736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,1,power_law_1.2,0.1970944046974182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,4,power_law_1.01,0.055219197273254396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,1,power_law_1.2,0.21904640197753905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,1,power_law_1.2,0.271776008605957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,4,power_law_1.01,0.06097279787063599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,1,power_law_1.2,0.3106816053390503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,4,power_law_1.01,0.06499840021133423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,1,power_law_1.2,0.41211519241333006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,4,power_law_1.01,0.06310399770736694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,1,power_law_1.2,0.5041791915893554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,4,power_law_1.01,0.07253119945526124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,1,power_law_1.2,0.6946368217468262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,2,4,power_law_1.2,9.82265625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,4,power_law_1.01,0.08120959997177124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,1,power_law_1.2,0.8825152397155762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,4,power_law_1.01,0.10053759813308716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,1,power_law_1.2,1.2637887954711915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,4,power_law_1.01,0.11463040113449097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,4,power_law_1.01,0.14469759464263915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,1,power_law_1.2,1.6745344161987306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,4,power_law_1.01,0.18209919929504395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,1,power_law_1.2,2.0694271087646485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,4,power_law_1.01,0.24158720970153807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,4,power_law_1.01,0.3428992033004761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,1,power_law_1.2,3.132761573791504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,4,power_law_1.01,0.4779776096343994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,4,power_law_1.01,0.6220672130584717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,4,power_law_1.01,0.7490240097045898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,1,power_law_1.2,6.072844696044922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,64,power_law_1.2,0.03649280071258545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,4,power_law_1.01,1.2142271995544434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,64,power_law_1.2,0.03490560054779053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,64,power_law_1.2,0.03341439962387085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,64,power_law_1.2,0.033055999875068666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,64,power_law_1.2,0.03370240032672882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,64,power_law_1.2,0.03426559865474701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,4,power_law_1.01,2.1589567184448244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,64,power_law_1.2,0.03500800132751465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,64,power_law_1.2,0.0341376006603241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,64,power_law_1.2,0.03404799997806549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,64,power_law_1.2,0.034220799803733826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,64,power_law_1.2,0.035129600763320924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,64,power_law_1.2,0.03573119938373566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,64,power_law_1.2,0.03601920008659363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,64,power_law_1.2,0.03723520040512085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,64,power_law_1.2,0.03986560106277466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,64,power_law_1.2,0.039929598569869995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,64,power_law_1.2,0.041152000427246094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,64,power_law_1.2,0.0446399986743927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,64,power_law_1.2,0.049369600415229795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,4,4,power_law_1.01,0.07505279779434204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,4,4,power_law_1.01,0.0702015995979309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,4,4,power_law_1.01,0.0793727993965149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,64,power_law_1.2,0.05373439788818359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,4,4,power_law_1.01,0.07541760206222534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,4,4,power_law_1.01,0.07581440210342408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,4,4,power_law_1.01,0.07613440155982971
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,64,power_law_1.2,0.061440002918243405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,4,4,power_law_1.01,0.07904000282287597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,4,4,power_law_1.01,0.08380799889564514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,4,4,power_law_1.01,0.08421760201454162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,4,4,power_law_1.01,0.09125760197639465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,64,power_law_1.2,0.07539839744567871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,4,4,power_law_1.01,0.08410239815711976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,4,4,power_law_1.01,0.09249280095100403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,4,4,power_law_1.01,0.09510400295257568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,64,power_law_1.2,0.09127680063247681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,4,4,power_law_1.01,0.09683840274810791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,4,4,power_law_1.01,0.11413120031356812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,4,4,power_law_1.01,0.1168895959854126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,4,4,power_law_1.01,0.13045120239257812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,64,power_law_1.2,0.12550400495529174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,4,4,power_law_1.01,0.15012480020523072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,4,4,power_law_1.01,0.18875520229339598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,4,4,power_law_1.01,0.23708159923553468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,4,4,power_law_1.01,0.29254400730133057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,64,power_law_1.2,0.16268800497055053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,4,4,power_law_1.01,0.35866239070892336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,4,8,power_law_1.01,0.034246399998664856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,4,4,power_law_1.01,0.5384768009185791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,4,8,power_law_1.01,0.03389439880847931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,64,power_law_1.2,0.24057600498199463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,4,4,power_law_1.01,0.7931968212127686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,4,8,power_law_1.01,0.03304960131645203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,4,4,power_law_1.01,1.1090304374694824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,4,8,power_law_1.01,0.03326080143451691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,4,8,power_law_1.01,0.03375999927520752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,4,4,power_law_1.01,1.3575167655944824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,64,power_law_1.2,0.3435456037521362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,4,8,power_law_1.01,0.0350847989320755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,4,8,power_law_1.01,0.036774399876594546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,4,4,power_law_1.01,1.6949567794799805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,4,8,power_law_1.01,0.03735679984092712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,4,8,power_law_1.01,0.0379584014415741
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,64,power_law_1.2,0.3077631950378418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,4,4,power_law_1.01,1.9249408721923829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,4,8,power_law_1.01,0.04257279932498932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,4,8,power_law_1.01,0.04190720021724701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,4,4,power_law_1.01,2.5635135650634764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,4,8,power_law_1.01,0.04390400052070618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,64,power_law_1.2,0.701913595199585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,4,8,power_law_1.01,0.04665600061416626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,4,8,power_law_1.01,0.04513919949531555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,4,8,power_law_1.01,0.05246080160140991
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,4,8,power_law_1.01,0.0566976010799408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,4,4,power_law_1.01,7.586988830566407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,4,8,power_law_1.01,0.062003201246261595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,4,8,power_law_1.01,0.07451519966125489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,64,power_law_1.2,1.4562687873840332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,4,8,power_law_1.01,0.0814848005771637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,4,8,power_law_1.01,0.10663039684295654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,4,8,power_law_1.01,0.12791680097579955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,4,8,power_law_1.01,0.15768959522247314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,4,8,power_law_1.01,0.20241279602050782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,4,8,power_law_1.01,0.27854080200195314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,4,8,power_law_1.01,0.34226560592651367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,4,8,power_law_1.01,0.49120640754699707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,4,8,power_law_1.01,0.6416128158569336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,4,8,power_law_1.01,0.8118975639343262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,4,8,power_law_1.01,1.2263168334960937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,4,8,power_law_1.01,2.4182336807250975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,128,power_law_1.2,0.036032000184059144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,16,power_law_1.01,0.03546879887580871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,128,power_law_1.2,0.03562879860401154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,16,power_law_1.01,0.0298880010843277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,128,power_law_1.2,0.036287999153137206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,16,power_law_1.01,0.029043200612068176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,128,power_law_1.2,0.037376001477241516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,16,power_law_1.01,0.02876800000667572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,128,power_law_1.2,0.03646720051765442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,16,power_law_1.01,0.029875200986862183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,128,power_law_1.2,0.037503999471664426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,16,power_law_1.01,0.031007999181747438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,128,power_law_1.2,0.03761920034885406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,16,power_law_1.01,0.03125759959220886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,128,power_law_1.2,0.03781760036945343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,16,power_law_1.01,0.031564798951148984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,16,power_law_1.01,0.031744000315666196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,128,power_law_1.2,0.03787519931793213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,128,power_law_1.2,0.03797119855880737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,16,power_law_1.01,0.0322816014289856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,128,power_law_1.2,0.038873600959777835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,16,power_law_1.01,0.03248000144958496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,128,power_law_1.2,0.03933440148830414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,16,power_law_1.01,0.03369599878787995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,128,power_law_1.2,0.039980798959732056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,16,power_law_1.01,0.03473919928073883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,128,power_law_1.2,0.040064001083374025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,16,power_law_1.01,0.03532800078392029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,128,power_law_1.2,0.0435263991355896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,16,power_law_1.01,0.03830400109291077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,64,power_law_1.01,0.07923840284347534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,128,power_law_1.2,0.046412798762321475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,16,power_law_1.01,0.04032639861106872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,128,power_law_1.2,0.04656000137329101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,16,power_law_1.01,0.0414463996887207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,128,power_law_1.2,0.05082240104675293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,16,power_law_1.01,0.04420480132102966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,128,power_law_1.2,0.05559039711952209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,16,power_law_1.01,0.048198398947715757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,128,power_law_1.2,0.0667967975139618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,16,power_law_1.01,0.05717759728431702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,128,power_law_1.2,0.07642239928245545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,16,power_law_1.01,0.061292797327041626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,64,power_law_1.01,0.08254719972610473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,128,power_law_1.2,0.11859840154647827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,16,power_law_1.01,0.07624959945678711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,128,power_law_1.2,0.14303359985351563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,16,power_law_1.01,0.09109119772911071
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,128,power_law_1.2,0.2059391975402832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,16,power_law_1.01,0.12551039457321167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,128,power_law_1.2,0.2549184083938599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,16,power_law_1.01,0.1641152024269104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,64,power_law_1.01,0.06074240207672119
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,128,power_law_1.2,0.370252799987793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,16,power_law_1.01,0.23290879726409913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,128,power_law_1.2,0.4953343868255615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,16,power_law_1.01,0.276908802986145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,128,power_law_1.2,0.6136703968048096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,16,power_law_1.01,0.35374720096588136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,64,power_law_1.01,0.06346240043640136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,16,power_law_1.01,0.5561279773712158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,128,power_law_1.2,0.9531904220581054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,64,power_law_1.01,0.07402880191802978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,16,power_law_1.01,1.105836772918701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,128,power_law_1.2,1.8630336761474608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,64,power_law_1.01,0.07836160063743591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,64,power_law_1.01,0.07697920203208923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,64,power_law_1.01,0.07686399817466735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,64,power_law_1.01,0.077920001745224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,64,power_law_1.01,0.07935360074043274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,64,power_law_1.01,0.08010879755020142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,1,2,balanced,0.1288320024808248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,1,2,balanced,0.21522667010625204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,1,2,balanced,0.4323306481043498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,1,2,balanced,0.3959733247756958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,1,2,balanced,0.3933440049489339
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,1,2,balanced,0.39076801141103107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,1,2,balanced,0.39500268300374347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,1,2,balanced,0.394538680712382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,1,2,balanced,0.3938080072402954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,1,2,balanced,0.3964159886042277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,1,2,balanced,0.40547200043996173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,1,2,balanced,0.40036265055338544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,1,2,balanced,0.4039466778437297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,1,2,balanced,0.43212799231211346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,1,2,balanced,0.43929600715637207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,1,2,balanced,0.441861351331075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,1,2,balanced,0.4477653503417969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,1,2,balanced,0.5355093479156494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,1,2,balanced,0.5324106613794962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,64,power_law_1.01,0.08390399813652039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,8,power_law_1.01,0.08634880185127258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,1,2,balanced,0.6988106568654379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,1,2,balanced,0.7629439830780029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,1,2,balanced,1.019488016764323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,1,2,balanced,1.2733439604441326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,1,2,balanced,1.8106667200724285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,1,2,balanced,2.3913920720418296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,8,power_law_1.01,0.1208575963973999
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,64,power_law_1.01,0.08483840227127075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,1,2,balanced,3.5339787801106772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,1,2,balanced,4.684672037760417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,32,balanced,0.053530668218930565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,32,balanced,0.04585599899291992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,32,balanced,0.04378133515516917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,32,balanced,0.052239999175071716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,32,balanced,0.07450133562088013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,32,balanced,0.08149866759777069
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,32,balanced,0.08194666604201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,32,balanced,0.08266666531562805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,32,balanced,0.08003733555475871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,8,power_law_1.01,0.0683456003665924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,32,balanced,0.08275733391443889
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,32,balanced,0.08195200065771739
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,32,balanced,0.08213866750399272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,32,balanced,0.08296533425649007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,32,balanced,0.08480532964070638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,32,balanced,0.0885653297106425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,32,balanced,0.09047999978065491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,1,2,balanced,5.862042744954427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,32,balanced,0.09297066926956177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,32,balanced,0.10022399822870891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,32,balanced,0.09984532992045085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,32,balanced,0.11574932932853699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,64,power_law_1.01,0.08591359853744507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,32,balanced,0.11404800415039062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,32,balanced,0.14033066232999167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,32,balanced,0.14543466766675314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,32,balanced,0.20508799950281778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,32,balanced,0.2165493369102478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,32,balanced,0.2818293372790019
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,32,balanced,0.34913067022959393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,32,balanced,0.4074666500091553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,32,balanced,0.5946293274561564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,8,power_law_1.01,0.08238080143928528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,32,balanced,1.1115893522898357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,1,2,balanced,9.44221305847168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,64,power_law_1.01,0.09331200122833253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,8,power_law_1.01,0.0976639986038208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,64,power_law_1.01,0.09482880234718323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,8,power_law_1.01,0.13175679445266725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,1,2,balanced,19.441275278727215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,8,power_law_1.01,0.13201919794082642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,64,power_law_1.01,0.10224640369415283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,8,power_law_1.01,0.13492480516433716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,64,power_law_1.01,0.12040319442749023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,8,power_law_1.01,0.14199039936065674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,64,power_law_1.01,0.12449920177459717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,8,power_law_1.01,0.14039039611816406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,8,power_law_1.01,0.1456768035888672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,64,power_law_1.01,0.1568384051322937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,64,power_law_1.2,0.082777601480484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,8,power_law_1.01,0.15009280443191528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,64,power_law_1.01,0.1728384017944336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,64,power_law_1.2,0.08147839903831482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,8,power_law_1.01,0.14883840084075928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,4,balanced,0.05203733344872793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,4,balanced,0.05593066910902659
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,4,balanced,0.06798933446407318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,4,balanced,0.096778670946757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,4,balanced,0.13134933511416116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,4,balanced,0.16301332910855612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,4,balanced,0.1630826691786448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,4,balanced,0.16355733076731363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,4,balanced,0.16591999928156534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,4,balanced,0.16594133774439493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,64,power_law_1.2,0.07896320223808288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,4,balanced,0.16766399145126343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,4,balanced,0.170250674088796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,4,balanced,0.1721173326174418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,4,balanced,0.17529600858688354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,4,balanced,0.18281600872675577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,4,balanced,0.1894986629486084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,4,balanced,0.19562667608261108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,4,balanced,0.21929067373275757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,4,balanced,0.23031999667485556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,8,power_law_1.01,0.15499520301818848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,4,balanced,0.29761600494384766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,4,balanced,0.31808000802993774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,4,balanced,0.4790026744206746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,64,power_law_1.01,0.22615039348602295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,4,balanced,0.4725866715113322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,64,power_law_1.2,0.06319360136985779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,4,balanced,0.6821866830190023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,4,balanced,0.7771680355072021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,4,balanced,1.1321760018666585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,4,balanced,1.3797972997029622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,8,power_law_1.01,0.16633599996566772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,64,power_law_1.2,0.0636352002620697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,4,4,balanced,1.7333760261535645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,4,4,balanced,2.580746650695801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,64,power_law_1.01,0.23315839767456054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,64,power_law_1.2,0.07002239823341369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,8,power_law_1.01,0.16812800168991088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,4,4,balanced,5.165338516235352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,64,power_law_1.2,0.07969279885292054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,8,power_law_1.01,0.18001919984817505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,64,power_law_1.01,0.31826560497283934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,64,power_law_1.2,0.08112639784812928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,8,power_law_1.01,0.1997056007385254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,64,power_law_1.2,0.0773248016834259
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,64,power_law_1.2,0.08046079874038696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,64,power_law_1.01,0.3831039905548096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,8,power_law_1.01,0.22069120407104492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,64,power_law_1.2,0.07921919822692872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,8,power_law_1.01,0.2655872106552124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,64,power_law_1.2,0.08351359963417053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,64,power_law_1.01,0.5739840030670166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,8,power_law_1.01,0.28854401111602784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,64,power_law_1.2,0.08309760093688964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,64,power_law_1.2,0.08277119994163513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,8,power_law_1.01,0.3587071895599365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,64,power_law_1.2,0.09056000113487243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,64,power_law_1.01,0.6639616012573242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,32,power_law_1.01,0.03814400136470795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,32,power_law_1.01,0.0353983998298645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,64,power_law_1.2,0.0931007981300354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,8,power_law_1.01,0.4035200119018555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,32,power_law_1.01,0.03395200073719025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,32,power_law_1.01,0.03248000144958496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,64,power_law_1.2,0.09950079917907714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,32,power_law_1.01,0.03342080116271973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,32,power_law_1.01,0.03397760093212128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,8,power_law_1.01,0.5493760108947754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,64,power_law_1.2,0.1129472017288208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,32,power_law_1.01,0.03436160087585449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,1,64,power_law_1.01,0.794163179397583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,32,power_law_1.01,0.03475840091705322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,64,power_law_1.2,0.12149759531021118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,32,power_law_1.01,0.03463680148124695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,32,power_law_1.01,0.03506560027599335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,8,power_law_1.01,0.6772096157073975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,64,power_law_1.2,0.14797439575195312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,32,power_law_1.01,0.03530240058898926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,32,power_law_1.01,0.036236798763275145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,32,power_law_1.01,0.03672960102558136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,64,power_law_1.2,0.18027520179748535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,32,power_law_1.01,0.037574398517608645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,32,power_law_1.01,0.03980799913406372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,8,power_law_1.01,0.9095423698425293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,64,power_law_1.2,0.23363199234008789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,32,power_law_1.01,0.040761598944664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,1,64,power_law_1.01,1.2214847564697267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,32,power_law_1.01,0.04196479916572571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,64,power_law_1.2,0.2901632070541382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,32,power_law_1.01,0.04457600116729736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,32,power_law_1.01,0.04834559857845307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,32,power_law_1.01,0.05259519815444946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,64,power_law_1.2,0.39073278903961184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,8,power_law_1.01,1.0853952407836913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,32,power_law_1.01,0.05935360193252563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,32,power_law_1.01,0.07158399820327759
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,32,power_law_1.01,0.07422720193862915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,64,power_law_1.2,0.5077568054199219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,32,power_law_1.01,0.11106560230255128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,32,power_law_1.01,0.1376255989074707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,64,power_law_1.2,0.7209343910217285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,4,8,power_law_1.01,1.267251205444336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,32,power_law_1.01,0.18228479623794555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,32,power_law_1.01,0.23560960292816163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,64,power_law_1.2,0.9647168159484864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,1,64,power_law_1.01,2.462828826904297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,32,power_law_1.01,0.28707840442657473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,64,power_law_1.2,1.1843071937561036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,32,power_law_1.01,0.4935296058654785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,4,8,power_law_1.01,1.9839487075805664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,64,power_law_1.2,1.867673683166504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,32,power_law_1.01,0.9361215591430664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,64,power_law_1.2,3.8629249572753905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,4,8,power_law_1.01,4.346656036376953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,8,4,power_law_1.01,0.05596799850463867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,8,4,power_law_1.01,0.05803520083427429
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,8,4,power_law_1.01,0.056806397438049314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,8,4,power_law_1.01,0.06366080045700073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,8,4,power_law_1.01,0.06813439726829529
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,4,power_law_1.2,0.0469760000705719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,8,4,power_law_1.01,0.0798143982887268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,4,power_law_1.2,0.052671998739242554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,8,4,power_law_1.01,0.08432000279426574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,8,4,power_law_1.01,0.08818560242652893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,4,power_law_1.2,0.05161600112915039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,8,4,power_law_1.01,0.08823680281639099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,8,8,balanced,0.03751999884843826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,8,8,balanced,0.03749866783618927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,8,8,balanced,0.03741333385308584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,8,8,balanced,0.035546667873859406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,8,8,balanced,0.03616533428430557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,4,power_law_1.2,0.05459200143814087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,8,8,balanced,0.03700266778469086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,8,8,balanced,0.03750933210055033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,8,8,balanced,0.03760000069936117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,8,8,balanced,0.036805334190527596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,8,8,balanced,0.03775999943415324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,8,8,balanced,0.03933866570393244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,8,8,balanced,0.037717332442601524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,8,8,balanced,0.03953066716591517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,8,8,balanced,0.03948266555865606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,8,4,power_law_1.01,0.09137279987335205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,8,8,balanced,0.043893332282702126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,8,8,balanced,0.045498669147491455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,8,8,balanced,0.04573333263397217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,8,8,balanced,0.048010667165120445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,8,8,balanced,0.05374933282534281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,8,8,balanced,0.05815466741720835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,8,8,balanced,0.06422933439413707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,8,8,balanced,0.08162133395671844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,8,8,balanced,0.10114666819572449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,8,8,balanced,0.12384000420570374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,8,8,balanced,0.14453333616256714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,8,8,balanced,0.19498133659362793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,8,8,balanced,0.24305067459742227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,4,power_law_1.2,0.06182399988174438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,8,8,balanced,0.29160000880559284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,8,4,power_law_1.01,0.09168639779090881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,8,8,balanced,0.4332746664683024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,8,8,balanced,0.8091626962025961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,8,4,power_law_1.01,0.09465600252151489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,4,power_law_1.2,0.07382400035858154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,8,power_law_1.2,0.18617600202560425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,8,4,power_law_1.01,0.09731199741363525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,4,power_law_1.2,0.07926399707794189
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,8,4,power_law_1.01,0.10138880014419556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,8,power_law_1.2,0.12421120405197143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,4,power_law_1.2,0.09106559753417968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,8,4,power_law_1.01,0.1120959997177124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,8,4,power_law_1.01,0.11338239908218384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,4,power_law_1.2,0.09370239973068237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,8,power_law_1.2,0.11192959547042847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,8,4,power_law_1.01,0.11754239797592163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,4,power_law_1.2,0.09525759816169739
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,8,4,power_law_1.01,0.1376512050628662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,8,power_law_1.2,0.17487360239028932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,8,4,power_law_1.01,0.15101439952850343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,4,power_law_1.2,0.10079360008239746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,8,4,power_law_1.01,0.17493760585784912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,8,power_law_1.2,0.21159679889678956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,4,power_law_1.2,0.10687999725341797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,8,4,power_law_1.01,0.20631039142608643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,4,power_law_1.2,0.10780160427093506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,8,power_law_1.2,0.32894721031188967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,8,4,power_law_1.01,0.2805056095123291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,4,4,power_law_1.01,0.03902080059051514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,4,power_law_1.2,0.11492480039596557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,4,4,power_law_1.01,0.03825919926166534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,8,4,power_law_1.01,0.3419519901275635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,4,4,power_law_1.01,0.035750401020050046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,8,power_law_1.2,0.3600575923919678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,4,4,power_law_1.01,0.03573760092258453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,4,4,power_law_1.01,0.03654400110244751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,4,4,power_law_1.01,0.037503999471664426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,4,4,power_law_1.01,0.037625598907470706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,4,power_law_1.2,0.1308351993560791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,4,4,power_law_1.01,0.03901439905166626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,8,4,power_law_1.01,0.4524543762207031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,4,4,power_law_1.01,0.039852800965309146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,4,4,power_law_1.01,0.040582400560379026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,4,4,power_law_1.01,0.042483198642730716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,4,4,power_law_1.01,0.04537599980831146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,8,power_law_1.2,0.39166080951690674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,4,4,power_law_1.01,0.04430719912052154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,4,4,power_law_1.01,0.04588800072669983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,4,power_law_1.2,0.13029119968414307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,4,4,power_law_1.01,0.0553600013256073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,8,4,power_law_1.01,0.6155136108398438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,4,4,power_law_1.01,0.05768960118293762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,4,4,power_law_1.01,0.06947839856147767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,4,4,power_law_1.01,0.07571840286254883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,4,4,power_law_1.01,0.08606079816818238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,4,4,power_law_1.01,0.10559359788894654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,4,power_law_1.2,0.1500864028930664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,8,power_law_1.2,0.40147838592529295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,4,4,power_law_1.01,0.12080639600753784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,4,4,power_law_1.01,0.15162880420684816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,8,4,power_law_1.01,0.8365056037902832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,4,4,power_law_1.01,0.17207679748535157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,4,4,power_law_1.01,0.3060992002487183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,4,4,power_law_1.01,0.3819583892822266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,4,power_law_1.2,0.1761664032936096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,4,4,power_law_1.01,0.42761597633361814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,8,power_law_1.2,0.39957759380340574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,4,4,power_law_1.01,0.609164810180664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,8,4,power_law_1.01,1.120569610595703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,4,4,power_law_1.01,0.7649919986724854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,4,power_law_1.2,0.1784767985343933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,4,4,power_law_1.01,1.2758655548095703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,8,power_law_1.2,0.40945920944213865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,8,4,power_law_1.01,1.2949440002441406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,4,power_law_1.2,0.21991679668426514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,4,4,power_law_1.01,2.2464704513549805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,8,power_law_1.2,0.416377592086792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,4,power_law_1.2,0.25569279193878175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,8,4,power_law_1.01,1.92410888671875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,4,power_law_1.2,0.3441087961196899
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,8,power_law_1.2,0.42981758117675783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,4,power_law_1.2,0.4348608016967773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,8,power_law_1.2,0.44851198196411135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,8,4,power_law_1.01,3.9328449249267576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,4,power_law_1.2,0.5841599941253662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,8,power_law_1.2,0.4638336181640625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,4,power_law_1.2,0.7488319873809814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,8,power_law_1.2,0.478380823135376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,4,power_law_1.2,1.0878527641296387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,8,power_law_1.2,0.5031167984008789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,4,power_law_1.2,1.423699188232422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,8,power_law_1.2,0.5788991928100586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,4,power_law_1.2,0.07134720087051391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,4,power_law_1.2,1.7818496704101563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,8,power_law_1.2,0.5750592231750489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,4,power_law_1.2,0.11809920072555542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,8,power_law_1.2,0.6868927955627442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,4,power_law_1.2,0.08662400245666504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,4,power_law_1.2,2.7648767471313476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,2,2,power_law_1.2,0.05578240156173706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,2,2,power_law_1.2,0.07199360132217407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,4,power_law_1.2,0.08355839848518372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,2,2,power_law_1.2,0.06563839912414551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,8,power_law_1.2,0.7753024101257324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,2,2,power_law_1.2,0.08492799997329711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,2,2,power_law_1.2,0.08517760038375854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,2,2,power_law_1.2,0.08365439772605895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,4,power_law_1.2,0.11781120300292969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,2,2,power_law_1.2,0.09114239811897278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,2,2,power_law_1.2,0.09365760087966919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,2,2,power_law_1.2,0.0953279972076416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,8,power_law_1.2,0.9490943908691406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,2,2,power_law_1.2,0.09482240080833435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,4,power_law_1.2,0.15437439680099488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,2,2,power_law_1.2,0.09304320216178893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,4,power_law_1.2,5.102092742919922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,2,2,power_law_1.2,0.09562240242958069
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,2,2,power_law_1.2,0.09715200066566468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,2,2,power_law_1.2,0.1088703989982605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,4,power_law_1.2,0.16918400526046753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,2,2,power_law_1.2,0.11839359998703003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,2,2,power_law_1.2,0.13093760013580322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,8,power_law_1.2,0.9625408172607421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,2,2,power_law_1.2,0.13370879888534545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,2,2,power_law_1.2,0.15198080539703368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,4,power_law_1.2,0.22612481117248534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,2,2,power_law_1.2,0.192249596118927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,2,2,power_law_1.2,0.21235840320587157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,2,2,power_law_1.2,0.28314239978790284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,4,power_law_1.2,0.2394239902496338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,8,power_law_1.2,1.1718976020812988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,2,2,power_law_1.2,0.37667839527130126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,2,2,power_law_1.2,0.43848958015441897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,4,power_law_1.2,0.25668480396270754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,2,2,power_law_1.2,0.7112703800201416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,2,2,power_law_1.2,0.8997119903564453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,2,2,power_law_1.2,1.3184576034545898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,4,power_law_1.2,0.26847999095916747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,8,power_law_1.2,1.5395392417907714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,2,2,power_law_1.2,1.5646464347839355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,4,power_law_1.2,0.29198079109191893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,2,2,power_law_1.2,2.2234624862670898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,4,power_law_1.2,0.2864703893661499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,2,2,power_law_1.2,3.248275375366211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,8,power_law_1.2,2.012985610961914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,4,power_law_1.2,0.2975231885910034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,2,2,power_law_1.2,7.13372802734375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,16,balanced,0.04806933303674062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,16,balanced,0.051167999704678856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,16,balanced,0.0537066658337911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,4,power_law_1.2,0.30691840648651125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,16,balanced,0.08004266520341237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,16,balanced,0.10167466600735982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,16,balanced,0.1051573355992635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,16,balanced,0.10834667086601257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,16,balanced,0.10652800401051839
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,16,balanced,0.1088053286075592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,16,balanced,0.10692800084749858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,16,balanced,0.10801600416501363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,16,balanced,0.11215466260910034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,16,balanced,0.11064533392588298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,16,balanced,0.11284266908963521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,16,balanced,0.11626666784286499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,16,balanced,0.11741866668065389
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,16,balanced,0.12367467085520427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,16,balanced,0.13462400436401367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,16,balanced,0.1413386662801107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,16,balanced,0.1747093399365743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,16,balanced,0.18424532810846964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,16,balanced,0.2587520082791646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,4,power_law_1.2,0.3178623914718628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,16,balanced,0.24548800786336264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,16,balanced,0.31282132863998413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,8,power_law_1.2,2.6506559371948244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,16,balanced,0.356170654296875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,16,balanced,0.49945068359375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,16,balanced,0.5737226804097494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,4,16,balanced,0.7148906389872233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,4,16,balanced,1.0252959728240967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,4,power_law_1.2,0.3436863899230957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,4,16,balanced,2.105834643046061
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,4,power_law_1.2,0.3827008008956909
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,4,power_law_1.2,0.40755839347839357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,1,8,power_law_1.2,2.814828872680664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,4,power_law_1.2,0.4558527946472168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,4,power_law_1.2,0.5407487869262695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,4,power_law_1.2,0.6640575885772705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,1,8,power_law_1.2,4.904383850097656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,4,power_law_1.2,0.7762879848480224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,4,power_law_1.2,1.0288576126098632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,2,power_law_1.01,0.0966592013835907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,16,2,power_law_1.01,0.030380800366401672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,16,2,power_law_1.01,0.030502399802207945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,16,2,power_law_1.01,0.030828800797462464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,4,power_law_1.2,1.1406784057617188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,16,2,power_law_1.01,0.031174400448799135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,16,2,power_law_1.01,0.031948798894882204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,16,2,power_law_1.01,0.03281280100345611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,16,2,power_law_1.01,0.03357439935207367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,16,2,power_law_1.01,0.034067198634147644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,16,2,power_law_1.01,0.03589119911193848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,16,2,power_law_1.01,0.036620798707008365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,16,2,power_law_1.01,0.037273600697517395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,16,2,power_law_1.01,0.03865599930286408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,16,2,power_law_1.01,0.03973760008811951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,4,power_law_1.2,1.5910400390625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,16,2,power_law_1.01,0.03993600010871887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,2,power_law_1.01,0.12611839771270753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,16,2,power_law_1.01,0.04533120095729828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,16,2,power_law_1.01,0.04992640018463135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,16,2,power_law_1.01,0.055174398422241214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,16,2,power_law_1.01,0.06207360029220581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,16,2,power_law_1.01,0.07442560195922851
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,16,2,power_law_1.01,0.10239360332489014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,16,2,power_law_1.01,0.12275840044021606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,16,2,power_law_1.01,0.14759039878845215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,16,2,power_law_1.01,0.17696640491485596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,4,power_law_1.2,1.9049856185913085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,16,2,power_law_1.01,0.2704319953918457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,16,2,power_law_1.01,0.30804479122161865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,2,power_law_1.01,0.18468480110168456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,16,2,power_law_1.01,0.44762239456176756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,1,8,power_law_1.2,8.89200668334961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,16,2,power_law_1.01,0.6528768062591552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,16,2,power_law_1.01,0.7730688095092774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,16,2,power_law_1.01,1.2060159683227538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,4,power_law_1.2,2.308902359008789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,2,power_law_1.01,0.13699840307235717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,16,2,power_law_1.01,2.510348892211914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,2,power_law_1.01,0.23099520206451415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,4,power_law_1.2,3.3849472045898437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,2,power_law_1.01,0.33977599143981935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,2,power_law_1.01,0.404307222366333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,4,power_law_1.2,6.8300926208496096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,2,power_law_1.01,0.5045887947082519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,2,power_law_1.01,0.5255616188049317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,2,balanced,0.10134399930636089
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,2,balanced,0.15029333035151163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,2,balanced,0.1696959932645162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,2,power_law_1.01,0.5512639999389648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,2,balanced,0.26125866174697876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,2,balanced,0.4304693142573039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,2,balanced,0.5892106691996256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,2,balanced,0.6043573220570883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,2,balanced,0.607264002164205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,2,balanced,0.607040007909139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,2,balanced,0.6105279922485352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,2,balanced,0.6187413136164347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,2,balanced,0.626202662785848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,2,balanced,0.6293546756108602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,2,balanced,0.6429493427276611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,2,balanced,0.6530719995498657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,2,power_law_1.01,0.5928959846496582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,2,balanced,0.6617066860198975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,2,balanced,0.6837013562520345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,2,balanced,0.7325279712677002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,2,balanced,0.7758293151855469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,2,balanced,1.1059040228525798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,2,balanced,0.9849653244018555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,4,power_law_1.2,0.07263360023498536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,2,balanced,1.3998613357543945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,2,balanced,1.4421067237854004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,2,power_law_1.01,0.6023551940917968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,2,balanced,2.139498710632324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,2,balanced,2.4101759592692056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,4,power_law_1.2,0.0735871970653534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,2,balanced,3.612250645955404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,2,power_law_1.01,0.6326464176177978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,2,balanced,4.312608083089192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,4,power_law_1.2,0.06401280164718628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,1,2,balanced,5.353637059529622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,2,power_law_1.01,0.6265600204467774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,4,power_law_1.2,0.07893120050430298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,1,2,balanced,8.755061467488607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,32,power_law_1.01,0.058259201049804685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,2,power_law_1.01,0.6752319812774659
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,4,power_law_1.2,0.08869119882583618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,32,power_law_1.01,0.0578495979309082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,2,power_law_1.01,0.6653759956359864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,1,2,balanced,16.316853841145832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,4,power_law_1.2,0.12746880054473878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,32,power_law_1.01,0.04913919866085052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,2,power_law_1.01,0.6860479831695556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,4,power_law_1.2,0.1349951982498169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,32,power_law_1.01,0.05045120120048523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,4,power_law_1.2,0.14102400541305543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,2,power_law_1.01,0.7513728141784668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,32,power_law_1.01,0.05392000079154968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,4,power_law_1.2,0.14453760385513306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,2,power_law_1.01,0.7773439884185791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,32,power_law_1.01,0.06039680242538452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,4,power_law_1.2,0.1451200008392334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,32,power_law_1.01,0.05948160290718078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,2,power_law_1.01,0.8713343620300293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,4,power_law_1.2,0.1508031964302063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,32,power_law_1.01,0.061337602138519284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,2,power_law_1.01,0.9651776313781738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,4,power_law_1.2,0.15965440273284912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,32,power_law_1.01,0.058822399377822875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,4,power_law_1.2,0.1637760043144226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,2,power_law_1.01,1.1461695671081542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,32,power_law_1.01,0.06380800008773804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,4,power_law_1.2,0.17095680236816407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,2,power_law_1.01,1.2402496337890625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,32,power_law_1.01,0.06467840075492859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,4,power_law_1.2,0.18096640110015869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,32,power_law_1.01,0.0681984007358551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,2,power_law_1.01,1.5397631645202636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,4,power_law_1.2,0.18679039478302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,32,power_law_1.01,0.07141119837760926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,4,power_law_1.2,0.1982848048210144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,2,power_law_1.01,1.8838720321655273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,32,power_law_1.01,0.07553920149803162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,4,power_law_1.2,0.22515199184417725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,32,power_law_1.01,0.08278399705886841
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,2,power_law_1.01,2.465203285217285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,4,power_law_1.2,0.24791040420532226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,32,power_law_1.01,0.08611840009689331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,32,power_law_1.01,0.088646399974823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,4,power_law_1.2,0.3122944116592407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,2,power_law_1.01,3.371193695068359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,32,power_law_1.01,0.10264960527420045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,4,power_law_1.2,0.36659200191497804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,32,power_law_1.01,0.11179519891738891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,2,power_law_1.01,3.768819046020508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,4,power_law_1.2,0.488972806930542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,32,power_law_1.01,0.13848960399627686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,4,power_law_1.2,0.5457536220550537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,32,power_law_1.01,0.17007360458374024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,2,power_law_1.01,5.701670455932617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,4,power_law_1.2,0.7844351768493653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,32,power_law_1.01,0.2140671968460083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,4,power_law_1.2,0.8906559944152832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,32,power_law_1.01,0.22784640789031982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,32,power_law_1.01,0.2946432113647461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,2,power_law_1.01,10.488921356201171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,4,power_law_1.2,1.411897563934326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,32,power_law_1.01,0.37735040187835694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,4,power_law_1.2,1.6859840393066405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,32,power_law_1.01,0.498521614074707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,8,4,power_law_1.2,2.1566144943237306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,32,power_law_1.01,0.6374976158142089
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,4,32,power_law_1.01,0.7561088085174561
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,16,balanced,0.047637333472569786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,16,balanced,0.04740799963474274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,16,balanced,0.0476693312327067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,16,balanced,0.05375466744105021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,8,4,power_law_1.2,3.2333568572998046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,16,balanced,0.0812799980243047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,16,balanced,0.08595200379689534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,16,balanced,0.08666666348775227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,16,balanced,0.08608532945315044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,16,balanced,0.0860746701558431
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,16,balanced,0.08850133419036865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,16,balanced,0.08705066641171773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,16,balanced,0.08911466598510742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,16,balanced,0.09037866195042928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,16,balanced,0.09276800354321797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,16,balanced,0.0969493289788564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,16,balanced,0.09994666775067647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,16,balanced,0.10217066605885823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,16,balanced,0.11635733644167583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,16,balanced,0.11698133746782939
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,16,balanced,0.13743467132250467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,16,balanced,0.15128533045450845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,16,balanced,0.1941279967625936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,16,balanced,0.21477333704630533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,16,balanced,0.3197759985923767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,16,balanced,0.3524320125579834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,16,balanced,0.510095993677775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,16,balanced,0.6215893427530924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,16,balanced,0.7288479804992676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,16,balanced,1.1326026916503906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,4,32,power_law_1.01,1.2106111526489258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,16,balanced,2.168821334838867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,8,4,power_law_1.2,6.5866752624511715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,4,32,power_law_1.01,2.342815971374512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,4,power_law_1.2,0.04987519979476929
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,4,power_law_1.2,0.04580479860305786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,4,power_law_1.2,0.05066239833831787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,16,1,power_law_1.2,0.03485440015792847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,16,1,power_law_1.2,0.035769599676132205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,4,power_law_1.2,0.05398399829864502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,16,1,power_law_1.2,0.0353408008813858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,16,1,power_law_1.2,0.03813120126724243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,16,1,power_law_1.2,0.03863039910793305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,16,1,power_law_1.2,0.03920640051364899
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,16,1,power_law_1.2,0.03976320028305054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,16,1,power_law_1.2,0.04017919898033142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,4,power_law_1.2,0.05880320072174072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,16,1,power_law_1.2,0.04094719886779785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,16,1,power_law_1.2,0.04056319892406464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,16,1,power_law_1.2,0.04229759871959686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,16,1,power_law_1.2,0.04539520144462585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,16,1,power_law_1.2,0.04688639938831329
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,4,power_law_1.2,0.06659839749336242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,16,1,power_law_1.2,0.04842880070209503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,16,1,power_law_1.2,0.05568000078201294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,16,1,power_law_1.2,0.05866879820823669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,16,1,power_law_1.2,0.06348800063133239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,16,1,power_law_1.2,0.07182719707489013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,16,1,power_law_1.2,0.08055679798126221
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,4,power_law_1.2,0.07147520184516906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,16,1,power_law_1.2,0.10461440086364746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,16,1,power_law_1.2,0.11608959436416626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,16,1,power_law_1.2,0.1492735981941223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,16,1,power_law_1.2,0.18158719539642335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,4,power_law_1.2,0.07343360185623168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,16,1,power_law_1.2,0.25311999320983886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,16,1,power_law_1.2,0.3250816106796265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,16,1,power_law_1.2,0.46392321586608887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,4,power_law_1.2,0.0767359972000122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,16,1,power_law_1.2,0.6061056137084961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,16,1,power_law_1.2,0.7387263774871826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,4,power_law_1.2,0.07883520126342773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,16,1,power_law_1.2,1.139622402191162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,4,power_law_1.2,0.08168960213661194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,16,1,power_law_1.2,2.2302656173706055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,4,power_law_1.2,0.0887935996055603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,4,power_law_1.2,0.09064319729804993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,4,power_law_1.2,0.08878080248832702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,4,power_law_1.2,0.09710080027580262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,4,power_law_1.2,0.10099200010299683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,4,power_law_1.2,0.1108672022819519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,4,power_law_1.2,0.1265023946762085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,4,power_law_1.2,0.14524799585342407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,4,power_law_1.2,0.17955199480056763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,4,power_law_1.2,0.23520638942718505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,4,power_law_1.2,0.2888704061508179
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,4,power_law_1.2,0.35751678943634035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,4,power_law_1.01,0.03526400029659271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,4,power_law_1.2,0.47861762046813966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,4,power_law_1.01,0.03824000060558319
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,4,power_law_1.01,0.03959679901599884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,4,power_law_1.2,0.6648255825042725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,4,power_law_1.01,0.04797439873218536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,4,power_law_1.2,0.9913215637207031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,4,power_law_1.01,0.057062399387359616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,4,power_law_1.01,0.07367039918899536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,4,power_law_1.01,0.0726207971572876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,4,power_law_1.2,1.1882559776306152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,4,power_law_1.01,0.07495679855346679
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,4,power_law_1.01,0.0753600001335144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,4,power_law_1.01,0.07680000066757202
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,4,power_law_1.2,1.5301312446594237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,4,power_law_1.01,0.08062080144882203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,4,power_law_1.01,0.08341119885444641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,4,power_law_1.01,0.08558719754219055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,4,power_law_1.01,0.08550400137901307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,4,power_law_1.01,0.09185280203819275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,4,power_law_1.2,2.7546112060546877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,4,power_law_1.01,0.0974399983882904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,4,power_law_1.01,0.09703680276870727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,4,power_law_1.01,0.11089919805526734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,4,power_law_1.01,0.11875840425491332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,4,power_law_1.01,0.13511680364608764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,4,power_law_1.01,0.15895040035247804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,4,power_law_1.2,5.233113479614258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,4,power_law_1.01,0.2067199945449829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,4,power_law_1.01,0.2549312114715576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,4,power_law_1.01,0.32648959159851076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,4,power_law_1.01,0.38618879318237304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,4,power_law_1.01,0.5755712032318115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,4,power_law_1.01,0.6410240173339844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,4,power_law_1.01,0.9156479835510254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,4,power_law_1.01,1.272979164123535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,4,power_law_1.01,2.5575231552124023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,4,power_law_1.01,0.03560320138931274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,4,power_law_1.01,0.03548159897327423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,4,power_law_1.01,0.03689599931240082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,4,power_law_1.01,0.03943679928779602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,4,power_law_1.01,0.04270719885826111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,4,power_law_1.01,0.05898879766464234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,4,power_law_1.01,0.05854079723358154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,4,power_law_1.01,0.06041600108146668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,4,power_law_1.01,0.06120319962501526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,4,power_law_1.01,0.06162559986114502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,4,power_law_1.01,0.06391680240631104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,4,power_law_1.01,0.06579840183258057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,4,power_law_1.01,0.07017599940299987
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,4,power_law_1.01,0.07442560195922851
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,4,power_law_1.01,0.07928959727287292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,4,power_law_1.01,0.08615679740905761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,4,power_law_1.01,0.08509439826011658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,4,power_law_1.01,0.09852799773216248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,4,power_law_1.01,0.11012480258941651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,4,power_law_1.01,0.12952959537506104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,4,power_law_1.01,0.15342719554901124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,4,power_law_1.01,0.20227200984954835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,4,power_law_1.01,0.26483840942382814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,4,power_law_1.01,0.3323456048965454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,4,power_law_1.01,0.459987211227417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,4,power_law_1.01,0.6293056011199951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,4,power_law_1.01,0.7851327896118164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,4,power_law_1.01,1.0564352035522462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,4,power_law_1.01,1.5384511947631836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,4,power_law_1.01,2.9895231246948244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,16,power_law_1.01,0.0661952018737793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,16,power_law_1.01,0.08097919821739197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,16,power_law_1.01,0.0871616005897522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,16,power_law_1.01,0.06002560257911682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,16,power_law_1.01,0.06028159856796265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,16,power_law_1.01,0.06652160286903382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,16,power_law_1.01,0.06959360241889953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,16,power_law_1.01,0.0766207993030548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,16,power_law_1.01,0.07628160119056701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,16,power_law_1.01,0.07684479951858521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,16,power_law_1.01,0.08775039911270141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,16,power_law_1.01,0.08110079765319825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,16,power_law_1.01,0.08329600095748901
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,16,power_law_1.01,0.08984959721565247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,16,power_law_1.01,0.09501439929008484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,16,power_law_1.01,0.09296640157699584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,16,power_law_1.01,0.10325759649276733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,16,power_law_1.01,0.11781120300292969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,16,power_law_1.01,0.12520320415496827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,16,power_law_1.01,0.15631359815597534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,16,power_law_1.01,0.17781120538711548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,16,power_law_1.01,0.2309567928314209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,16,power_law_1.01,0.2557760000228882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,16,power_law_1.01,0.3447423934936523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,16,power_law_1.01,0.4323904037475586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,16,power_law_1.01,0.6078080177307129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,16,power_law_1.01,0.8071295738220214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,16,power_law_1.01,0.9176383972167969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,16,power_law_1.01,1.4440640449523925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,16,power_law_1.01,2.8883840560913088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,2,8,power_law_1.01,0.0378495991230011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,2,8,power_law_1.01,0.038099199533462524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,2,8,power_law_1.01,0.037248000502586365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,2,8,power_law_1.01,0.03718400001525879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,2,8,power_law_1.01,0.03922559916973114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,2,8,power_law_1.01,0.0459199994802475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,2,8,power_law_1.01,0.04956159889698029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,2,8,power_law_1.01,0.05129600167274475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,2,8,power_law_1.01,0.055257600545883176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,2,8,power_law_1.01,0.056415998935699464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,2,8,power_law_1.01,0.05813120007514953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,2,8,power_law_1.01,0.062067198753356936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,2,8,power_law_1.01,0.06613119840621948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,2,8,power_law_1.01,0.06340479850769043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,2,8,power_law_1.01,0.072953599691391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,2,8,power_law_1.01,0.07987200021743775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,2,8,power_law_1.01,0.08236799836158752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,2,8,power_law_1.01,0.10043519735336304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,2,8,power_law_1.01,0.1134719967842102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,2,8,power_law_1.01,0.1545215964317322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,2,8,power_law_1.01,0.1773375988006592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,2,8,power_law_1.01,0.24783360958099365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,2,8,power_law_1.01,0.2764863967895508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,2,8,power_law_1.01,0.4170559883117676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,2,8,power_law_1.01,0.5481791973114014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,2,8,power_law_1.01,0.8249728202819824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,2,8,power_law_1.01,1.1042112350463866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,2,8,power_law_1.01,1.283993625640869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,2,8,power_law_1.01,2.054355239868164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,2,8,power_law_1.01,3.6718910217285154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,4,power_law_1.01,0.04505600035190582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,4,power_law_1.01,0.043942400813102724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,4,power_law_1.01,0.04366720020771027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,4,power_law_1.01,0.048256000876426695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,4,power_law_1.01,0.05386880040168762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,4,power_law_1.01,0.0681984007358551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,4,power_law_1.01,0.07059199810028076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,4,power_law_1.01,0.0704255998134613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,4,power_law_1.01,0.07272959947586059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,4,power_law_1.01,0.07345280051231384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,4,power_law_1.01,0.07581440210342408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,4,power_law_1.01,0.07749760150909424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,4,power_law_1.01,0.07878400087356567
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,4,power_law_1.01,0.08194559812545776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,4,power_law_1.01,0.08837760090827942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,4,power_law_1.01,0.09208959937095643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,4,power_law_1.01,0.09360640048980713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,4,power_law_1.01,0.10362880229949951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,4,power_law_1.01,0.11918079853057861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,4,power_law_1.01,0.1416767954826355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,4,power_law_1.01,0.16310399770736694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,4,power_law_1.01,0.1961856007575989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,4,power_law_1.01,0.25148160457611085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,4,power_law_1.01,0.3424384117126465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,4,power_law_1.01,0.42993922233581544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,4,power_law_1.01,0.5663487911224365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,4,power_law_1.01,0.8179840087890625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,4,power_law_1.01,0.9231167793273926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,4,power_law_1.01,1.5080896377563477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,4,power_law_1.01,3.0828479766845702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,32,8,balanced,0.03955733279387156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,32,8,balanced,0.03620799879233042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,32,8,balanced,0.03350399931271871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,32,8,balanced,0.03323200096686681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,32,8,balanced,0.033770665526390076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,32,8,balanced,0.035461333890755974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,32,8,balanced,0.0330079992612203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,32,8,balanced,0.03457599878311157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,32,8,balanced,0.03374933451414108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,32,8,balanced,0.034714666505654655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,32,8,balanced,0.03617066641648611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,32,8,balanced,0.03542399903138479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,32,8,balanced,0.03449599941571554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,32,8,balanced,0.036576000352700554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,32,8,balanced,0.038880000511805214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,32,8,balanced,0.04060266663630804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,32,8,balanced,0.040394666294256844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,32,8,balanced,0.041008000572522484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,32,8,balanced,0.04484800000985464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,32,8,balanced,0.046181331078211464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,32,8,balanced,0.051226665576299034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,32,8,balanced,0.05905066430568695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,32,8,balanced,0.06919999917348225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,32,8,balanced,0.09185066819190979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,32,8,balanced,0.10091200470924377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,32,8,balanced,0.13822399576505026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,32,8,balanced,0.16214932998021445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,32,8,balanced,0.20966933170954385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,32,8,balanced,0.27194666862487793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,32,8,balanced,0.4941759904225667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,32,power_law_1.01,0.11943680047988892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,32,power_law_1.01,0.16655999422073364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,32,power_law_1.01,0.15815039873123168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,32,power_law_1.01,0.09119359850883484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,32,power_law_1.01,0.09532160162925721
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,32,power_law_1.01,0.1051967978477478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,32,power_law_1.01,0.12083200216293336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,32,power_law_1.01,0.1238144040107727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,16,balanced,0.05012799799442291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,16,balanced,0.054229333996772766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,16,balanced,0.04993066688378652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,16,balanced,0.05403733253479004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,32,power_law_1.01,0.14337919950485228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,16,balanced,0.06289066871007283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,16,balanced,0.0754613329966863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,16,balanced,0.07642666498819987
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,16,balanced,0.07793599863847096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,16,balanced,0.07931733131408691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,16,balanced,0.07819200058778127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,16,balanced,0.07858133316040039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,16,balanced,0.07950399816036224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,16,balanced,0.08059733112653096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,16,balanced,0.08226133386294048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,16,balanced,0.085999995470047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,16,balanced,0.0867199997107188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,16,balanced,0.08925333619117737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,16,balanced,0.09629866480827332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,16,balanced,0.10249066352844238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,16,balanced,0.12078400452931722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,16,balanced,0.12549333771069845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,16,balanced,0.1552853286266327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,16,balanced,0.1734773317972819
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,16,balanced,0.23172267278035483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,16,balanced,0.27196266253789264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,16,balanced,0.3693600098292033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,16,balanced,0.46110932032267254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,16,balanced,0.5427039861679077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,32,power_law_1.01,0.14603519439697266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,16,balanced,0.8403306802113851
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,16,balanced,1.5948266983032227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,32,power_law_1.01,0.1570623993873596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,32,power_law_1.01,0.14005759954452515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,32,power_law_1.01,0.1518336057662964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,32,power_law_1.01,0.14586880207061767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,32,power_law_1.01,0.15728000402450562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,32,power_law_1.01,0.16279040575027465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,32,power_law_1.01,0.16514559984207153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,128,balanced,0.054192001620928444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,128,balanced,0.05411200225353241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,128,balanced,0.05225066840648651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,128,balanced,0.06853333115577698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,128,balanced,0.07977599898974101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,128,balanced,0.08095466593901317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,128,balanced,0.07982400059700012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,128,balanced,0.08281599978605907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,128,balanced,0.08230400085449219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,128,balanced,0.08157333234945933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,128,balanced,0.08082666496435802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,128,balanced,0.08357333143552144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,128,balanced,0.08474666873613994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,128,balanced,0.08506666620572408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,128,balanced,0.09016000231107076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,128,balanced,0.09092266360918681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,128,balanced,0.09313066800435384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,128,balanced,0.10790399710337321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,128,balanced,0.10583999752998352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,128,balanced,0.13261866569519043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,128,balanced,0.13647466897964478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,32,power_law_1.01,0.17280639410018922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,128,balanced,0.17923200130462646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,128,balanced,0.1657386620839437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,128,balanced,0.21623466412226358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,128,balanced,0.22692267100016275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,128,balanced,0.2999040087064107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,128,balanced,0.34329064687093097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,1,128,balanced,0.4177866776784261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,1,128,balanced,0.5849386850992838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,1,128,balanced,1.247754653294881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,32,power_law_1.01,0.182425594329834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,32,power_law_1.01,0.20553600788116455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,32,power_law_1.01,0.21505920886993407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,32,power_law_1.01,0.27250559329986573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,32,power_law_1.01,0.32291200160980227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,32,power_law_1.01,0.4027967929840088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,32,power_law_1.01,0.49831042289733884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,2,8,power_law_1.01,0.057222402095794676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,2,8,power_law_1.01,0.08202239871025085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,2,8,power_law_1.01,0.057011198997497556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,2,8,power_law_1.01,0.057043200731277464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,32,power_law_1.01,0.6966400146484375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,2,8,power_law_1.01,0.05799679756164551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,2,8,power_law_1.01,0.06076800227165222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,2,8,power_law_1.01,0.06259840130805969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,2,8,power_law_1.01,0.07356799840927124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,2,8,power_law_1.01,0.26121599674224855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,2,8,power_law_1.01,0.38650879859924314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,2,8,power_law_1.01,0.06527360081672669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,2,8,power_law_1.01,0.07264639735221863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,2,8,power_law_1.01,0.07689599990844727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,2,8,power_law_1.01,0.0794368028640747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,2,8,power_law_1.01,0.0997439980506897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,2,8,power_law_1.01,0.10039680004119873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,32,power_law_1.01,0.8716992378234864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,2,8,power_law_1.01,0.10281599760055542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,2,8,power_law_1.01,0.11578240394592285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,2,8,power_law_1.01,0.14826879501342774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,2,8,power_law_1.01,0.21665918827056885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,2,8,power_law_1.01,0.24045441150665284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,2,8,power_law_1.01,0.3389120101928711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,2,8,power_law_1.01,0.3997312068939209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,2,8,power_law_1.01,0.5612415790557861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,2,8,power_law_1.01,0.7629695892333984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,32,power_law_1.01,1.081881618499756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,2,8,power_law_1.01,1.2299776077270508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,2,8,power_law_1.01,1.395315170288086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,2,8,power_law_1.01,1.8358463287353515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,2,8,power_law_1.01,3.037241554260254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,32,power_law_1.01,1.6668672561645508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,2,8,power_law_1.01,5.501612854003906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,32,power_law_1.01,2.975718307495117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,32,power_law_1.2,0.0379967987537384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,32,power_law_1.2,0.034355199337005614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,32,power_law_1.2,0.03116160035133362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,32,power_law_1.2,0.030822399258613586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,32,power_law_1.2,0.03203839957714081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,32,power_law_1.2,0.03319680094718933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,32,power_law_1.2,0.03383040130138397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,32,power_law_1.2,0.033344000577926636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,32,power_law_1.2,0.0335999995470047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,32,power_law_1.2,0.03346560001373291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,32,power_law_1.2,0.034745600819587705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,32,power_law_1.2,0.03601279854774475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,32,power_law_1.2,0.03576320111751556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,32,power_law_1.2,0.03763839900493622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,32,power_law_1.2,0.04166400134563446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,32,power_law_1.2,0.0416128009557724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,32,power_law_1.2,0.04455040097236633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,32,power_law_1.2,0.05107839703559876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,32,power_law_1.2,0.056857597827911374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,32,power_law_1.2,0.07047039866447449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,32,power_law_1.2,0.07637760043144226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,32,power_law_1.2,0.10837119817733765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,32,power_law_1.2,0.12969599962234496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,32,power_law_1.2,0.1525439977645874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,32,power_law_1.2,0.20032639503479005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,32,power_law_1.2,0.2988863945007324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,32,power_law_1.2,0.47603840827941896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,32,power_law_1.2,0.6509952068328857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,32,power_law_1.2,0.8196800231933594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,32,power_law_1.2,1.9756671905517578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,2,8,balanced,0.02700799951950709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,2,8,balanced,0.024959998826185863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,2,8,balanced,0.026863999664783478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,2,8,balanced,0.0288426677385966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,2,8,balanced,0.029813334345817566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,2,8,balanced,0.031301334500312805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,2,8,balanced,0.030645333230495453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,2,8,balanced,0.031157332162062328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,2,8,balanced,0.031104000906149547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,2,8,balanced,0.031023999055226643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,2,8,balanced,0.03181333343187968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,2,8,balanced,0.03253333270549774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,2,8,balanced,0.03326933334271113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,2,8,balanced,0.0388373335202535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,2,8,balanced,0.04101333270470301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,2,8,balanced,0.04162133236726125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,2,8,balanced,0.04474666714668274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,2,8,balanced,0.04743466774622599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,8,balanced,0.05275199810663859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,8,balanced,0.06508799890677135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,8,balanced,0.06057066718737284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,8,balanced,0.07807999849319458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,8,balanced,0.08713066577911377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,8,balanced,0.10626133282979329
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,8,balanced,0.13090667128562927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,8,balanced,0.16922666629155478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,8,balanced,0.20002132654190063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,8,balanced,0.24290132522583008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,8,balanced,0.3531200091044108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,8,balanced,0.6487626632054647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,128,balanced,0.06595733265082042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,128,balanced,0.0655626654624939
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,128,balanced,0.06712000072002411
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,128,balanced,0.06504533191521962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,128,balanced,0.06555200119813283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,128,balanced,0.07149866720040639
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,128,balanced,0.07576533158620198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,128,balanced,0.06948266426722209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,128,balanced,0.07622399926185608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,128,balanced,0.0738560010989507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,128,balanced,0.07585066556930542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,128,balanced,0.07948799928029378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,128,balanced,0.0784693310658137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,128,balanced,0.07674666742483775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,128,balanced,0.08569066723187764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,128,balanced,0.08604799707730611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,128,balanced,0.08796266714731853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,128,balanced,0.09391466776529948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,128,balanced,0.09843732913335164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,128,balanced,0.10758399963378906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,128,balanced,0.11668266852696736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,128,balanced,0.13665599624315897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,128,balanced,0.15398933490117392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,128,balanced,0.19538666804631552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,128,balanced,0.23102933168411255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,128,balanced,0.30766934156417847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,128,balanced,0.3802613417307536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,128,balanced,0.47089068094889325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,128,balanced,0.6927253405253092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,128,balanced,1.3497120539347331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,1,1,power_law_1.2,0.07594239711761475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,1,1,power_law_1.2,0.08847360014915466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,1,1,power_law_1.2,0.11187200546264649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,1,1,power_law_1.2,0.15918079614639283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,1,1,power_law_1.2,0.1598528027534485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,1,1,power_law_1.2,0.16462080478668212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,1,1,power_law_1.2,0.16106879711151123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,1,1,power_law_1.2,0.16536320447921754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,1,1,power_law_1.2,0.16656639575958251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,1,1,power_law_1.2,0.1686336040496826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,1,1,power_law_1.2,0.17461119890213012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,1,1,power_law_1.2,0.17868160009384154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,1,1,power_law_1.2,0.1841920018196106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,1,1,power_law_1.2,0.18776960372924806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,1,1,power_law_1.2,0.2066431999206543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,1,1,power_law_1.2,0.20981121063232422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,1,1,power_law_1.2,0.2266495943069458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,1,1,power_law_1.2,0.26439039707183837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,1,1,power_law_1.2,0.3025216102600098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,1,1,power_law_1.2,0.3873984098434448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,1,1,power_law_1.2,0.47047038078308107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,1,1,power_law_1.2,0.6200895786285401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,1,1,power_law_1.2,0.7716991901397705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,1,1,power_law_1.2,1.0913344383239747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,1,1,power_law_1.2,1.4408767700195313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,1,1,power_law_1.2,2.1617984771728516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,1,1,power_law_1.2,2.677996826171875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,1,1,power_law_1.2,3.314656066894531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,1,1,power_law_1.2,4.992761611938477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,1,1,power_law_1.2,9.809458923339843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,2,2,power_law_1.01,0.06636160016059875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,2,2,power_law_1.01,0.07496320009231568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,2,2,power_law_1.01,0.08060799837112427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,2,2,power_law_1.01,0.09625599980354309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,2,2,power_law_1.01,0.0944703996181488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,2,2,power_law_1.01,0.09699199795722961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,2,2,power_law_1.01,0.09710720181465149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,2,2,power_law_1.01,0.09815040230751038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,2,2,power_law_1.01,0.09497600197792053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,2,2,power_law_1.01,0.09879680275917054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,2,2,power_law_1.01,0.09905279874801635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,2,2,power_law_1.01,0.10280959606170655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,2,2,power_law_1.01,0.10496000051498414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,2,2,power_law_1.01,0.11671680212020874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,2,2,power_law_1.01,0.12873599529266358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,2,2,power_law_1.01,0.12752000093460084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,2,2,power_law_1.01,0.12782080173492433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,2,2,power_law_1.01,0.16933759450912475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,2,2,power_law_1.01,0.18240640163421631
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,2,2,power_law_1.01,0.2505919933319092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,2,2,power_law_1.01,0.3096447944641113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,2,2,power_law_1.01,0.3471296072006226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,2,2,power_law_1.01,0.435148811340332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,2,2,power_law_1.01,0.6554880142211914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,2,2,power_law_1.01,0.8483839988708496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,2,2,power_law_1.01,1.1339072227478026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,2,2,power_law_1.01,1.635532760620117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,2,2,power_law_1.01,2.2016767501831054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,2,2,power_law_1.01,2.724608039855957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,2,2,power_law_1.01,6.274604797363281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,16,4,power_law_1.2,0.03975679874420166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,16,4,power_law_1.2,0.0389055997133255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,16,4,power_law_1.2,0.03759360015392303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,16,4,power_law_1.2,0.03712640106678009
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,16,4,power_law_1.2,0.03747200071811676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,16,4,power_law_1.2,0.0383679986000061
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,16,4,power_law_1.2,0.037939199805259706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,16,4,power_law_1.2,0.038950398564338684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,16,4,power_law_1.2,0.03835520148277283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,16,4,power_law_1.2,0.038796800374984744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,16,4,power_law_1.2,0.0403328001499176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,16,4,power_law_1.2,0.041356799006462094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,16,4,power_law_1.2,0.04264959990978241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,16,4,power_law_1.2,0.04211199879646301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,16,4,power_law_1.2,0.0486847996711731
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,16,4,power_law_1.2,0.05144960284233093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,16,4,power_law_1.2,0.054579198360443115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,16,4,power_law_1.2,0.06587520241737366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,16,4,power_law_1.2,0.06872320175170898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,16,4,power_law_1.2,0.09088000059127807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,16,4,power_law_1.2,0.11685760021209717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,16,4,power_law_1.2,0.14689279794692994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,16,4,power_law_1.2,0.19552639722824097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,16,4,power_law_1.2,0.2274240016937256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,16,4,power_law_1.2,0.3466111898422241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,16,4,power_law_1.2,0.4652224063873291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,16,4,power_law_1.2,0.607206392288208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,16,4,power_law_1.2,0.830412769317627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,16,4,power_law_1.2,1.2455039978027345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,16,4,power_law_1.2,2.5032960891723635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,4,power_law_1.2,0.05408639907836914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,4,power_law_1.2,0.04783360064029694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,4,power_law_1.2,0.04743039906024933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,4,power_law_1.2,0.060140800476074216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,4,power_law_1.2,0.06949759721755981
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,4,power_law_1.2,0.10395519733428955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,8,1,power_law_1.01,0.06763520240783691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,4,power_law_1.2,0.1087231993675232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,4,power_law_1.2,0.11181440353393554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,8,1,power_law_1.01,0.07246720194816589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,4,power_law_1.2,0.11729279756546021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,4,power_law_1.2,0.11401599645614624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,4,power_law_1.2,0.11802879571914673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,8,1,power_law_1.01,0.08647680282592773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,4,power_law_1.2,0.11825920343399048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,4,power_law_1.2,0.12220159769058228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,8,1,power_law_1.01,0.10380159616470337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,4,power_law_1.2,0.12447999715805054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,4,power_law_1.2,0.1307263970375061
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,8,1,power_law_1.01,0.12988799810409546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,4,power_law_1.2,0.13186559677124024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,4,power_law_1.2,0.14428160190582276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,8,1,power_law_1.01,0.1758015990257263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,4,power_law_1.2,0.15544960498809815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,4,power_law_1.2,0.17096960544586182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,8,1,power_law_1.01,0.18517760038375855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,4,power_law_1.2,0.21694719791412354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,8,1,power_law_1.01,0.19522559642791748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,4,power_law_1.2,0.2275712013244629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,4,power_law_1.2,0.2585472106933594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,8,1,power_law_1.01,0.1987328052520752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,4,power_law_1.2,0.32286078929901124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,8,1,power_law_1.01,0.2005311965942383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,4,power_law_1.2,0.41821441650390623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,8,1,power_law_1.01,0.20619521141052247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,4,power_law_1.2,0.5128640174865723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,8,1,power_law_1.01,0.2124351978302002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,4,power_law_1.2,0.8184255599975586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,8,1,power_law_1.01,0.21616640090942382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,4,power_law_1.2,0.9666175842285156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,8,1,power_law_1.01,0.22832000255584717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,4,power_law_1.2,1.3232512474060059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,8,1,power_law_1.01,0.2372607946395874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,8,1,power_law_1.01,0.24763519763946534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,4,power_law_1.2,2.017900848388672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,8,1,power_law_1.01,0.26759040355682373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,8,1,power_law_1.01,0.31604480743408203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,8,1,power_law_1.01,0.35335679054260255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,4,power_law_1.2,3.4850879669189454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,8,1,power_law_1.01,0.44161281585693357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,8,1,power_law_1.01,0.5156415939331055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,8,1,power_law_1.01,0.6868415832519531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,8,1,power_law_1.01,0.8459199905395508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,8,1,power_law_1.01,1.1675264358520507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,8,1,power_law_1.01,1.5644543647766114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,8,1,power_law_1.01,2.185785675048828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,8,1,power_law_1.01,2.7760128021240233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,8,1,power_law_1.01,3.4137409210205076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,8,1,power_law_1.01,5.341318511962891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,8,1,power_law_1.01,10.469491577148437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,8,power_law_1.01,0.06861439943313599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,8,power_law_1.01,0.08736000061035157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,8,power_law_1.01,0.1185920000076294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,8,power_law_1.01,0.06934400200843811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,8,power_law_1.01,0.08531200289726257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,8,power_law_1.01,0.10213119983673095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,8,power_law_1.01,0.1068992018699646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,8,power_law_1.01,0.11921919584274292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,8,power_law_1.01,0.1283136010169983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,8,power_law_1.01,0.12775039672851562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,8,power_law_1.01,0.13203840255737304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,8,power_law_1.01,0.1341312050819397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,8,power_law_1.01,0.1385472059249878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,8,power_law_1.01,0.1400256037712097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,8,power_law_1.01,0.15155199766159058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,8,power_law_1.01,0.15194239616394042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,8,power_law_1.01,0.16696319580078126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,8,power_law_1.01,0.18530559539794922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,8,power_law_1.01,0.19890559911727906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,8,power_law_1.01,0.2373823881149292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,4,power_law_1.2,0.05451520085334778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,4,power_law_1.2,0.06896640062332153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,8,power_law_1.01,0.25706241130828855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,4,power_law_1.2,0.053286397457122804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,4,power_law_1.2,0.06818559765815735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,4,power_law_1.2,0.08138239979743958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,8,power_law_1.01,0.33263359069824217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,4,power_law_1.2,0.09310719966888428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,4,power_law_1.2,0.11954560279846191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,8,power_law_1.01,0.3775360107421875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,4,power_law_1.2,0.12554240226745605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,4,power_law_1.2,0.1314303994178772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,8,power_law_1.01,0.49534077644348146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,4,power_law_1.2,0.1298624038696289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,4,power_law_1.2,0.14291199445724487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,4,power_law_1.2,0.14748799800872803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,8,power_law_1.01,0.6114304065704346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,4,power_law_1.2,0.14991999864578248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,4,power_law_1.2,0.1651520013809204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,8,power_law_1.01,0.8450240135192871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,4,power_law_1.2,0.17196799516677858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,4,power_law_1.2,0.18071680068969725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,4,power_law_1.2,0.1768704056739807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,8,power_law_1.01,1.1006719589233398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,4,power_law_1.2,0.20421121120452881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,4,power_law_1.2,0.22250878810882568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,4,power_law_1.2,0.2654400110244751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,8,power_law_1.01,1.3658880233764648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,4,power_law_1.2,0.27952001094818113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,4,power_law_1.2,0.3486783981323242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,4,power_law_1.2,0.4398655891418457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,8,power_law_1.01,2.010380744934082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,4,power_law_1.2,0.5516992092132569
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,4,power_law_1.2,0.6577087879180908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,4,power_law_1.2,0.9892160415649414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,8,power_law_1.01,3.8483646392822264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,4,power_law_1.2,1.3181952476501464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,4,power_law_1.2,1.4302335739135743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,4,power_law_1.2,2.36362247467041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,64,balanced,0.032816000282764435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,64,balanced,0.03105599929889043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,64,balanced,0.02938666691382726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,64,balanced,0.030858665704727173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,64,balanced,0.031157332162062328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,64,balanced,0.03173866619666418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,64,balanced,0.033439998825391136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,64,balanced,0.03139200061559677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,64,balanced,0.03339199970165888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,64,balanced,0.031343999008337654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,64,balanced,0.0335359995563825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,64,balanced,0.0335413341720899
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,64,balanced,0.03378133227427801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,64,balanced,0.035546667873859406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,64,balanced,0.03745600084463755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,64,balanced,0.03955200066169103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,64,balanced,0.03813866774241129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,64,balanced,0.03952533255020777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,64,balanced,0.03956266740957896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,64,balanced,0.04321600000063578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,64,balanced,0.04387733340263367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,64,balanced,0.048826664686203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,64,balanced,0.05209066470464071
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,64,balanced,0.07167999943097432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,64,balanced,0.09071999788284302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,64,balanced,0.12244266271591187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,64,balanced,0.14853333433469137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,64,balanced,0.1785386602083842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,64,balanced,0.25625065962473553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,64,balanced,0.4553333520889282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,4,power_law_1.2,4.600556945800781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,1,power_law_1.01,0.09833599925041199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,1,power_law_1.01,0.14710400104522706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,1,power_law_1.01,0.17620480060577393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,1,power_law_1.01,0.3505728006362915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,32,balanced,0.056074668963750206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,32,balanced,0.05781333148479462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,32,balanced,0.06004266440868378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,32,balanced,0.08065066734949748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,32,balanced,0.09315199653307597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,32,balanced,0.09473599990208943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,32,balanced,0.09514133135477702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,32,balanced,0.09500799576441447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,32,balanced,0.09550399581591289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,32,balanced,0.09591466188430786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,32,balanced,0.09551466504732768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,32,balanced,0.09919466574986775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,32,balanced,0.09804266691207886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,32,balanced,0.09897067149480183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,32,balanced,0.10376532872517903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,32,balanced,0.10557867089907329
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,32,balanced,0.11124799648920695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,32,balanced,0.121888001759847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,32,balanced,0.12959466377894083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,32,balanced,0.16014933586120605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,32,balanced,0.1716853380203247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,32,balanced,0.2246613303820292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,1,power_law_1.01,0.5496960163116456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,32,balanced,0.22366933027903238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,32,balanced,0.2927786707878113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,32,balanced,0.3300693432490031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,32,balanced,0.45524799823760986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,32,balanced,0.5395679871241251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,1,32,balanced,0.6640906731287638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,1,32,balanced,0.9700640042622884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,1,32,balanced,2.023983955383301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,1,power_law_1.01,0.7691711902618408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,1,power_law_1.01,1.092684841156006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,1,power_law_1.01,1.1826239585876466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,1,power_law_1.01,1.2395903587341308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,1,power_law_1.01,1.2248576164245606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,1,power_law_1.01,1.2932671546936034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,1,power_law_1.01,1.367039966583252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,1,power_law_1.01,1.3866751670837403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,4,power_law_1.01,0.045900800824165346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,4,power_law_1.01,0.048281601071357726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,1,power_law_1.01,1.422879981994629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,4,power_law_1.01,0.045798400044441225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,4,power_law_1.01,0.053286397457122804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,4,power_law_1.01,0.05682560205459595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,1,power_law_1.01,1.4816960334777831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,4,power_law_1.01,0.06470400094985962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,4,power_law_1.01,0.06913279891014099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,1,power_law_1.01,1.4896896362304688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,4,power_law_1.01,0.0712768018245697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,4,power_law_1.01,0.07289599776268005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,1,power_law_1.01,1.5673727989196777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,4,power_law_1.01,0.07649919986724854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,4,power_law_1.01,0.07763199806213379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,4,power_law_1.01,0.08352000117301941
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,1,power_law_1.01,1.7005376815795898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,4,power_law_1.01,0.08828160166740417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,4,power_law_1.01,0.08804479837417603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,1,power_law_1.01,1.7040767669677734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,4,power_law_1.01,0.09585279822349549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,4,power_law_1.01,0.099891197681427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,4,power_law_1.01,0.10622719526290894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,1,power_law_1.01,1.9226751327514648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,4,power_law_1.01,0.12211840152740479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,4,power_law_1.01,0.14354560375213624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,1,power_law_1.01,1.9716032028198243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,4,power_law_1.01,0.17375999689102173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,4,power_law_1.01,0.2133120059967041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,1,power_law_1.01,2.3366527557373047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,4,power_law_1.01,0.27548799514770506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,4,power_law_1.01,0.3667327880859375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,1,power_law_1.01,2.5436607360839845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,4,power_law_1.01,0.469817590713501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,4,power_law_1.01,0.6231103897094726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,1,power_law_1.01,3.154636764526367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,4,power_law_1.01,0.8879039764404297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,4,power_law_1.01,1.1210495948791503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,1,power_law_1.01,3.581734466552734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,4,power_law_1.01,1.440556812286377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,1,power_law_1.01,4.668159866333008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,4,power_law_1.01,2.24019832611084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,1,power_law_1.01,5.737036895751953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,4,power_law_1.01,4.417721557617187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,1,power_law_1.01,6.8641212463378904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,1,power_law_1.01,10.195359802246093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,1,power_law_1.01,19.012921142578126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,128,power_law_1.01,0.04012160003185272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,128,power_law_1.01,0.03941119909286499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,128,power_law_1.01,0.040320000052452086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,128,power_law_1.01,0.03896960020065308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,128,power_law_1.01,0.04046080112457275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,128,power_law_1.01,0.04164479970932007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,128,power_law_1.01,0.04184960126876831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,128,power_law_1.01,0.040729600191116336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,128,power_law_1.01,0.04079360067844391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,128,power_law_1.01,0.0412416011095047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,128,power_law_1.01,0.041254401206970215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,128,power_law_1.01,0.043110400438308716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,128,power_law_1.01,0.04252159893512726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,128,power_law_1.01,0.04337919950485229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,128,power_law_1.01,0.04583039879798889
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,128,power_law_1.01,0.04570879936218262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,128,power_law_1.01,0.04838399887084961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,128,power_law_1.01,0.05063679814338684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,128,power_law_1.01,0.05542399883270264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,128,power_law_1.01,0.061638402938842776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,128,power_law_1.01,0.06947839856147767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,128,power_law_1.01,0.08158720135688782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,128,power_law_1.01,0.1020799994468689
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,128,power_law_1.01,0.13222399950027466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,128,power_law_1.01,0.1582527995109558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,128,power_law_1.01,0.24867839813232423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,128,power_law_1.01,0.28589439392089844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,128,power_law_1.01,0.362227201461792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,128,power_law_1.01,0.5723648071289062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,128,power_law_1.01,1.04269437789917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,8,power_law_1.2,0.04945279955863953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,8,power_law_1.2,0.04526079893112182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,8,power_law_1.2,0.04762240052223206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,8,power_law_1.2,0.05221760272979736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,8,power_law_1.2,0.05351679921150208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,8,power_law_1.2,0.05873280167579651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,8,power_law_1.2,0.06459519863128663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,8,power_law_1.2,0.06677759885787964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,8,power_law_1.2,0.06809599995613098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,8,power_law_1.2,0.06907520294189454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,8,power_law_1.2,0.07141759991645813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,8,power_law_1.2,0.07643520236015319
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,8,power_law_1.2,0.08020480275154114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,8,power_law_1.2,0.08101119995117187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,8,power_law_1.2,0.08727040290832519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,8,power_law_1.2,0.09157760143280029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,8,power_law_1.2,0.09733759760856628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,8,power_law_1.2,0.10933760404586793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,8,power_law_1.2,0.12351360321044921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,8,power_law_1.2,0.15371520519256593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,8,power_law_1.2,0.18172800540924072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,8,power_law_1.2,0.24273920059204102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,8,power_law_1.2,0.3161151885986328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,8,power_law_1.2,0.4491903781890869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,8,power_law_1.2,0.5159999847412109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,8,power_law_1.2,0.769375991821289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,8,power_law_1.2,0.9667519569396973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,8,power_law_1.2,1.284780788421631
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,8,power_law_1.2,1.9124351501464845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,8,power_law_1.2,4.776582336425781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,16,4,power_law_1.2,0.03747200071811676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,16,4,power_law_1.2,0.03191039860248566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,16,4,power_law_1.2,0.031455999612808226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,16,4,power_law_1.2,0.03413119912147522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,16,4,power_law_1.2,0.0336896002292633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,16,4,power_law_1.2,0.03432320058345795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,16,4,power_law_1.2,0.03413119912147522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,16,4,power_law_1.2,0.03415040075778961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,16,4,power_law_1.2,0.03573760092258453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,16,4,power_law_1.2,0.03749119937419891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,16,4,power_law_1.2,0.036287999153137206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,16,4,power_law_1.2,0.03678719997406006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,16,4,power_law_1.2,0.040575999021530154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,16,4,power_law_1.2,0.03954559862613678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,16,4,power_law_1.2,0.04815999865531921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,16,4,power_law_1.2,0.0499455988407135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,16,4,power_law_1.2,0.048607999086380006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,16,4,power_law_1.2,0.058316802978515624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,16,4,power_law_1.2,0.07061120271682739
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,16,4,power_law_1.2,0.0889855980873108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,16,4,power_law_1.2,0.09450240135192871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,16,4,power_law_1.2,0.12532479763031007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,16,4,power_law_1.2,0.15324800014495848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,16,4,power_law_1.2,0.20884480476379394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,16,4,power_law_1.2,0.2294912099838257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,16,4,power_law_1.2,0.38709120750427245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,16,4,power_law_1.2,0.5501247882843018
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,16,4,power_law_1.2,0.5368959903717041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,16,4,power_law_1.2,0.8626432418823242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,16,4,power_law_1.2,1.6428415298461914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,128,balanced,0.055434669057528176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,128,balanced,0.05710933109124502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,128,balanced,0.05486933390299479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,128,balanced,0.05716800192991892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,128,balanced,0.057402665416399636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,128,balanced,0.06277333199977875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,128,balanced,0.06539733211199443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,128,balanced,0.06701866785685222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,128,balanced,0.07028266787528992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,128,balanced,0.06484800080458324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,128,balanced,0.06919999917348225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,128,balanced,0.06683200101057689
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,128,balanced,0.07061333457628886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,128,balanced,0.07627200086911519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,128,balanced,0.07620800038178761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,128,balanced,0.07940799991289775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,128,balanced,0.08619200189908345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,128,balanced,0.09472533067067464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,128,balanced,0.10736533006032307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,128,balanced,0.1325759987036387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,128,balanced,0.15068800250689188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,128,balanced,0.1909546653429667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,128,balanced,0.2286346753438314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,128,balanced,0.3067573308944702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,128,balanced,0.3863626718521118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,128,balanced,0.5398133198420206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,128,balanced,0.6935786406199137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,128,balanced,0.8560853004455566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,128,balanced,1.3203893502553303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,128,balanced,2.5851200421651206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,8,balanced,0.0535093347231547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,8,balanced,0.05331199864546458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,8,balanced,0.05363733569780985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,8,balanced,0.05731200178464254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,8,balanced,0.07221333185831706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,8,balanced,0.08060266574223836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,8,balanced,0.08182933429876964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,8,balanced,0.08341333270072937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,8,balanced,0.08380800485610962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,8,balanced,0.08405333757400513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,8,balanced,0.08428266644477844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,8,balanced,0.08433600266774495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,8,balanced,0.0860693355401357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,8,balanced,0.08846933643023173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,8,balanced,0.09160533547401428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,8,balanced,0.09489066402117412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,8,balanced,0.09884799520174663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,8,balanced,0.10645332932472229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,8,balanced,0.11598933736483256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,8,balanced,0.14552533626556396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,8,balanced,0.14595199624697366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,8,balanced,0.19154133399327597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,8,balanced,0.21734933058420816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,8,balanced,0.3025493423144023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,8,balanced,0.349018653233846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,8,balanced,0.48045865694681805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,8,balanced,0.6250933408737183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,8,balanced,0.7430826822916666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,8,balanced,1.1441493034362793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,8,balanced,2.2018826802571616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,64,balanced,0.04669333497683207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,64,balanced,0.036176001032193504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,64,balanced,0.030821333328882854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,64,balanced,0.033071999748547874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,64,balanced,0.033973333736260734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,64,balanced,0.03154666721820831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,64,balanced,0.03347733368476232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,64,balanced,0.03283733377854029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,64,balanced,0.03289599965016047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,64,balanced,0.03306133300065994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,64,balanced,0.031925333042939506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,64,balanced,0.032485333581765495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,64,balanced,0.032442666590213776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,64,balanced,0.03325333446264267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,64,balanced,0.0366239994764328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,64,balanced,0.03473600000143051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,64,balanced,0.03694933404525121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,64,balanced,0.03856533269087473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,64,balanced,0.03758399933576584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,64,balanced,0.04048533240954081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,64,balanced,0.03909866760174433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,64,balanced,0.04444799820582072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,64,balanced,0.040789333482583366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,64,balanced,0.049072002371152244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,64,balanced,0.05295466880003611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,64,balanced,0.06606400012969971
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,64,balanced,0.07355733215808868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,4,64,balanced,0.08197333415349324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,4,64,balanced,0.1116480032602946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,4,64,balanced,0.17413334051767984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,32,2,power_law_1.2,0.036396801471710205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,32,2,power_law_1.2,0.03299840092658997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,32,2,power_law_1.2,0.035097599029541016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,32,2,power_law_1.2,0.03608959913253784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,32,2,power_law_1.2,0.03938559889793396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,32,2,power_law_1.2,0.04844799935817719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,32,2,power_law_1.2,0.05058559775352478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,32,2,power_law_1.2,0.052908802032470705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,32,2,power_law_1.2,0.05381119847297668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,32,2,power_law_1.2,0.05260159969329834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,32,2,power_law_1.2,0.056678402423858645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,32,2,power_law_1.2,0.060096001625061034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,32,2,power_law_1.2,0.06098560094833374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,32,2,power_law_1.2,0.06568319797515869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,32,2,power_law_1.2,0.07334399819374085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,32,2,power_law_1.2,0.07807999849319458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,32,2,power_law_1.2,0.08488320112228394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,32,2,power_law_1.2,0.09505919814109802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,32,2,power_law_1.2,0.11207679510116578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,32,2,power_law_1.2,0.13639039993286134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,32,2,power_law_1.2,0.17843199968338014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,32,2,power_law_1.2,0.2388672113418579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,32,2,power_law_1.2,0.2957823991775513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,32,2,power_law_1.2,0.37115519046783446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,32,2,power_law_1.2,0.5148287773132324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,32,2,power_law_1.2,0.7343743801116943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,32,2,power_law_1.2,0.8795007705688477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,32,2,power_law_1.2,1.1211520195007325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,32,2,power_law_1.2,1.9682367324829102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,32,2,power_law_1.2,3.680972671508789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,1,balanced,0.07354666789372762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,1,balanced,0.10171199838320415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,1,balanced,0.15716266632080078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,1,balanced,0.2593546708424886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,1,balanced,0.4678933223088582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,1,balanced,0.5755199988683065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,1,balanced,0.5783893267313639
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,1,balanced,0.5816533168156942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,1,balanced,0.5840640068054199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,1,balanced,0.5875413417816162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,1,balanced,0.5909706751505533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,1,balanced,0.5989386638005575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,1,balanced,0.6023573478062948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,1,balanced,0.606661319732666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,1,balanced,0.6188426812489828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,1,balanced,0.6284106572469076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,1,balanced,0.6426399946212769
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,1,balanced,0.7159573237101237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,1,balanced,0.7069013118743896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,1,balanced,0.858400026957194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,1,balanced,0.8363893032073975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,1,balanced,1.0541813373565674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,1,balanced,1.1204853057861328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,1,balanced,1.5834186871846516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,1,balanced,1.815615971883138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,1,balanced,2.6735413869222007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,1,balanced,3.3283840815226235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,1,balanced,3.71448548634847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,1,balanced,5.920010884602864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,1,balanced,11.264970143636068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,2,power_law_1.01,0.03402239978313446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,2,power_law_1.01,0.034118399024009705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,2,power_law_1.01,0.035724800825119016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,2,power_law_1.01,0.03854719996452331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,2,power_law_1.01,0.041843199729919435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,2,power_law_1.01,0.052127999067306516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,2,power_law_1.01,0.05511040091514587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,2,power_law_1.01,0.05482239723205566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,2,power_law_1.01,0.0574783980846405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,2,power_law_1.01,0.05856639742851257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,2,power_law_1.01,0.0630079984664917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,2,power_law_1.01,0.06700159907341004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,2,power_law_1.01,0.07294719815254211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,2,power_law_1.01,0.06791679859161377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,2,power_law_1.01,0.07567999958992004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,2,power_law_1.01,0.08086400032043457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,2,power_law_1.01,0.09017599821090698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,2,power_law_1.01,0.10641920566558838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,2,power_law_1.01,0.1295680046081543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,2,power_law_1.01,0.15237120389938355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,2,power_law_1.01,0.1841472029685974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,2,power_law_1.01,0.24552319049835206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,2,power_law_1.01,0.3153343915939331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,2,power_law_1.01,0.45432319641113283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,2,power_law_1.01,0.5690303802490234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,2,power_law_1.01,0.8080960273742676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,2,power_law_1.01,1.0568575859069824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,2,power_law_1.01,1.3711872100830078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,2,power_law_1.01,2.002982330322266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,2,power_law_1.01,4.22303352355957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,4,power_law_1.01,0.05482239723205566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,4,power_law_1.01,0.05459840297698974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,4,power_law_1.01,0.05233280062675476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,4,power_law_1.01,0.06947839856147767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,4,power_law_1.01,0.07336320281028748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,4,power_law_1.01,0.10750080347061157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,4,power_law_1.01,0.11127680540084839
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,4,power_law_1.01,0.10935039520263672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,4,power_law_1.01,0.11230080127716065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,4,power_law_1.01,0.11684479713439941
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,4,power_law_1.01,0.11855360269546508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,4,power_law_1.01,0.11838079690933227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,4,power_law_1.01,0.11839359998703003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,4,power_law_1.01,0.12586239576339722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,4,power_law_1.01,0.13135360479354857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,4,power_law_1.01,0.13463040590286254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,4,power_law_1.01,0.13448959589004517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,4,power_law_1.01,0.15537919998168945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,4,power_law_1.01,0.1725119948387146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,4,power_law_1.01,0.19007999897003175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,4,power_law_1.01,0.21413760185241698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,4,power_law_1.01,0.25505919456481935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,4,power_law_1.01,0.30342400074005127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,4,power_law_1.01,0.39336960315704345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,4,power_law_1.01,0.4922368049621582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,4,power_law_1.01,0.7079808235168457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,4,power_law_1.01,0.8551039695739746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,4,power_law_1.01,0.9979328155517578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,4,power_law_1.01,1.6910591125488281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,4,power_law_1.01,3.173427200317383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,2,128,power_law_1.2,0.03255679905414581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,2,128,power_law_1.2,0.03253760039806366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,2,128,power_law_1.2,0.03511039912700653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,2,128,power_law_1.2,0.03343999981880188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,2,128,power_law_1.2,0.033855998516082765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,2,128,power_law_1.2,0.034764799475669864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,2,128,power_law_1.2,0.03470079898834229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,2,128,power_law_1.2,0.034643200039863584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,2,128,power_law_1.2,0.03438720107078552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,2,128,power_law_1.2,0.03455359935760498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,2,128,power_law_1.2,0.03514240086078644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,2,128,power_law_1.2,0.035641598701477054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,2,128,power_law_1.2,0.035129600763320924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,2,128,power_law_1.2,0.036601600050926206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,2,128,power_law_1.2,0.038335999846458434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,2,128,power_law_1.2,0.038848000764846805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,2,128,power_law_1.2,0.03992320001125336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,2,128,power_law_1.2,0.040889599919319154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,128,power_law_1.2,0.0436928004026413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,128,power_law_1.2,0.04627200067043304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,128,power_law_1.2,0.05230720043182373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,128,power_law_1.2,0.06040319800376892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,128,power_law_1.2,0.07559679746627808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,128,power_law_1.2,0.09623039960861206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,128,power_law_1.2,0.11902719736099243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,128,power_law_1.2,0.15438719987869262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,128,power_law_1.2,0.2119488000869751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,128,power_law_1.2,0.25724799633026124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,32,4,balanced,0.04462933540344238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,128,power_law_1.2,0.3844095945358276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,32,4,balanced,0.04611733555793762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,32,4,balanced,0.048154667019844055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,32,4,balanced,0.05162133276462555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,32,4,balanced,0.060405333836873375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,32,4,balanced,0.08937600255012512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,32,4,balanced,0.09082667032877605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,32,4,balanced,0.09241599837938945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,32,4,balanced,0.09370666742324829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,32,4,balanced,0.09412266810735066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,32,4,balanced,0.09330667058626811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,32,4,balanced,0.09702400366465251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,32,4,balanced,0.09673600395520528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,32,4,balanced,0.0983893374602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,32,4,balanced,0.10637332995732625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,32,4,balanced,0.10930132865905762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,32,4,balanced,0.11162666479746501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,32,4,balanced,0.12542933225631714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,32,4,balanced,0.13537066181500754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,32,4,balanced,0.15529599785804749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,32,4,balanced,0.17383466164271036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,32,4,balanced,0.23287999629974365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,32,4,balanced,0.26898666222890216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,32,4,balanced,0.38441598415374756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,32,4,balanced,0.46883734067281085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,32,4,balanced,0.6780532995859782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,32,4,balanced,0.8923306465148926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,32,4,balanced,1.0965653260548909
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,32,4,balanced,1.6321652730305989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,128,power_law_1.2,0.72707200050354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,32,4,balanced,3.2155946095784507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,1,balanced,0.06883200009663899
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,1,balanced,0.073594664533933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,1,balanced,0.08888000249862671
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,1,balanced,0.12291733423868816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,1,balanced,0.1810879906018575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,1,balanced,0.2772853374481201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,1,balanced,0.2821226716041565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,1,balanced,0.28408533334732056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,1,balanced,0.2865599989891052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,1,balanced,0.28562132517496747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,1,balanced,0.2916319966316223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,1,balanced,0.2983466585477193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,1,balanced,0.30154667297999066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,1,balanced,0.307258665561676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,1,balanced,0.317248006661733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,1,balanced,0.3265226682027181
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,1,balanced,0.3376213312149048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,1,balanced,0.37801067034403485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,1,balanced,0.40826133886973065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,1,balanced,0.4822506507237752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,1,balanced,0.5436586538950602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,1,balanced,0.6816480159759521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,1,balanced,0.8165120283762614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,1,balanced,1.207546631495158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,1,balanced,1.4874614079793294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,1,balanced,2.1505759557088218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,1,balanced,2.7962719599405923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,1,balanced,3.4522453943888345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,1,balanced,5.410373051961263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,1,balanced,10.627994537353516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,32,power_law_1.2,0.09733759760856628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,32,power_law_1.2,0.10335359573364258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,32,power_law_1.2,0.05443840026855469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,32,power_law_1.2,0.06321280002593994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,32,power_law_1.2,0.07705600261688232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,32,power_law_1.2,0.08280320167541504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,32,power_law_1.2,0.09173120260238647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,32,power_law_1.2,0.08972799777984619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,32,power_law_1.2,0.09140480160713196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,32,power_law_1.2,0.09148160219192505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,32,power_law_1.2,0.0938431978225708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,32,power_law_1.2,0.09451519846916198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,32,power_law_1.2,0.09717760086059571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,32,power_law_1.2,0.101254403591156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,8,power_law_1.2,0.07274240255355835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,32,power_law_1.2,0.10689280033111573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,8,power_law_1.2,0.058899199962615965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,8,power_law_1.2,0.05187199711799621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,32,power_law_1.2,0.11026560068130493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,8,power_law_1.2,0.06104320287704468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,32,power_law_1.2,0.11607040166854858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,8,power_law_1.2,0.06897280216217042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,8,power_law_1.2,0.10968960523605346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,32,power_law_1.2,0.13223040103912354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,8,power_law_1.2,0.11109119653701782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,8,power_law_1.2,0.1162559986114502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,32,power_law_1.2,0.13999999761581422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,8,power_law_1.2,0.11333119869232178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,8,power_law_1.2,0.11984000205993653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,32,power_law_1.2,0.17893760204315184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,8,power_law_1.2,0.11901439428329467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,32,power_law_1.2,0.17904000282287597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,8,power_law_1.2,0.11918079853057861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,8,power_law_1.2,0.12423039674758911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,32,power_law_1.2,0.21132159233093262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,8,power_law_1.2,0.12959359884262084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,8,power_law_1.2,0.13815679550170898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,32,power_law_1.2,0.25753600597381593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,8,power_law_1.2,0.14680960178375244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,8,power_law_1.2,0.1385599970817566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,32,power_law_1.2,0.3249216079711914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,8,power_law_1.2,0.1587007999420166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,8,power_law_1.2,0.16336640119552612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,32,power_law_1.2,0.4611775875091553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,8,power_law_1.2,0.18462719917297363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,8,power_law_1.2,0.21822719573974608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,32,power_law_1.2,0.596889591217041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,8,power_law_1.2,0.2871232032775879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,8,power_law_1.2,0.31868801116943357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,32,power_law_1.2,0.752953577041626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,8,power_law_1.2,0.44936318397521974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,8,power_law_1.2,0.5630784034729004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,32,power_law_1.2,1.0499648094177245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,8,power_law_1.2,0.8066368103027344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,8,power_law_1.2,1.0590335845947265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,32,power_law_1.2,1.7121856689453125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,8,power_law_1.2,1.5942272186279296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,8,power_law_1.2,1.9341951370239259
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,32,power_law_1.2,3.344121551513672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,8,power_law_1.01,0.03537279963493347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,8,power_law_1.01,0.03265919983386993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,8,power_law_1.01,0.030905601382255555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,8,power_law_1.01,0.03246079981327057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,8,power_law_1.2,4.421152114868164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,8,power_law_1.01,0.03290880024433136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,8,power_law_1.01,0.03720319867134094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,8,power_law_1.01,0.03761920034885406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,8,power_law_1.01,0.037964800000190736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,8,power_law_1.01,0.038649600744247434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,8,power_law_1.01,0.03941119909286499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,8,power_law_1.01,0.04116480052471161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,8,power_law_1.01,0.042342400550842284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,8,power_law_1.01,0.044684800505638125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,8,power_law_1.01,0.04341759979724884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,8,power_law_1.01,0.047091200947761536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,8,power_law_1.01,0.049235200881958006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,8,power_law_1.01,0.05378559827804565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,8,power_law_1.01,0.0615231990814209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,8,power_law_1.01,0.06880000233650208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,8,power_law_1.01,0.09827200174331666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,8,power_law_1.01,0.10683519840240478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,16,8,power_law_1.2,0.03624959886074066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,16,8,power_law_1.2,0.033817601203918454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,8,power_law_1.01,0.1400704026222229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,16,8,power_law_1.2,0.032025599479675294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,16,8,power_law_1.2,0.031615999341011045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,16,8,power_law_1.2,0.03289600014686585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,8,power_law_1.01,0.17485439777374268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,16,8,power_law_1.2,0.03313280045986176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,16,8,power_law_1.2,0.03367680013179779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,16,8,power_law_1.2,0.03424000144004822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,16,8,power_law_1.2,0.03563520014286041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,8,power_law_1.01,0.22880001068115235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,16,8,power_law_1.2,0.035519999265670774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,16,8,power_law_1.2,0.03555839955806732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,16,8,power_law_1.2,0.03663359880447388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,16,8,power_law_1.2,0.03696640133857727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,8,power_law_1.01,0.3121920108795166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,16,8,power_law_1.2,0.036320000886917114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,16,8,power_law_1.2,0.040940800309181215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,16,8,power_law_1.2,0.042131200432777405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,16,8,power_law_1.2,0.04450559914112091
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,16,8,power_law_1.2,0.051571202278137204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,8,power_law_1.01,0.47745280265808104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,16,8,power_law_1.2,0.05502079725265503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,16,8,power_law_1.2,0.06439039707183838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,16,8,power_law_1.2,0.07118719816207886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,16,8,power_law_1.2,0.09246079921722412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,16,8,power_law_1.2,0.11333760023117065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,8,power_law_1.01,0.5203199863433838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,16,8,power_law_1.2,0.15703680515289306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,16,8,power_law_1.2,0.1942720055580139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,16,8,power_law_1.2,0.2588288068771362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,16,8,power_law_1.2,0.33510398864746094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,8,power_law_1.01,0.7000895977020264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,16,8,power_law_1.2,0.43114237785339354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,16,8,power_law_1.2,0.6543615818023681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,8,power_law_1.01,1.3014592170715331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,16,8,power_law_1.2,1.2171327590942382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,8,power_law_1.01,2.438969612121582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,32,power_law_1.01,0.034534400701522826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,32,power_law_1.01,0.032153600454330446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,32,power_law_1.01,0.029190400242805482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,32,power_law_1.01,0.02815360128879547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,32,power_law_1.01,0.03054080009460449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,32,power_law_1.01,0.03155840039253235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,32,power_law_1.01,0.031481599807739256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,32,power_law_1.01,0.03196159899234772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,32,power_law_1.01,0.03190400004386902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,32,power_law_1.01,0.03184640109539032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,32,power_law_1.01,0.03273600041866302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,32,power_law_1.01,0.03387520015239716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,32,power_law_1.01,0.03447679877281189
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,32,power_law_1.01,0.03578239977359772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,32,power_law_1.01,0.038815999031066896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,32,power_law_1.01,0.04164479970932007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,32,power_law_1.01,0.04284160137176514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,32,power_law_1.01,0.0459199994802475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,32,power_law_1.01,0.05296639800071716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,32,power_law_1.01,0.0596671998500824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,32,power_law_1.01,0.06681600213050842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,32,power_law_1.01,0.08737919926643371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,32,power_law_1.01,0.10455039739608765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,32,power_law_1.01,0.12608640193939208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,32,power_law_1.01,0.15092480182647705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,32,power_law_1.01,0.21566081047058105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,32,power_law_1.01,0.30804479122161865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,32,power_law_1.01,0.3420095920562744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,32,power_law_1.01,0.4982592105865479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,32,power_law_1.01,0.93438720703125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,32,power_law_1.2,0.09582080245018006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,32,power_law_1.2,0.10931839942932128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,32,power_law_1.2,0.09916160106658936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,32,power_law_1.2,0.07646080255508422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,32,power_law_1.2,0.07950080037117005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,32,power_law_1.2,0.08819839954376221
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,32,power_law_1.2,0.09173120260238647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,32,power_law_1.2,0.09370239973068237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,32,power_law_1.2,0.09523199796676636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,32,power_law_1.2,0.09322879910469055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,32,power_law_1.2,0.09778559803962708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,32,power_law_1.2,0.10212479829788208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,32,power_law_1.2,0.1015552043914795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,32,power_law_1.2,0.10213119983673095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,32,power_law_1.2,0.10874240398406983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,32,power_law_1.2,0.11119359731674194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,32,power_law_1.2,0.11385600566864014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,32,power_law_1.2,0.12136319875717164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,32,power_law_1.2,0.13429759740829467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,32,power_law_1.2,0.1538815975189209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,32,power_law_1.2,0.16853760480880736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,32,power_law_1.2,0.22950398921966553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,32,power_law_1.2,0.2599936008453369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,32,power_law_1.2,0.3605184078216553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,32,power_law_1.2,0.43337597846984866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,32,power_law_1.2,0.5979135990142822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,32,power_law_1.2,0.806118392944336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,32,power_law_1.2,0.9992832183837891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,32,power_law_1.2,1.5943296432495118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,32,power_law_1.2,3.475116729736328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,8,balanced,0.04993066688378652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,8,balanced,0.05193066596984863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,8,balanced,0.054101333022117615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,8,balanced,0.07930666704972585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,8,balanced,0.09624532858530681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,8,balanced,0.11127466956774394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,8,balanced,0.11220799883206685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,8,balanced,0.11121066411336263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,8,balanced,0.11196266611417134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,8,balanced,0.11238400141398112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,8,balanced,0.11307733257611592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,8,balanced,0.11595199505488078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,8,balanced,0.11750933527946472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,8,balanced,0.11973333358764648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,8,balanced,0.12615999579429626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,8,balanced,0.12990400195121765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,8,balanced,0.13402133186658224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,8,balanced,0.15185599525769553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,8,balanced,0.16150933504104614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,8,balanced,0.20963732401529947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,8,balanced,0.22829333941141763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,8,balanced,0.308405339717865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,8,balanced,0.30985067288080853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,8,balanced,0.4376159906387329
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,8,balanced,0.5129333337148031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,8,balanced,0.7315413157145182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,8,balanced,0.8871839841206869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,4,8,balanced,1.108682632446289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,4,8,balanced,1.6428640683492024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,4,8,balanced,3.3208160400390625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,32,balanced,0.030591999491055805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,32,balanced,0.027621333797772724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,32,balanced,0.029472000896930695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,32,balanced,0.030095999439557392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,32,balanced,0.029818666477998097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,32,balanced,0.03145066648721695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,32,balanced,0.0317546675602595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,32,balanced,0.03181866556406021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,32,balanced,0.033488000432650246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,32,balanced,0.03180266668399175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,32,balanced,0.032618666688601174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,32,balanced,0.03405333310365677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,32,balanced,0.03368533402681351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,32,balanced,0.03357866654793421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,32,balanced,0.035749333600203194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,32,balanced,0.037776000797748566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,32,balanced,0.037989333271980286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,32,balanced,0.03958933303753535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,32,balanced,0.03980266551176707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,32,balanced,0.04385599990685781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,32,balanced,0.04452266792456309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,32,balanced,0.04990933338801066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,32,balanced,0.05630399783452352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,32,balanced,0.07455466687679291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,32,balanced,0.09658132990201314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,32,balanced,0.1358453333377838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,32,balanced,0.167194664478302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,32,balanced,0.19780266284942627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,32,balanced,0.2863413294156392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,32,balanced,0.5151146650314331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,2,4,power_law_1.2,0.11867519617080688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,2,4,power_law_1.2,0.12801920175552367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,2,4,power_law_1.2,0.10809600353240967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,2,4,power_law_1.2,0.11780480146408082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,2,4,power_law_1.2,0.11822079420089722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,2,4,power_law_1.2,0.12431360483169555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,2,4,power_law_1.2,0.14333440065383912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,2,4,power_law_1.2,0.12768640518188476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,2,4,power_law_1.2,0.1427135944366455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,2,4,power_law_1.2,0.1428096055984497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,2,4,power_law_1.2,0.14202879667282103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,2,4,power_law_1.2,0.15512319803237914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,2,4,power_law_1.2,0.16097279787063598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,2,4,power_law_1.2,0.1609344005584717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,2,4,power_law_1.2,0.18149759769439697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,2,4,power_law_1.2,0.2136768102645874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,2,4,power_law_1.2,0.2180799961090088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,2,4,power_law_1.2,0.2500607967376709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,2,4,power_law_1.2,0.29344639778137205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,2,4,power_law_1.2,0.39982080459594727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,2,4,power_law_1.2,0.5159103870391846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,2,4,power_law_1.2,0.8231103897094727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,2,4,power_law_1.2,0.8529279708862305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,2,4,power_law_1.2,1.3546175956726074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,2,4,power_law_1.2,2.0297279357910156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,2,4,power_law_1.2,3.0609600067138674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,2,4,power_law_1.2,4.042726516723633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,2,4,power_law_1.2,4.027443313598633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,2,4,power_law_1.2,7.863085174560547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,2,4,power_law_1.2,15.855244445800782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,32,power_law_1.01,0.04407680034637451
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,32,power_law_1.01,0.04035840034484863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,32,power_law_1.01,0.03116160035133362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,32,power_law_1.01,0.030630400776863097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,32,power_law_1.01,0.031200000643730165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,32,power_law_1.01,0.03304960131645203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,32,power_law_1.01,0.03268480002880096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,32,power_law_1.01,0.03281280100345611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,32,power_law_1.01,0.033024001121521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,32,power_law_1.01,0.033062401413917544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,32,power_law_1.01,0.035129600763320924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,32,power_law_1.01,0.03932799994945526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,32,power_law_1.01,0.04066559970378876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,32,power_law_1.01,0.04454399943351746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,16,balanced,0.0378506655494372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,16,balanced,0.034815999368826546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,32,power_law_1.01,0.04830079972743988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,16,balanced,0.034341332813103996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,16,balanced,0.03611200054486593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,16,balanced,0.03621333340803782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,16,balanced,0.038149334490299225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,16,balanced,0.036277333895365395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,16,balanced,0.037018666664759316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,16,balanced,0.038149334490299225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,32,power_law_1.01,0.049267199635505673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,16,balanced,0.03799466788768768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,16,balanced,0.03822933385769526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,16,balanced,0.0383093332250913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,16,balanced,0.038047999143600464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,16,balanced,0.04043733328580856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,16,balanced,0.04225599765777588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,16,balanced,0.042352000872294106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,32,power_law_1.01,0.05159040093421936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,16,balanced,0.04426133135954539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,16,balanced,0.04615999758243561
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,16,balanced,0.046442667643229164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,16,balanced,0.05053866902987162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,16,balanced,0.056346664826075234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,16,balanced,0.06117866436640421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,32,power_law_1.01,0.056857597827911374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,16,balanced,0.06875200072924297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,16,balanced,0.08698667089144389
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,16,balanced,0.11019733548164368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,16,balanced,0.14448533455530801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,32,power_law_1.01,0.06157439947128296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,16,balanced,0.1771999994913737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,16,balanced,0.21121599276860556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,16,balanced,0.3022720019022624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,16,balanced,0.5585653384526571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,32,power_law_1.01,0.07667199969291687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,32,power_law_1.01,0.08273280262947083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,32,power_law_1.01,0.11118079423904419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,32,power_law_1.01,0.11745280027389526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,32,power_law_1.01,0.14604159593582153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,32,power_law_1.01,0.17989120483398438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,32,power_law_1.01,0.29789440631866454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,32,power_law_1.01,0.31626880168914795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,32,power_law_1.01,0.3734015941619873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,32,power_law_1.01,0.546342420578003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,32,power_law_1.01,1.1424639701843262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,2,4,power_law_1.01,0.07246080040931702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,2,4,power_law_1.01,0.06392959952354431
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,2,4,power_law_1.01,0.0769599974155426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,2,4,power_law_1.01,0.07976959943771363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,2,4,power_law_1.01,0.07925760149955749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,2,4,power_law_1.01,0.08302720189094544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,2,4,power_law_1.01,0.08880640268325805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,2,4,power_law_1.01,0.08312320113182067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,2,4,power_law_1.01,0.08735359907150268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,2,4,power_law_1.01,0.09232640266418457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,2,4,power_law_1.01,0.08843520283699036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,2,4,power_law_1.01,0.09452800154685974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,2,4,power_law_1.01,0.09831680059432983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,2,4,power_law_1.01,0.10301439762115479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,2,4,power_law_1.01,0.11008000373840332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,2,4,power_law_1.01,0.12689919471740724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,2,4,power_law_1.01,0.13893120288848876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,2,4,power_law_1.01,0.1698367953300476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,2,4,power_law_1.01,0.18350720405578613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,2,4,power_law_1.01,0.23192958831787108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,2,4,power_law_1.01,0.3039423942565918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,2,4,power_law_1.01,0.35043199062347413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,2,4,power_law_1.01,0.5094655990600586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,2,4,power_law_1.01,0.9050304412841796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,2,4,power_law_1.01,1.159763240814209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,2,4,power_law_1.01,1.7226816177368165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,1,power_law_1.2,0.03198719918727875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,1,power_law_1.2,0.03454079926013946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,2,4,power_law_1.01,2.2674240112304687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,1,power_law_1.2,0.03951359987258911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,1,power_law_1.2,0.04429439902305603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,2,4,power_law_1.01,2.803980827331543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,1,power_law_1.2,0.054118400812149046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,1,power_law_1.2,0.07370880246162415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,2,4,power_law_1.01,4.049983978271484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,1,power_law_1.2,0.07817599773406983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,1,power_law_1.2,0.08115199804306031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,1,power_law_1.2,0.08346880078315735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,1,power_law_1.2,0.08702719807624817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,1,power_law_1.2,0.0907904028892517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,2,4,power_law_1.01,7.674028778076172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,1,power_law_1.2,0.09686400294303894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,1,power_law_1.2,0.10252159833908081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,1,power_law_1.2,0.10298880338668823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,1,power_law_1.2,0.11152000427246093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,1,power_law_1.2,0.1183359980583191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,1,power_law_1.2,0.12780159711837769
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,1,power_law_1.2,0.15900800228118897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,1,power_law_1.2,0.18937599658966064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,1,power_law_1.2,0.24304640293121338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,1,power_law_1.2,0.2946367979049683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,1,power_law_1.2,0.398905611038208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,1,power_law_1.2,0.504313611984253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,1,power_law_1.2,0.7078591823577881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,1,power_law_1.2,0.9221247673034668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,1,power_law_1.2,1.3356351852416992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,1,power_law_1.2,1.7408000946044921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,1,8,power_law_1.01,0.06766080260276794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,1,8,power_law_1.01,0.06593279838562012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,1,8,power_law_1.01,0.06638720035552978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,1,power_law_1.2,2.1669567108154295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,1,8,power_law_1.01,0.06666880249977111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,1,8,power_law_1.01,0.06607360243797303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,1,8,power_law_1.01,0.06570240259170532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,1,8,power_law_1.01,0.06833919882774353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,1,8,power_law_1.01,0.07100800275802613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,1,power_law_1.2,3.3892353057861326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,1,8,power_law_1.01,0.07347840070724487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,1,8,power_law_1.01,0.07704960107803345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,1,8,power_law_1.01,0.07287039756774902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,1,8,power_law_1.01,0.08121600151062011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,1,8,power_law_1.01,0.08076159954071045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,1,8,power_law_1.01,0.08318719863891602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,1,8,power_law_1.01,0.10115200281143188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,1,8,power_law_1.01,0.11733759641647339
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,1,8,power_law_1.01,0.11880960464477539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,1,8,power_law_1.01,0.1692031979560852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,1,power_law_1.2,6.687238311767578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,1,8,power_law_1.01,0.18140159845352172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,1,8,power_law_1.01,0.2163327932357788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,1,8,power_law_1.01,0.2871936082839966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,1,8,power_law_1.01,0.3368767976760864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,1,8,power_law_1.01,0.48981761932373047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,1,8,power_law_1.01,0.7829184055328369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,1,8,power_law_1.01,0.9550848007202148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,1,8,power_law_1.01,1.4495295524597167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,1,8,power_law_1.01,1.8278848648071289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,1,8,power_law_1.01,2.0954496383666994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,1,8,power_law_1.01,3.469075012207031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,1,8,power_law_1.01,7.7587135314941404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,4,1,power_law_1.2,0.047014400362968445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,4,1,power_law_1.2,0.056409597396850586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,4,1,power_law_1.2,0.060761600732803345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,4,1,power_law_1.2,0.0890175998210907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,4,1,power_law_1.2,0.08784000277519226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,4,1,power_law_1.2,0.08986240029335021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,4,1,power_law_1.2,0.09096959829330445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,4,1,power_law_1.2,0.09208959937095643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,4,1,power_law_1.2,0.09260799884796142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,4,1,power_law_1.2,0.09266560077667237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,4,1,power_law_1.2,0.09445760250091553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,4,1,power_law_1.2,0.09608319997787476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,4,1,power_law_1.2,0.09966719746589661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,4,1,power_law_1.2,0.10401920080184937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,4,1,power_law_1.2,0.11870720386505126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,4,1,power_law_1.2,0.11980799436569214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,4,1,power_law_1.2,0.12893439531326295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,4,1,power_law_1.2,0.1511680006980896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,4,1,power_law_1.2,0.17299840450286866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,4,1,power_law_1.2,0.21269121170043945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,4,1,power_law_1.2,0.2410048007965088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,4,1,power_law_1.2,0.3142528057098389
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,4,1,power_law_1.2,0.3822335958480835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,4,1,power_law_1.2,0.5237120151519775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,4,1,power_law_1.2,0.6629695892333984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,4,1,power_law_1.2,0.931385612487793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,4,1,power_law_1.2,1.214303970336914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,4,1,power_law_1.2,1.4939135551452636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,4,1,power_law_1.2,2.331052780151367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,4,1,power_law_1.2,4.560281753540039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,2,power_law_1.01,0.06749439835548401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,2,power_law_1.01,0.07232000231742859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,2,power_law_1.01,0.08729599714279175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,2,power_law_1.01,0.08716800212860107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,2,power_law_1.01,0.11329920291900634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,4,2,power_law_1.01,0.03671680092811584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,4,2,power_law_1.01,0.04255360066890716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,4,2,power_law_1.01,0.04958719909191132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,4,2,power_law_1.01,0.05692160129547119
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,4,2,power_law_1.01,0.055180799961090085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,4,2,power_law_1.01,0.058713597059249875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,4,2,power_law_1.01,0.06161919832229614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,4,2,power_law_1.01,0.061343997716903687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,4,2,power_law_1.01,0.06320000290870667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,4,2,power_law_1.01,0.06416000127792358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,2,power_law_1.01,0.1411712050437927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,4,2,power_law_1.01,0.06574079990386963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,4,2,power_law_1.01,0.06700159907341004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,4,2,power_law_1.01,0.06824960112571717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,4,2,power_law_1.01,0.07704960107803345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,4,2,power_law_1.01,0.08541439771652222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,4,2,power_law_1.01,0.09285759925842285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,4,2,power_law_1.01,0.08499199748039246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,4,2,power_law_1.01,0.0998080015182495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,4,2,power_law_1.01,0.11970560550689698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,4,2,power_law_1.01,0.14243199825286865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,2,power_law_1.01,0.15818239450454713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,4,2,power_law_1.01,0.16372480392456054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,4,2,power_law_1.01,0.23701119422912598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,4,2,power_law_1.01,0.29351038932800294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,4,2,power_law_1.01,0.40769281387329104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,4,2,power_law_1.01,0.46472320556640623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,4,2,power_law_1.01,0.7371391773223877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,2,power_law_1.01,0.18010239601135253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,4,2,power_law_1.01,1.079411220550537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,4,2,power_law_1.01,1.17958402633667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,4,2,power_law_1.01,1.830169677734375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,2,power_law_1.01,0.1897536039352417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,4,2,power_law_1.01,3.6842689514160156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,2,power_law_1.01,0.1964735984802246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,2,power_law_1.01,0.20026240348815919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,2,power_law_1.01,0.20889599323272706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,2,power_law_1.01,0.2120448112487793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,2,power_law_1.01,0.21778559684753418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,2,power_law_1.01,0.2351680040359497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,2,power_law_1.01,0.2370368003845215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,64,power_law_1.01,0.06947839856147767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,2,power_law_1.01,0.249452805519104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,64,power_law_1.01,0.06986240148544312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,2,power_law_1.01,0.27508480548858644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,64,power_law_1.01,0.06125440001487732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,64,power_law_1.01,0.0609279990196228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,2,power_law_1.01,0.3038912057876587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,64,power_law_1.01,0.06469759941101075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,64,power_law_1.01,0.06711040139198303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,2,power_law_1.01,0.35647358894348147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,64,power_law_1.01,0.06735360026359558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,64,power_law_1.01,0.0702015995979309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,2,power_law_1.01,0.4025856018066406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,64,power_law_1.01,0.06685439944267273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,64,power_law_1.01,0.07096319794654846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,2,power_law_1.01,0.5229311943054199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,64,power_law_1.01,0.06920959949493408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,64,power_law_1.01,0.0710591971874237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,2,power_law_1.01,0.6212287902832031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,64,power_law_1.01,0.0741312026977539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,64,power_law_1.01,0.07661439776420594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,2,power_law_1.01,0.8355263710021973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,64,power_law_1.01,0.08044160008430482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,64,power_law_1.01,0.08241919875144958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,64,power_law_1.01,0.08803200125694274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,2,power_law_1.01,1.1104640007019042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,64,power_law_1.01,0.10301439762115479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,64,power_law_1.01,0.11026560068130493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,2,power_law_1.01,1.5138943672180176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,64,power_law_1.01,0.13230719566345214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,64,power_law_1.01,0.1537279963493347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,64,power_law_1.01,0.20380799770355223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,2,power_law_1.01,1.9855615615844726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,64,power_law_1.01,0.24399359226226808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,64,power_law_1.01,0.33196799755096434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,2,power_law_1.01,2.387276840209961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,64,power_law_1.01,0.43528318405151367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,64,power_law_1.01,0.600710391998291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,2,power_law_1.01,3.7071998596191404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,64,power_law_1.01,0.7754816055297852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,64,power_law_1.01,0.9925696372985839
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,64,power_law_1.01,1.624166488647461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,2,power_law_1.01,6.806566619873047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,64,power_law_1.01,3.195071983337402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,2,power_law_1.2,0.048153600096702574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,32,balanced,0.02940266579389572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,32,balanced,0.029450667401154835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,32,balanced,0.03086400032043457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,32,balanced,0.029685333371162415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,32,balanced,0.031445334355036415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,32,balanced,0.03311466674009959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,32,balanced,0.0334346666932106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,32,balanced,0.033626665671666466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,32,balanced,0.0317546675602595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,32,balanced,0.033376000821590424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,2,power_law_1.2,0.050310397148132326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,32,balanced,0.03356266766786575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,32,balanced,0.03384000062942505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,32,balanced,0.03422933320204417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,32,balanced,0.03362133353948593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,32,balanced,0.03946666667858759
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,32,balanced,0.039093332986036934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,32,balanced,0.038032000263532005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,32,balanced,0.0397119993964831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,32,balanced,0.04167999823888143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,32,balanced,0.04372799893220266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,32,balanced,0.04586666822433472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,32,balanced,0.053930665055910744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,32,balanced,0.06010666489601135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,32,balanced,0.086325337489446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,32,balanced,0.10941867033640544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,32,balanced,0.1483626663684845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,32,balanced,0.17697066068649292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,2,power_law_1.2,0.062067198753356936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,32,balanced,0.21256534258524576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,32,balanced,0.29628799359003705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,32,balanced,0.5465813477834066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,2,power_law_1.2,0.07922559976577759
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,2,power_law_1.2,0.08908159732818603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,2,power_law_1.2,0.134278404712677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,2,power_law_1.2,0.14249600172042848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,2,power_law_1.2,0.14648959636688233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,2,power_law_1.2,0.1487615942955017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,2,power_law_1.2,0.15456639528274535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,2,power_law_1.2,0.1544319987297058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,2,power_law_1.2,0.15852160453796388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,2,power_law_1.2,0.16238720417022706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,2,power_law_1.2,0.17191679477691652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,2,power_law_1.2,0.17485439777374268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,2,power_law_1.2,0.18544000387191772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,4,balanced,0.0354720006386439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,4,balanced,0.0663679987192154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,4,balanced,0.07684266567230225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,4,balanced,0.11921067039171855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,2,power_law_1.2,0.1960576057434082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,4,balanced,0.20415467023849487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,4,balanced,0.20219733317693075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,4,balanced,0.20428266127904257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,4,balanced,0.20530666907628378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,4,balanced,0.2066239913304647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,4,balanced,0.20599999030431113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,4,balanced,0.2054133415222168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,4,balanced,0.20641066630681357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,4,balanced,0.20879467328389487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,4,balanced,0.21006399393081665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,4,balanced,0.21661865711212158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,2,power_law_1.2,0.22805120944976806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,4,balanced,0.21846399704615274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,4,balanced,0.2227893273035685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,4,balanced,0.23143466313680014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,4,balanced,0.2401813268661499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,4,balanced,0.2667626738548279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,4,balanced,0.26976533730824787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,4,balanced,0.32215466101964313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,4,balanced,0.3399893442789714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,4,balanced,0.4452693462371826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,2,power_law_1.2,0.25319681167602537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,4,balanced,0.4960533380508423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,4,balanced,0.6642666657765707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,4,balanced,0.8431839942932129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,4,balanced,1.028117338816325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,2,power_law_1.2,0.3008064031600952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,4,balanced,1.5684107144673665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,4,balanced,3.0353759129842124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,2,power_law_1.2,0.3578687906265259
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,2,power_law_1.2,0.41984000205993655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,2,power_law_1.2,0.524345588684082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,2,power_law_1.2,0.6909567832946777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,2,power_law_1.2,0.8950719833374023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,2,power_law_1.2,1.3123968124389649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,2,power_law_1.2,1.6980031967163085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,2,power_law_1.2,2.133523178100586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,2,power_law_1.2,3.1990591049194337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,1,balanced,0.04991999765237173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,1,balanced,0.055914665261904396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,1,balanced,0.06453333298365276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,1,balanced,0.0912000040213267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,1,balanced,0.1360106666882833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,1,balanced,0.13828266660372415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,1,balanced,0.14325333635012308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,1,balanced,0.1422826647758484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,1,balanced,0.14404267072677612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,1,balanced,0.14601066708564758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,1,balanced,0.14757333199183145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,1,balanced,0.15271466970443726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,1,balanced,0.15331199765205383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,1,balanced,0.16148266196250916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,1,balanced,0.16685332854588827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,1,balanced,0.17641599973042807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,1,balanced,0.18741865952809653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,1,balanced,0.21802133321762085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,1,balanced,0.2406346599260966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,1,balanced,0.2977546652158101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,1,balanced,0.3497866789499919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,1,balanced,0.5076053142547607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,1,balanced,0.6098346710205078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,1,balanced,0.8893226782480875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,1,balanced,1.1309066613515217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,1,balanced,1.666320006052653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,1,balanced,2.20198392868042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,1,balanced,2.734421412150065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,2,power_law_1.2,6.140857696533203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,1,balanced,4.188864072163899
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,1,balanced,8.206192016601562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,32,power_law_1.2,0.06429439783096313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,32,power_law_1.2,0.06229760050773621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,32,power_law_1.2,0.04568319916725159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,32,power_law_1.2,0.047539201378822324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,32,power_law_1.2,0.04780800044536591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,32,power_law_1.2,0.05434240102767944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,32,power_law_1.2,0.05430399775505066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,32,power_law_1.2,0.055302399396896365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,32,power_law_1.2,0.05808640122413635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,32,power_law_1.2,0.056518399715423585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,32,power_law_1.2,0.0574783980846405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,32,power_law_1.2,0.056480002403259275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,32,power_law_1.2,0.05885440111160278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,32,power_law_1.2,0.06142079830169678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,32,power_law_1.2,0.06806399822235107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,32,power_law_1.2,0.06912639737129211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,32,power_law_1.2,0.07452160120010376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,32,power_law_1.2,0.0861952006816864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,32,power_law_1.2,0.0897599995136261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,32,power_law_1.2,0.11725440025329589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,32,power_law_1.2,0.13647359609603882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,32,power_law_1.2,0.16714240312576295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,32,power_law_1.2,0.1906175971031189
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,32,power_law_1.2,0.27013120651245115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,32,power_law_1.2,0.37036800384521484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,32,power_law_1.2,0.4764736175537109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,32,power_law_1.2,0.5764800071716308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,32,power_law_1.2,0.8758208274841308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,32,power_law_1.2,1.1791808128356933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,32,power_law_1.2,2.3928512573242187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,4,power_law_1.01,0.0964031994342804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,16,balanced,0.0436160018046697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,16,balanced,0.04391466577847799
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,16,balanced,0.04372799893220266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,16,balanced,0.04576533536116282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,16,balanced,0.04782933493455251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,16,balanced,0.049829334020614624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,16,balanced,0.05193066596984863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,16,balanced,0.052144000927607216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,16,balanced,0.05194666484991709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,16,balanced,0.05386666456858317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,16,balanced,0.05393599967161814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,16,balanced,0.052095999320348106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,16,balanced,0.054469332098960876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,16,balanced,0.05794133245944977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,16,balanced,0.06118399898211161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,4,power_law_1.01,0.1456447958946228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,16,balanced,0.06215466558933258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,16,balanced,0.0664106657107671
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,16,balanced,0.07236800094445546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,16,balanced,0.07879999776681264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,16,balanced,0.11002133289972942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,16,balanced,0.12553067008654276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,16,balanced,0.1646293302377065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,16,balanced,0.1876159906387329
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,16,balanced,0.2622720003128052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,16,balanced,0.3137226700782776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,16,balanced,0.4387306769688924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,16,balanced,0.5482399861017863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,16,balanced,0.664138674736023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,16,balanced,1.0170559883117676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,4,power_law_1.01,0.10705280303955078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,16,balanced,1.9841972986857097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,4,power_law_1.01,0.13664640188217164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,4,power_law_1.01,0.18234239816665648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,4,power_law_1.01,0.23788158893585204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,32,4,power_law_1.01,0.03429119884967804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,4,power_law_1.01,0.33327360153198243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,32,4,power_law_1.01,0.032313600182533264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,32,4,power_law_1.01,0.03076480031013489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,32,4,power_law_1.01,0.03274880051612854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,4,power_law_1.01,0.33615360260009763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,32,4,power_law_1.01,0.03427839875221252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,32,4,power_law_1.01,0.036959999799728395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,4,power_law_1.01,0.3612351894378662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,32,4,power_law_1.01,0.03781760036945343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,32,4,power_law_1.01,0.03832319974899292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,32,4,power_law_1.01,0.03868800103664398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,4,power_law_1.01,0.35089280605316164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,32,4,power_law_1.01,0.03903360068798065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,32,4,power_law_1.01,0.039680001139640805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,4,power_law_1.01,0.3825599908828735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,32,4,power_law_1.01,0.04047360122203827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,1,16,power_law_1.2,0.028358399868011475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,32,4,power_law_1.01,0.041945600509643556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,1,16,power_law_1.2,0.027001601457595826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,32,4,power_law_1.01,0.0455487996339798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,4,power_law_1.01,0.39001600742340087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,1,16,power_law_1.2,0.026035198569297792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,32,4,power_law_1.01,0.0488319993019104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,1,16,power_law_1.2,0.025785601139068602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,32,4,power_law_1.01,0.05144960284233093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,1,16,power_law_1.2,0.031142398715019226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,4,power_law_1.01,0.3990015983581543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,32,4,power_law_1.01,0.057599997520446776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,1,16,power_law_1.2,0.029971200227737426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,32,4,power_law_1.01,0.06280320286750793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,1,16,power_law_1.2,0.029926401376724244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,32,4,power_law_1.01,0.07559040188789368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,4,power_law_1.01,0.40369281768798826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,1,16,power_law_1.2,0.03335680067539215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,32,4,power_law_1.01,0.08899199962615967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,1,16,power_law_1.2,0.03521920144557953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,32,4,power_law_1.01,0.10604159832000733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,1,16,power_law_1.2,0.03576320111751556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,4,power_law_1.01,0.41950078010559083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,32,4,power_law_1.01,0.13326719999313355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,1,16,power_law_1.2,0.03925760090351105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,1,16,power_law_1.2,0.03980799913406372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,32,4,power_law_1.01,0.2063040018081665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,4,power_law_1.01,0.42421760559082033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,1,16,power_law_1.2,0.043756800889968875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,32,4,power_law_1.01,0.2659584045410156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,1,16,power_law_1.2,0.041043201088905336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,1,16,power_law_1.2,0.047391998767852786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,32,4,power_law_1.01,0.3542655944824219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,4,power_law_1.01,0.45362558364868166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,1,16,power_law_1.2,0.0490880012512207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,32,4,power_law_1.01,0.4569983959197998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,1,16,power_law_1.2,0.05124480128288269
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,1,16,power_law_1.2,0.05534719824790955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,4,power_law_1.01,0.49429121017456057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,16,power_law_1.2,0.05924479961395264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,32,4,power_law_1.01,0.6044735908508301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,16,power_law_1.2,0.06782079935073852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,4,power_law_1.01,0.5040575981140136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,16,power_law_1.2,0.07127040028572082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,32,4,power_law_1.01,0.7019519805908203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,16,power_law_1.2,0.1017151951789856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,16,power_law_1.2,0.1212928056716919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,4,power_law_1.01,0.5799871921539307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,32,4,power_law_1.01,1.215993595123291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,16,power_law_1.2,0.15932799577713014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,16,power_law_1.2,0.21531519889831544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,4,power_law_1.01,0.6032896041870117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,16,power_law_1.2,0.2955904006958008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,4,power_law_1.01,0.7147007942199707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,32,4,power_law_1.01,2.229984092712402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,16,power_law_1.2,0.3384959936141968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,4,power_law_1.01,0.8141951560974121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,16,power_law_1.2,0.4783616065979004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,4,power_law_1.01,1.0808064460754394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,16,power_law_1.2,0.7839231967926026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,4,power_law_1.01,1.2566720008850099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,16,power_law_1.2,1.7441535949707032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,4,power_law_1.01,1.6755647659301758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,4,power_law_1.01,2.0162559509277345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,4,power_law_1.01,2.5709503173828123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,2,8,power_law_1.2,0.03925119936466217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,2,8,power_law_1.2,0.04201599955558777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,2,8,power_law_1.2,0.039654400944709775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,2,8,power_law_1.2,0.039705601334571836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,2,8,power_law_1.2,0.039699199795722964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,2,8,power_law_1.2,0.04042240083217621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,2,8,power_law_1.2,0.040703999996185306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,2,8,power_law_1.2,0.04127359986305237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,2,8,power_law_1.2,0.04243200123310089
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,4,power_law_1.01,3.731827163696289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,2,8,power_law_1.2,0.04351359903812409
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,2,8,power_law_1.2,0.045158401131629944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,2,8,power_law_1.2,0.04873600006103516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,2,8,power_law_1.2,0.050355201959609984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,2,8,power_law_1.2,0.05135999917984009
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,2,8,power_law_1.2,0.060038399696350095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,2,8,power_law_1.2,0.06521599888801574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,2,8,power_law_1.2,0.06686720252037048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,2,8,power_law_1.2,0.0818943977355957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,2,8,power_law_1.2,0.09944959878921508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,2,8,power_law_1.2,0.11747200489044189
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,2,8,power_law_1.2,0.1230463981628418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,2,8,power_law_1.2,0.18315520286560058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,2,8,power_law_1.2,0.2331775903701782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,2,8,power_law_1.2,0.2938431978225708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,8,balanced,0.06098133325576782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,8,balanced,0.07146133482456207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,8,balanced,0.08585066596666972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,2,8,power_law_1.2,0.37150719165802004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,8,balanced,0.11857600013415019
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,8,balanced,0.15980799992879233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,8,balanced,0.19242133696873984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,8,balanced,0.1949653426806132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,8,balanced,0.19535466035207114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,2,8,power_law_1.2,0.5707776069641113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,8,balanced,0.19774399201075235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,8,balanced,0.19614932934443155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,8,balanced,0.20082134008407593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,8,balanced,0.20338666439056396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,8,balanced,0.20562134186426798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,2,8,power_law_1.2,0.7382656097412109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,8,balanced,0.2083573341369629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,8,balanced,0.21548267205556235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,8,balanced,0.21796266237894693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,4,power_law_1.01,6.928435516357422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,8,balanced,0.2269386649131775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,8,balanced,0.24251733223597208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,8,balanced,0.257152001063029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,2,8,power_law_1.2,0.8748607635498047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,8,balanced,0.3285920023918152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,8,balanced,0.33032532533009845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,8,balanced,0.469376007715861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,8,balanced,0.4659200112024943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,8,balanced,0.6746293703715006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,2,8,power_law_1.2,1.449708843231201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,8,balanced,0.7354880174001058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,8,balanced,1.0774346987406414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,8,balanced,1.2898026307423909
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,1,8,balanced,1.6166079839070637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,2,8,power_law_1.2,2.851519966125488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,1,8,balanced,2.448906739552816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,1,8,balanced,4.771029472351074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,4,2,power_law_1.01,0.028857600688934327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,4,2,power_law_1.01,0.02632960081100464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,4,2,power_law_1.01,0.030611199140548707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,2,1,power_law_1.2,0.11889280080795288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,2,1,power_law_1.2,0.17463040351867676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,4,2,power_law_1.01,0.03529599905014038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,2,1,power_law_1.2,0.20188159942626954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,2,1,power_law_1.2,0.35932800769805906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,2,1,power_law_1.2,0.35946879386901853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,2,power_law_1.01,0.051737600564956666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,4,2,power_law_1.01,0.040934398770332336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,2,1,power_law_1.2,0.3722111940383911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,2,1,power_law_1.2,0.3821183919906616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,2,1,power_law_1.2,0.3906048059463501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,2,power_law_1.01,0.06991360187530518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,2,1,power_law_1.2,0.3958976030349731
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,4,2,power_law_1.01,0.04822399914264679
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,2,1,power_law_1.2,0.4013951778411865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,2,1,power_law_1.2,0.3859647989273071
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,2,power_law_1.01,0.06807680130004883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,2,1,power_law_1.2,0.39609599113464355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,4,2,power_law_1.01,0.049747198820114136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,2,1,power_law_1.2,0.41066880226135255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,2,1,power_law_1.2,0.43318400382995603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,2,power_law_1.01,0.0864512026309967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,2,1,power_law_1.2,0.4462912082672119
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,4,2,power_law_1.01,0.053958398103713986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,2,1,power_law_1.2,0.47322878837585447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,2,1,power_law_1.2,0.4966271877288818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,2,power_law_1.01,0.10935039520263672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,2,1,power_law_1.2,0.5645376205444336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,2,1,power_law_1.2,0.6460224151611328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,4,2,power_law_1.01,0.055641597509384154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,2,power_law_1.01,0.15539200305938722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,2,1,power_law_1.2,0.7873856067657471
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,2,1,power_law_1.2,0.9086976051330566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,2,power_law_1.01,0.15612800121307374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,4,2,power_law_1.01,0.05644800066947937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,2,1,power_law_1.2,1.1934975624084472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,2,power_law_1.01,0.16662399768829345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,2,1,power_law_1.2,1.452883243560791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,4,2,power_law_1.01,0.05848960280418396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,2,power_law_1.01,0.17047040462493895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,2,1,power_law_1.2,2.048806381225586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,4,2,power_law_1.01,0.05845119953155518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,2,power_law_1.01,0.17288960218429567
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,2,1,power_law_1.2,2.665567970275879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,2,power_law_1.01,0.17671680450439453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,4,2,power_law_1.01,0.060083198547363284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,2,power_law_1.01,0.18454400300979615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,2,1,power_law_1.2,3.921171188354492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,4,2,power_law_1.01,0.06872320175170898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,2,power_law_1.01,0.18632320165634156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,4,2,power_law_1.01,0.07342720031738281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,2,power_law_1.01,0.19648640155792235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,2,1,power_law_1.2,5.109791946411133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,4,2,power_law_1.01,0.07642239928245545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,2,power_law_1.01,0.20683519840240477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,2,power_law_1.01,0.2191551923751831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,4,2,power_law_1.01,0.08010879755020142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,2,1,power_law_1.2,6.409996795654297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,2,power_law_1.01,0.22883200645446777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,4,2,power_law_1.01,0.08581119775772095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,2,power_law_1.01,0.2571903944015503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,2,power_law_1.01,0.10136959552764893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,2,power_law_1.01,0.2851263999938965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,2,power_law_1.01,0.11851520538330078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,2,power_law_1.01,0.3523519992828369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,2,1,power_law_1.2,9.985183715820312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,2,power_law_1.01,0.12226560115814208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,2,power_law_1.01,0.41609601974487304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,2,power_law_1.01,0.542080020904541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,2,power_law_1.01,0.15624959468841554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,2,power_law_1.01,0.6324031829833985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,2,power_law_1.01,0.19413119554519653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,2,power_law_1.01,0.8272128105163574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,2,power_law_1.01,0.2630975961685181
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,2,power_law_1.01,1.1376895904541016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,2,power_law_1.01,0.34270079135894777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,2,1,power_law_1.2,19.273370361328126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,2,power_law_1.01,1.4838335990905762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,2,power_law_1.01,0.4810431957244873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,2,power_law_1.01,0.6180928230285645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,2,power_law_1.01,1.8747072219848633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,2,power_law_1.01,0.7960383892059326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,2,power_law_1.01,2.3211904525756837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,2,power_law_1.01,1.2600831985473633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,2,power_law_1.01,3.6515201568603515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,2,power_law_1.01,2.1266752243041993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,2,power_law_1.01,6.821971130371094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,32,balanced,0.05881600081920624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,32,balanced,0.05994666616121928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,32,balanced,0.05991466840108236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,32,balanced,0.06260266900062561
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,32,balanced,0.07144000132878621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,32,balanced,0.0848640004793803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,32,balanced,0.1090186635653178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,32,balanced,0.10734933614730835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,32,balanced,0.10412800312042236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,32,balanced,0.10685867071151733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,32,balanced,0.10537599523862202
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,32,balanced,0.10415466626485188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,32,balanced,0.10728533069292705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,32,balanced,0.10716799894968669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,32,balanced,0.11110400160153706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,32,balanced,0.11339199542999268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,32,balanced,0.11737599968910217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,32,balanced,0.12385599811871846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,32,balanced,0.13425599535306296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,32,balanced,0.1495146652062734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,32,balanced,0.16531200210253397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,32,balanced,0.19778666893641153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,32,balanced,0.22592000166575113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,32,balanced,0.2884959975878398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,32,balanced,0.36613333225250244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,32,balanced,0.48847464720408124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,32,balanced,0.6347200075785319
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,32,balanced,0.7895253499348959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,32,balanced,1.1946346759796143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,32,balanced,2.3187626202901206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,32,4,power_law_1.2,0.04922879934310913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,32,4,power_law_1.2,0.04490239918231964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,8,power_law_1.2,0.032364800572395325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,8,power_law_1.2,0.03051519989967346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,32,4,power_law_1.2,0.0506816029548645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,8,power_law_1.2,0.030291199684143066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,32,4,power_law_1.2,0.05343359708786011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,8,power_law_1.2,0.031123200058937074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,8,power_law_1.2,0.03320960104465485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,32,4,power_law_1.2,0.05625600218772888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,8,power_law_1.2,0.036422398686408994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,8,power_law_1.2,0.03642880022525787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,32,4,power_law_1.2,0.06728320121765137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,8,power_law_1.2,0.03733119964599609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,8,power_law_1.2,0.03743360042572021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,32,4,power_law_1.2,0.06915839910507202
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,8,power_law_1.2,0.03797119855880737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,8,power_law_1.2,0.03976959884166718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,32,4,power_law_1.2,0.07324159741401673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,8,power_law_1.2,0.04081279933452606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,8,power_law_1.2,0.04355199933052063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,32,4,power_law_1.2,0.07395200133323669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,8,power_law_1.2,0.04323840141296387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,8,power_law_1.2,0.04658559858798981
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,32,4,power_law_1.2,0.07565439939498901
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,8,power_law_1.2,0.05070719718933105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,32,4,power_law_1.2,0.07874559760093688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,8,power_law_1.2,0.0526528000831604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,8,power_law_1.2,0.06252800226211548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,32,4,power_law_1.2,0.08273919820785522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,8,power_law_1.2,0.07144960165023803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,8,power_law_1.2,0.08367999792098998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,32,4,power_law_1.2,0.08357120156288148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,8,power_law_1.2,0.11079679727554322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,32,4,power_law_1.2,0.08895360231399536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,8,power_law_1.2,0.14886399507522582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,8,power_law_1.2,0.19633920192718507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,32,4,power_law_1.2,0.1009600043296814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,8,power_law_1.2,0.26791679859161377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,32,4,power_law_1.2,0.09960960149765015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,8,power_law_1.2,0.3379519939422607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,32,4,power_law_1.2,0.10076800584793091
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,8,power_law_1.2,0.4858367919921875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,32,4,power_law_1.2,0.11752959489822387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,8,power_law_1.2,0.7133376121520996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,32,4,power_law_1.2,0.13004159927368164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,8,power_law_1.2,0.7612351894378662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,32,4,power_law_1.2,0.1703935980796814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,32,4,power_law_1.2,0.20785279273986818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,8,power_law_1.2,1.4379008293151856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,32,4,power_law_1.2,0.2579967975616455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,32,4,power_law_1.2,0.32723839282989503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,32,4,power_law_1.2,0.4631296157836914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,8,power_law_1.2,2.984851264953613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,32,4,power_law_1.2,0.5315072059631347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,32,4,power_law_1.2,0.8356736183166504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,32,4,power_law_1.2,1.1705984115600585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,32,4,power_law_1.2,1.2708736419677735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,32,4,power_law_1.2,1.9914815902709961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,32,4,power_law_1.2,3.6980224609375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,16,power_law_1.01,0.04599680006504059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,16,power_law_1.01,0.04472959935665131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,16,power_law_1.01,0.035718399286270144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,16,power_law_1.01,0.038553598523139956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,16,power_law_1.01,0.03914879858493805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,16,power_law_1.01,0.04325119853019714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,16,power_law_1.01,0.04380159974098206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,16,power_law_1.01,0.04428159892559051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,16,power_law_1.01,0.04521600008010864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,16,power_law_1.01,0.04514560103416443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,16,power_law_1.01,0.04643200039863586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,16,power_law_1.01,0.04826239943504333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,16,power_law_1.01,0.04794879853725433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,16,power_law_1.01,0.05258880257606506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,16,power_law_1.01,0.055353599786758426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,16,power_law_1.01,0.06059520244598389
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,16,power_law_1.01,0.06541439890861511
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,16,power_law_1.01,0.07412480115890503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,16,power_law_1.01,0.0889855980873108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,16,power_law_1.01,0.09870719909667969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,16,power_law_1.01,0.10718719959259033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,8,4,power_law_1.01,0.04392960071563721
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,16,power_law_1.01,0.1338495969772339
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,8,4,power_law_1.01,0.03504000008106232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,8,4,power_law_1.01,0.03370240032672882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,16,power_law_1.01,0.16250879764556886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,8,4,power_law_1.01,0.03454079926013946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,8,4,power_law_1.01,0.034835198521614076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,16,power_law_1.01,0.20670719146728517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,8,4,power_law_1.01,0.03581439852714539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,8,4,power_law_1.01,0.041817599534988405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,16,power_law_1.01,0.2438528060913086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,8,4,power_law_1.01,0.0414463996887207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,8,4,power_law_1.01,0.04523519873619079
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,8,4,power_law_1.01,0.04620800018310547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,16,power_law_1.01,0.3379584074020386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,8,4,power_law_1.01,0.046265599131584165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,8,4,power_law_1.01,0.04821760058403015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,8,4,power_law_1.01,0.050963199138641356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,16,power_law_1.01,0.5173503875732421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,8,4,power_law_1.01,0.05018879771232605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,8,4,power_law_1.01,0.060102397203445436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,8,4,power_law_1.01,0.06172800064086914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,8,4,power_law_1.01,0.06960639953613282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,16,power_law_1.01,0.5508927822113037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,8,4,power_law_1.01,0.08321279883384705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,8,4,power_law_1.01,0.07845759987831116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,8,4,power_law_1.01,0.10524159669876099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,8,4,power_law_1.01,0.12472959756851196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,8,4,power_law_1.01,0.14981119632720946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,16,power_law_1.01,0.9765119552612305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,8,4,power_law_1.01,0.19085439443588256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,8,4,power_law_1.01,0.2599040031433105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,8,4,power_law_1.01,0.33603200912475584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,8,4,power_law_1.01,0.6220416069030762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,8,4,power_law_1.01,0.6649024009704589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,16,power_law_1.01,1.6682048797607423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,8,4,power_law_1.01,0.7557824134826661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,8,4,power_law_1.01,1.165664005279541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,8,4,power_law_1.01,2.2166656494140624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,1,power_law_1.01,0.09573119878768921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,2,8,power_law_1.01,0.02993920147418976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,1,power_law_1.01,0.14531840085983277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,2,8,power_law_1.01,0.026150399446487428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,1,power_law_1.01,0.24063360691070557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,2,8,power_law_1.01,0.025574401021003723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,1,power_law_1.01,0.29745919704437257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,2,8,power_law_1.01,0.02563839852809906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,1,power_law_1.01,0.5935232162475585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,2,8,power_law_1.01,0.025727999210357667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,2,8,power_law_1.01,0.028569599986076354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,1,power_law_1.01,0.9080575942993164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,2,8,power_law_1.01,0.02876800000667572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,1,power_law_1.01,1.0612735748291016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,2,8,power_law_1.01,0.03031040132045746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,2,8,power_law_1.01,0.033024001121521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,1,power_law_1.01,1.4321023941040039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,2,8,power_law_1.01,0.032467201352119446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,2,8,power_law_1.01,0.033855998516082765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,1,power_law_1.01,1.5303936004638672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,2,8,power_law_1.01,0.03857280015945434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,2,8,power_law_1.01,0.03898240029811859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,1,power_law_1.01,1.5749183654785157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,2,8,power_law_1.01,0.04162560105323791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,1,power_law_1.01,1.65098876953125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,2,8,power_law_1.01,0.046028798818588255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,2,8,power_law_1.01,0.04751999974250794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,1,power_law_1.01,1.7984447479248047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,2,8,power_law_1.01,0.04899199903011322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,2,8,power_law_1.01,0.05678719878196716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,1,power_law_1.01,1.7711616516113282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,8,power_law_1.01,0.06559360027313232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,8,power_law_1.01,0.07333120107650756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,1,power_law_1.01,1.8508159637451171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,8,power_law_1.01,0.07735679745674133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,8,power_law_1.01,0.08840320110321045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,1,power_law_1.01,1.9114879608154296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,8,power_law_1.01,0.11390080451965331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,1,power_law_1.01,1.964806365966797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,8,power_law_1.01,0.13471360206604005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,8,power_law_1.01,0.1725119948387146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,1,power_law_1.01,2.037887954711914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,8,power_law_1.01,0.25259521007537844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,1,power_law_1.01,2.1805183410644533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,8,power_law_1.01,0.30027520656585693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,1,power_law_1.01,2.3007232666015627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,8,power_law_1.01,0.3691456079483032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,1,power_law_1.01,2.6093311309814453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,8,power_law_1.01,0.5990911960601807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,1,power_law_1.01,2.583647918701172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,8,power_law_1.01,0.9743743896484375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,1,power_law_1.01,2.9935808181762695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,1,power_law_1.01,3.094918441772461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,1,power_law_1.01,3.8955711364746093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,1,power_law_1.01,4.1920318603515625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,1,power_law_1.01,5.325881576538086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,1,power_law_1.01,0.07511039972305297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,1,power_law_1.01,6.853657531738281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,1,power_law_1.01,0.0917631983757019
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,1,power_law_1.01,0.11096960306167603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,1,power_law_1.01,7.875698852539062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,1,power_law_1.01,0.1529536008834839
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,1,power_law_1.01,0.19550080299377443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,1,power_law_1.01,0.3115839958190918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,1,power_law_1.01,10.46951675415039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,1,power_law_1.01,0.33133440017700194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,1,power_law_1.01,0.344320011138916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,1,power_law_1.01,0.3556544065475464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,1,power_law_1.01,0.36903679370880127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,1,power_law_1.01,0.37319040298461914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,1,power_law_1.01,0.38606719970703124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,1,power_law_1.01,0.39327359199523926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,1,power_law_1.01,19.49901428222656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,1,power_law_1.01,0.4083712100982666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,1,power_law_1.01,0.42828798294067383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,1,power_law_1.01,0.4385663986206055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,1,power_law_1.01,0.4708672046661377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,1,power_law_1.01,0.5303679943084717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,1,power_law_1.01,0.5921855926513672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,1,power_law_1.01,0.7176383972167969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,1,power_law_1.01,0.767955207824707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,1,power_law_1.01,0.9692159652709961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,1,power_law_1.01,1.168070411682129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,16,balanced,0.04791999856630961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,16,balanced,0.050106664498647056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,16,balanced,0.051882664362589516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,16,balanced,0.07249600191911061
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,1,power_law_1.01,1.529190444946289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,16,balanced,0.09549867113431294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,16,balanced,0.14851733048756918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,16,balanced,0.1963040033976237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,16,balanced,0.20018132527669272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,16,balanced,0.19885333379109701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,16,balanced,0.1999733249346415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,16,balanced,0.1997386614481608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,16,balanced,0.20363199710845947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,16,balanced,0.2010293404261271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,16,balanced,0.2041119933128357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,16,balanced,0.21079466740290323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,16,balanced,0.2101973295211792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,16,balanced,0.2167253295580546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,16,balanced,0.2222506602605184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,16,balanced,0.22959999243418375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,16,balanced,0.2446933388710022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,16,balanced,0.259770671526591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,16,balanced,0.2909653385480245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,16,balanced,0.3206239938735962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,1,power_law_1.01,1.8850496292114258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,16,balanced,0.37329065799713135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,16,balanced,0.4512853225072225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,16,balanced,0.5536693334579468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,16,balanced,0.7079199949900309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,16,balanced,0.9051413536071777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,16,balanced,1.2203253110249836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,16,balanced,2.3497066497802734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,1,power_law_1.01,2.613209533691406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,1,power_law_1.01,3.3896385192871095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,1,power_law_1.01,4.098982238769532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,1,power_law_1.01,6.282073593139648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,2,4,balanced,0.02532266577084859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,2,4,balanced,0.029311999678611755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,2,4,balanced,0.031328000128269196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,2,4,balanced,0.03121600051720937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,2,4,balanced,0.06043200194835663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,2,4,balanced,0.05291733145713806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,2,4,balanced,0.05356800059477488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,2,4,balanced,0.052613332867622375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,2,4,balanced,0.053226664662361145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,2,4,balanced,0.05425066749254862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,2,4,balanced,0.05177066723505656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,2,4,balanced,0.05260799825191498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,2,4,balanced,0.053674668073654175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,2,4,balanced,0.05889600018660227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,2,4,balanced,0.06425599753856659
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,2,4,balanced,0.06513066589832306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,2,4,balanced,0.06469866633415222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,2,4,balanced,0.07045333087444305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,4,balanced,0.08067733546098073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,4,balanced,0.08994133273760478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,4,balanced,0.09119466940561931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,4,balanced,0.10910933216412862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,4,balanced,0.13037866353988647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,4,balanced,0.1706399917602539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,4,balanced,0.20175999402999878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,4,balanced,0.277402659257253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,4,balanced,0.3423786560694377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,1,power_law_1.01,12.015801239013673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,4,balanced,0.4150826533635457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,4,balanced,0.6255626678466797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,4,balanced,1.1811199982961018
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,32,power_law_1.01,0.09727360010147094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,16,power_law_1.2,0.03410559892654419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,32,power_law_1.01,0.09247360229492188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,16,power_law_1.2,0.030163198709487915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,32,power_law_1.01,0.07923840284347534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,16,power_law_1.2,0.03049600124359131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,32,power_law_1.01,0.06725119948387145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,16,power_law_1.2,0.030079999566078187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,32,power_law_1.01,0.07754240036010743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,16,power_law_1.2,0.03142400085926056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,32,power_law_1.01,0.0833728015422821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,16,power_law_1.2,0.03209599852561951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,32,power_law_1.01,0.08689919710159302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,16,power_law_1.2,0.032576000690460204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,32,power_law_1.01,0.08972799777984619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,16,power_law_1.2,0.03223679959774017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,32,power_law_1.01,0.09340159893035889
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,16,power_law_1.2,0.03370240032672882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,32,power_law_1.01,0.09482880234718323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,16,power_law_1.2,0.0328575998544693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,32,power_law_1.01,0.09578880071640014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,16,power_law_1.2,0.03341439962387085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,16,power_law_1.2,0.034668800234794614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,32,power_law_1.01,0.0958400011062622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,16,power_law_1.2,0.03611519932746887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,32,power_law_1.01,0.09500799775123596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,16,power_law_1.2,0.03687039911746979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,32,power_law_1.01,0.09617279767990113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,16,power_law_1.2,0.03953279852867127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,32,power_law_1.01,0.10584959983825684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,16,power_law_1.2,0.040345600247383116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,32,power_law_1.01,0.10894720554351807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,16,power_law_1.2,0.04589439928531647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,16,power_law_1.2,0.05080320239067078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,32,power_law_1.01,0.11126400232315063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,16,power_law_1.2,0.0551360011100769
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,32,power_law_1.01,0.12305279970169067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,16,power_law_1.2,0.07379840016365051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,32,power_law_1.01,0.12578560113906861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,16,power_law_1.2,0.07281919717788696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,32,power_law_1.01,0.15324800014495848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,16,power_law_1.2,0.09284480214118958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,32,power_law_1.01,0.15994880199432374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,16,power_law_1.2,0.11871360540390015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,32,power_law_1.01,0.1906496047973633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,16,power_law_1.2,0.20791680812835694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,32,power_law_1.01,0.2294464111328125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,16,power_law_1.2,0.22128000259399414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,32,power_law_1.01,0.3115328073501587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,16,power_law_1.2,0.2920959949493408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,32,power_law_1.01,0.3835007905960083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,16,power_law_1.2,0.40128002166748045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,32,power_law_1.01,0.5471168041229248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,16,power_law_1.2,0.47516160011291503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,32,power_law_1.01,0.627558422088623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,16,power_law_1.2,1.018995189666748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,32,power_law_1.01,0.7859903812408447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,32,power_law_1.01,1.1885696411132813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,16,power_law_1.2,1.6978752136230468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,32,power_law_1.01,2.3604095458984373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,2,64,power_law_1.01,0.0247871994972229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,2,64,power_law_1.01,0.024915200471878052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,2,64,power_law_1.01,0.024928000569343568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,8,power_law_1.01,0.05123839974403381
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,2,64,power_law_1.01,0.025228801369667053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,8,power_law_1.01,0.050809597969055174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,2,64,power_law_1.01,0.025036799907684325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,8,power_law_1.01,0.04888960123062134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,2,64,power_law_1.01,0.02632960081100464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,8,power_law_1.01,0.05111680030822754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,2,64,power_law_1.01,0.02725760042667389
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,8,power_law_1.01,0.05496960282325745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,2,64,power_law_1.01,0.026719999313354493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,8,power_law_1.01,0.06001920104026794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,2,64,power_law_1.01,0.030112001299858093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,8,power_law_1.01,0.07048320174217224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,2,64,power_law_1.01,0.029817599058151244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,2,64,power_law_1.01,0.03150080144405365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,8,power_law_1.01,0.07150080204010009
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,2,64,power_law_1.01,0.03079040050506592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,8,power_law_1.01,0.07269759774208069
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,2,64,power_law_1.01,0.031046399474143983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,8,power_law_1.01,0.07426559925079346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,2,64,power_law_1.01,0.030822399258613586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,8,power_law_1.01,0.07420160174369812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,2,64,power_law_1.01,0.03332479894161224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,8,power_law_1.01,0.07379840016365051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,2,64,power_law_1.01,0.033855998516082765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,8,power_law_1.01,0.07841920256614685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,2,64,power_law_1.01,0.035596799850463864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,8,power_law_1.01,0.08379520177841186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,2,64,power_law_1.01,0.036556801199913024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,16,power_law_1.01,0.053958398103713986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,16,power_law_1.01,0.04933120012283325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,64,power_law_1.01,0.04119040071964264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,8,power_law_1.01,0.0890496015548706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,16,power_law_1.01,0.04316799938678741
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,64,power_law_1.01,0.050169599056243894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,8,power_law_1.01,0.09270399808883667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,16,power_law_1.01,0.044486400485038755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,64,power_law_1.01,0.04692479968070984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,8,power_law_1.01,0.10298880338668823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,16,power_law_1.01,0.044684800505638125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,64,power_law_1.01,0.05860480070114136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,16,power_law_1.01,0.04821119904518127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,8,power_law_1.01,0.10896639823913574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,16,power_law_1.01,0.05006719827651977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,64,power_law_1.01,0.06232960224151611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,8,power_law_1.01,0.1229248046875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,16,power_law_1.01,0.05141119956970215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,64,power_law_1.01,0.07971839904785157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,16,power_law_1.01,0.05188480019569397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,8,power_law_1.01,0.1394752025604248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,16,power_law_1.01,0.05167359709739685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,64,power_law_1.01,0.08543360233306885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,8,power_law_1.01,0.17550719976425172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,16,power_law_1.01,0.05272960066795349
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,16,power_law_1.01,0.05676800012588501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,64,power_law_1.01,0.12188160419464111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,8,power_law_1.01,0.21047680377960204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,16,power_law_1.01,0.059987199306488034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,16,power_law_1.01,0.06403840184211732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,8,power_law_1.01,0.2556351900100708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,64,power_law_1.01,0.13564800024032592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,16,power_law_1.01,0.06690559983253479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,8,power_law_1.01,0.3333631992340088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,16,power_law_1.01,0.06943359971046448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,64,power_law_1.01,0.16812800168991088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,16,power_law_1.01,0.07588480114936828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,8,power_law_1.01,0.4220479965209961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,16,power_law_1.01,0.08392320275306701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,16,power_law_1.01,0.0967743992805481
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,64,power_law_1.01,0.3028991937637329
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,8,power_law_1.01,0.5803455829620361
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,16,power_law_1.01,0.11574399471282959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,16,power_law_1.01,0.14428160190582276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,8,power_law_1.01,0.717414379119873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,16,power_law_1.01,0.19572479724884034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,64,power_law_1.01,0.6943679809570312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,16,power_law_1.01,0.23025920391082763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,8,power_law_1.01,1.0362496376037598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,16,power_law_1.01,0.3304960012435913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,16,power_law_1.01,0.3793728113174438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,8,power_law_1.01,1.3652480125427247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,16,power_law_1.01,0.5592576026916504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,16,power_law_1.01,0.8065919876098633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,16,power_law_1.01,0.9107392311096192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,8,power_law_1.01,2.7423423767089843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,16,power_law_1.01,1.4588800430297852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,16,power_law_1.01,2.6980096817016603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,8,power_law_1.2,0.15761280059814453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,8,power_law_1.2,0.11447679996490479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,8,power_law_1.2,0.106278395652771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,8,power_law_1.2,0.13776639699935914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,8,power_law_1.2,0.17509119510650634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,8,power_law_1.2,0.25477120876312254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,8,power_law_1.2,0.27441279888153075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,8,2,balanced,0.03177600105603536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,8,2,balanced,0.03235200047492981
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,8,2,balanced,0.035904000202814736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,8,2,balanced,0.036464000741640724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,8,2,balanced,0.036687999963760376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,8,2,balanced,0.03586133321126302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,8,2,balanced,0.03585066646337509
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,8,2,balanced,0.03800000001986822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,8,2,balanced,0.037802666425704956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,8,2,balanced,0.03788266579310099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,8,2,balanced,0.04195733368396759
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,8,2,balanced,0.04180799921353658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,8,2,balanced,0.04196266829967499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,8,2,balanced,0.04446933170159658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,8,2,balanced,0.05671999851862589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,8,2,balanced,0.058143998185793556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,8,2,balanced,0.061199997862180076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,8,2,balanced,0.07542933523654938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,8,2,balanced,0.07944533228874207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,8,2,balanced,0.09661333759625752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,8,2,balanced,0.11153599619865417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,8,power_law_1.2,0.293721604347229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,8,2,balanced,0.1426346699396769
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,8,2,balanced,0.16994667053222656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,8,2,balanced,0.23473066091537476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,8,2,balanced,0.2893386681874593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,8,2,balanced,0.4094346761703491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,8,2,balanced,0.5271253188451132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,8,2,balanced,0.643178661664327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,8,2,balanced,0.9862293402353922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,8,power_law_1.2,0.2813184022903442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,8,2,balanced,1.900752067565918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,8,power_law_1.2,0.2897599935531616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,8,power_law_1.2,0.30915839672088624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,8,power_law_1.2,0.31254398822784424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,8,power_law_1.2,0.31779839992523196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,8,power_law_1.2,0.3374079942703247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,8,power_law_1.2,0.34209280014038085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,8,power_law_1.2,0.36382079124450684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,8,power_law_1.2,0.3566783905029297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,8,power_law_1.2,0.39964799880981444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,8,power_law_1.2,0.46350722312927245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,8,power_law_1.2,0.5560704231262207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,32,balanced,0.03384000062942505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,32,balanced,0.03367999941110611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,32,balanced,0.03229333211978277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,32,balanced,0.034074666599432625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,32,balanced,0.03658133248488108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,32,balanced,0.035904000202814736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,32,balanced,0.03775466730197271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,32,balanced,0.03601066768169403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,32,balanced,0.03606399893760681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,8,power_law_1.2,0.6374207973480225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,32,balanced,0.037989333271980286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,32,balanced,0.03811199963092804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,32,balanced,0.03965866565704346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,32,balanced,0.038176000118255615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,32,balanced,0.040448000033696495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,32,balanced,0.04574400186538696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,32,balanced,0.04404800136884054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,32,balanced,0.04808000226815542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,32,balanced,0.05056533217430115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,32,balanced,0.05592533449331919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,32,balanced,0.07300800085067749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,32,balanced,0.07935466865698497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,32,balanced,0.09896533687909444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,32,balanced,0.11155200004577637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,32,balanced,0.1405173341433207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,32,balanced,0.17121066649754843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,8,power_law_1.2,0.8023743629455566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,32,balanced,0.23007466395696005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,32,balanced,0.28467732667922974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,32,balanced,0.3415199915568034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,32,balanced,0.5190666516621908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,32,balanced,0.9781760374704996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,8,power_law_1.2,0.7376768112182617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,8,power_law_1.2,0.9725760459899903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,8,power_law_1.2,1.4448448181152345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,8,power_law_1.2,1.7609920501708984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,8,power_law_1.2,2.194534492492676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,8,power_law_1.2,2.6304832458496095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,8,power_law_1.2,4.258425521850586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,64,balanced,0.03230933348337809
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,64,balanced,0.03202133377393087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,64,balanced,0.03164266546567281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,64,balanced,0.03001066545645396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,64,balanced,0.032773333291212715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,64,balanced,0.033717334270477295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,64,balanced,0.03355200091997782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,64,balanced,0.03385066737731298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,64,balanced,0.03387200087308884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,64,balanced,0.03393599887688955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,64,balanced,0.035504000882307686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,64,balanced,0.033786666889985405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,64,balanced,0.03580799947182337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,64,balanced,0.037477334340413414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,64,balanced,0.03994666785001755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,64,balanced,0.03995733211437861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,64,balanced,0.03994133323431015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,64,balanced,0.043765331308046974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,64,balanced,0.043920000394185386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,64,balanced,0.04757866760094961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,64,balanced,0.05009066561857859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,64,balanced,0.06238399942715963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,64,balanced,0.07870933413505554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,64,balanced,0.11248000462849934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,64,balanced,0.14171733458836874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,64,balanced,0.1930453379948934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,64,balanced,0.231989324092865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,64,balanced,0.28387200832366943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,64,balanced,0.4118719895680745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,64,balanced,0.77182936668396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,8,power_law_1.2,7.839180755615234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,32,8,balanced,0.03148266673088074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,32,8,balanced,0.031354665756225586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,32,8,balanced,0.03137599925200144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,32,8,balanced,0.029711998999118805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,32,8,balanced,0.029578665892283123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,32,8,balanced,0.03147733211517334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,32,8,balanced,0.029711998999118805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,32,8,balanced,0.031162666777769726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,32,8,balanced,0.03183999905983607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,4,power_law_1.2,0.09648000001907349
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,32,8,balanced,0.03068266560633977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,32,8,balanced,0.031290667752424874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,32,8,balanced,0.03160000095764796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,32,8,balanced,0.031290667752424874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,32,8,balanced,0.03218133250872294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,32,8,balanced,0.03365866591533025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,32,8,balanced,0.03555733213822047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,32,8,balanced,0.03737066686153412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,32,8,balanced,0.03734933336575826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,32,8,balanced,0.039503999054431915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,32,8,balanced,0.04095466683308283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,32,8,balanced,0.04411733150482178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,32,8,balanced,0.049728001157442726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,32,8,balanced,0.051967998345692955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,32,8,balanced,0.07678933441638947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,32,8,balanced,0.0853760043780009
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,32,8,balanced,0.11314666271209717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,8,1,power_law_1.2,0.04320639967918396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,32,8,balanced,0.12590400377909342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,32,8,balanced,0.15891200304031372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,32,8,balanced,0.20723734299341837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,32,8,balanced,0.36342934767405194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,8,1,power_law_1.2,0.04447360038757324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,4,power_law_1.2,0.14909440279006958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,8,1,power_law_1.2,0.04989440143108368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,8,1,power_law_1.2,0.0643775999546051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,8,1,power_law_1.2,0.06442880034446716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,8,1,power_law_1.2,0.06665599942207337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,4,power_law_1.2,0.15217920541763305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,8,1,power_law_1.2,0.06638720035552978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,8,1,power_law_1.2,0.06800000071525573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,8,1,power_law_1.2,0.06700800061225891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,8,1,power_law_1.2,0.06739839911460876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,8,1,power_law_1.2,0.07163519859313965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,4,power_law_1.2,0.12480640411376953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,8,1,power_law_1.2,0.07336320281028748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,8,1,power_law_1.2,0.07560960054397584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,8,1,power_law_1.2,0.07869439721107482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,8,1,power_law_1.2,0.0866047978401184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,4,power_law_1.2,0.1732416033744812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,8,1,power_law_1.2,0.08902400135993957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,8,1,power_law_1.2,0.09782400131225585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,8,1,power_law_1.2,0.1071679949760437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,8,1,power_law_1.2,0.12409600019454955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,4,power_law_1.2,0.24413440227508545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,8,1,power_law_1.2,0.15096319913864137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,8,1,power_law_1.2,0.17412480115890502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,8,1,power_law_1.2,0.2259455919265747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,8,1,power_law_1.2,0.28016641139984133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,4,power_law_1.2,0.2928960084915161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,8,1,power_law_1.2,0.38073599338531494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,8,1,power_law_1.2,0.4800704002380371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,8,1,power_law_1.2,0.6829184055328369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,4,power_law_1.2,0.40369281768798826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,8,1,power_law_1.2,0.8848832130432129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,8,1,power_law_1.2,1.0825152397155762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,4,power_law_1.2,0.4191743850708008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,8,1,power_law_1.2,1.6757568359375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,4,power_law_1.2,0.42673277854919434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,8,4,power_law_1.2,0.054579198360443115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,8,1,power_law_1.2,3.2657470703125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,4,power_law_1.2,0.45172481536865233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,8,4,power_law_1.2,0.05162879824638367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,8,4,power_law_1.2,0.05614719986915588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,4,power_law_1.2,0.4677567958831787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,8,4,power_law_1.2,0.06096000075340271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,8,4,power_law_1.2,0.06537600159645081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,4,power_law_1.2,0.49660801887512207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,8,4,power_law_1.2,0.07598080039024353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,8,4,power_law_1.2,0.0864960014820099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,4,power_law_1.2,0.5014527797698974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,8,4,power_law_1.2,0.08623999953269959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,8,4,power_law_1.2,0.08920959830284118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,4,power_law_1.2,0.5050559997558594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,8,4,power_law_1.2,0.08839679956436157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,8,4,power_law_1.2,0.09142400026321411
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,4,power_law_1.2,0.5604800224304199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,8,4,power_law_1.2,0.09516800045967103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,8,4,power_law_1.2,0.09798399806022644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,8,4,power_law_1.2,0.10377600193023681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,4,power_law_1.2,0.6002111911773682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,8,4,power_law_1.2,0.11242239475250244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,8,4,power_law_1.2,0.11756800413131714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,4,power_law_1.2,0.6032192230224609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,8,4,power_law_1.2,0.12156800031661988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,4,power_law_1.2,0.6907072067260742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,8,4,power_law_1.2,0.13980159759521485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,64,balanced,0.048485333720842995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,64,balanced,0.03843733419974645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,64,balanced,0.029839999973773956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,64,balanced,0.0341333324710528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,64,balanced,0.03379733363787333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,64,balanced,0.033941333492596946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,64,balanced,0.033802665770053864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,64,balanced,0.03404266635576884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,8,4,power_law_1.2,0.15457919836044312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,64,balanced,0.03190399954716364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,64,balanced,0.03381866713364919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,64,balanced,0.03429866582155228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,64,balanced,0.03392533212900162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,64,balanced,0.033770665526390076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,64,balanced,0.03596800069014231
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,64,balanced,0.037791999677817024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,64,balanced,0.038293334345022835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,64,balanced,0.039359999199708305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,8,4,power_law_1.2,0.18389760255813598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,4,power_law_1.2,0.8142271995544433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,64,balanced,0.04166933397452036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,64,balanced,0.0410453329483668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,64,balanced,0.04563733438650767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,64,balanced,0.04326933125654856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,64,balanced,0.05213866631189982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,64,balanced,0.05166399975617727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,64,balanced,0.0677706648906072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,8,4,power_law_1.2,0.22748799324035646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,64,balanced,0.06333866715431213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,64,balanced,0.08178666730721791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,64,balanced,0.09284266829490662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,1,64,balanced,0.10114666819572449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,1,64,balanced,0.13455466429392496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,1,64,balanced,0.20968000094095865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,4,power_law_1.2,0.8035391807556153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,8,4,power_law_1.2,0.29023358821868894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,8,4,power_law_1.2,0.36814720630645753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,4,power_law_1.2,0.9217023849487305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,8,4,power_law_1.2,0.47251200675964355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,8,4,power_law_1.2,0.6554368019104004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,4,power_law_1.2,1.0746879577636719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,8,4,power_law_1.2,0.9608575820922851
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,4,power_law_1.2,1.34204158782959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,8,4,power_law_1.2,1.1358464241027832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,4,power_law_1.2,1.4254655838012695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,8,4,power_law_1.2,1.4142720222473144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,4,power_law_1.2,1.7780351638793945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,8,4,power_law_1.2,2.409254455566406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,4,power_law_1.2,2.276032066345215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,8,4,power_law_1.2,4.648652648925781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,64,power_law_1.2,0.03596160113811493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,4,power_law_1.2,2.5720703125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,64,power_law_1.2,0.030028799176216127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,64,power_law_1.2,0.029471999406814574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,64,power_law_1.2,0.0304639995098114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,4,power_law_1.2,3.888780975341797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,64,power_law_1.2,0.030534398555755616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,64,power_law_1.2,0.03198080062866211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,64,power_law_1.2,0.0313728004693985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,64,power_law_1.2,0.031615999341011045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,64,power_law_1.2,0.03139199912548065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,4,power_law_1.2,7.18853759765625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,64,power_law_1.2,0.03249920010566711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,64,power_law_1.2,0.03304319977760315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,64,power_law_1.2,0.03299199938774109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,64,power_law_1.2,0.03335680067539215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,64,power_law_1.2,0.034118399024009705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,16,power_law_1.2,0.08570240139961242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,64,power_law_1.2,0.03790720105171204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,16,power_law_1.2,0.12577919960021972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,64,power_law_1.2,0.04056960046291351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,16,power_law_1.2,0.05985919833183288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,64,power_law_1.2,0.04056319892406464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,64,power_law_1.2,0.04366720020771027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,16,power_law_1.2,0.06839680075645446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,64,power_law_1.2,0.04736000001430511
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,16,power_law_1.2,0.07900800108909607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,64,power_law_1.2,0.05614079833030701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,16,power_law_1.2,0.08614400029182434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,64,power_law_1.2,0.06432639956474304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,16,power_law_1.2,0.10414079427719117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,64,power_law_1.2,0.0782144010066986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,16,power_law_1.2,0.09916800260543823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,64,power_law_1.2,0.08549759984016418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,4,balanced,0.05247466762860616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,4,balanced,0.07623466849327087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,4,balanced,0.09715732932090759
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,4,balanced,0.1441386640071869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,16,power_law_1.2,0.10160640478134156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,64,power_law_1.2,0.13212800025939941
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,4,balanced,0.24374399582544962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,4,balanced,0.4397439956665039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,4,balanced,0.4435360034306844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,4,balanced,0.4433759848276774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,4,balanced,0.4434826771418254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,4,balanced,0.4432640075683594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,4,balanced,0.4506346782048543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,4,balanced,0.45191999276479083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,4,balanced,0.4548693497975667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,16,power_law_1.2,0.10401279926300049
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,4,balanced,0.4581386645634969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,64,power_law_1.2,0.17820800542831422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,4,balanced,0.46718935171763104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,4,balanced,0.47003201643625897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,4,balanced,0.4774133364359538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,4,balanced,0.500437339146932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,4,balanced,0.5136959950129191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,4,balanced,0.5466613372166952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,4,balanced,0.5893119970957438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,16,power_law_1.2,0.107315194606781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,4,balanced,0.6563626527786255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,64,power_law_1.2,0.254367995262146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,4,balanced,0.7077866395314535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,4,balanced,0.9250826835632324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,4,balanced,0.9916906356811523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,16,power_law_1.2,0.111353600025177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,4,balanced,1.3698560396830242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,4,balanced,1.6205226580301921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,64,power_law_1.2,0.3575360059738159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,4,balanced,2.169381300608317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,16,power_law_1.2,0.11006720066070556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,4,balanced,3.0288801193237305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,16,power_law_1.2,0.11411839723587036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,64,power_law_1.2,0.39731199741363527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,4,balanced,5.871386845906575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,16,power_law_1.2,0.12353279590606689
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,16,power_law_1.2,0.12466559410095215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,64,power_law_1.2,0.748044776916504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,16,power_law_1.2,0.13504639863967896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,16,power_law_1.2,0.14421119689941406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,16,power_law_1.2,0.1629696011543274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,64,power_law_1.2,1.62608642578125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,16,power_law_1.2,0.18997759819030763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,16,power_law_1.2,0.21990399360656737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,16,power_law_1.2,0.28830718994140625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,16,power_law_1.2,0.3256511926651001
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,16,power_law_1.2,0.4055935859680176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,16,power_law_1.2,0.5018176078796387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,1,power_law_1.01,0.03529599905014038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,1,power_law_1.01,0.0427839994430542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,16,power_law_1.2,0.8729344367980957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,1,power_law_1.01,0.04926080107688904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,1,power_law_1.01,0.0658623993396759
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,1,power_law_1.01,0.07694079875946044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,16,power_law_1.2,0.8999872207641602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,1,power_law_1.01,0.10774400234222412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,1,power_law_1.01,0.11650559902191163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,1,power_law_1.01,0.11749759912490845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,16,power_law_1.2,1.1409407615661622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,1,power_law_1.01,0.12295039892196655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,1,power_law_1.01,0.1242751955986023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,1,power_law_1.01,0.1320896029472351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,1,power_law_1.01,0.13498879671096803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,16,power_law_1.2,1.9950464248657227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,1,power_law_1.01,0.1384384036064148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,1,power_law_1.01,0.1459712028503418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,1,power_law_1.01,0.15394560098648072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,1,power_law_1.01,0.16224000453948975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,1,power_law_1.01,0.17011200189590453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,1,power_law_1.01,0.19574400186538696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,16,power_law_1.2,3.656441497802734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,1,power_law_1.01,0.21540479660034179
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,1,power_law_1.01,0.26847999095916747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,1,power_law_1.01,0.307475209236145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,1,power_law_1.01,0.40991997718811035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,1,power_law_1.01,0.5035967826843262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,1,power_law_1.01,0.6933184146881104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,1,power_law_1.01,0.8833791732788085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,1,power_law_1.01,1.2664064407348632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,1,power_law_1.01,1.6677888870239257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,1,power_law_1.01,2.0554431915283202
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,1,power_law_1.01,3.1261695861816405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,4,8,balanced,0.04333333174387614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,4,8,balanced,0.043925335009892784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,4,8,balanced,0.04586133360862732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,4,8,balanced,0.04433600107828776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,4,8,balanced,0.046021332343419395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,4,8,balanced,0.04572799801826477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,4,8,balanced,0.045738667249679565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,4,8,balanced,0.036117332677046456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,4,8,balanced,0.03579733272393545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,4,8,balanced,0.03737066686153412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,4,8,balanced,0.04008533308903376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,4,8,balanced,0.03956266740957896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,4,8,balanced,0.039706667264302574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,4,8,balanced,0.04162666698296865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,4,8,balanced,0.04961066444714864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,4,8,balanced,0.051669334371884666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,4,8,balanced,0.053082664807637535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,4,8,balanced,0.06438933312892914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,4,8,balanced,0.05996266504128774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,4,8,balanced,0.07582933207352956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,4,8,balanced,0.07674666742483775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,4,8,balanced,0.09609066446622212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,4,8,balanced,0.10162132978439331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,4,8,balanced,0.13410666584968567
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,4,8,balanced,0.152319997549057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,4,8,balanced,0.21429866552352905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,4,8,balanced,0.2585866649945577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,4,8,balanced,0.3208000063896179
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,4,8,balanced,0.46347200870513916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,4,8,balanced,0.8834986686706543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,1,power_law_1.01,6.087014389038086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,32,4,power_law_1.01,0.04602240025997162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,32,4,power_law_1.01,0.04831359982490539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,32,4,power_law_1.01,0.04585599899291992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,32,4,power_law_1.01,0.05267840027809143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,32,4,power_law_1.01,0.055199998617172244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,32,4,power_law_1.01,0.06409599781036376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,32,4,power_law_1.01,0.06574079990386963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,32,4,power_law_1.01,0.07004160284996033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,32,4,power_law_1.01,0.07109760046005249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,32,4,power_law_1.01,0.07416319847106934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,32,4,power_law_1.01,0.07477120161056519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,32,4,power_law_1.01,0.0791487991809845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,32,4,power_law_1.01,0.08227199912071229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,32,4,power_law_1.01,0.08542079925537109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,32,4,power_law_1.01,0.09590399861335755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,32,4,power_law_1.01,0.1002303957939148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,32,4,power_law_1.01,0.09887359738349914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,32,4,power_law_1.01,0.11411199569702149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,32,4,power_law_1.01,0.12361600399017333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,32,4,power_law_1.01,0.15132800340652466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,32,4,power_law_1.01,0.19975680112838745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,32,4,power_law_1.01,0.23985280990600585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,32,4,power_law_1.01,0.30151679515838625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,32,4,power_law_1.01,0.4173888206481934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,32,4,power_law_1.01,0.5202112197875977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,32,4,power_law_1.01,0.7493696212768555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,32,4,power_law_1.01,0.957094383239746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,32,4,power_law_1.01,1.1625791549682618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,32,4,power_law_1.01,1.711084747314453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,32,4,power_law_1.01,3.506560134887695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,4,balanced,0.0518506666024526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,4,balanced,0.055871998270352684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,4,balanced,0.06887466708819072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,4,balanced,0.09566932916641235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,4,balanced,0.12957866986592612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,4,balanced,0.15997333327929178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,4,balanced,0.16221867005030313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,4,balanced,0.16275200247764587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,4,balanced,0.162581334511439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,4,balanced,0.16274666786193848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,4,balanced,0.1657866636912028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,4,balanced,0.16782933473587036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,4,balanced,0.1720906694730123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,4,balanced,0.17585599422454834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,4,balanced,0.18220800161361694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,4,balanced,0.18598399559656778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,4,balanced,0.19144533077875772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,4,balanced,0.2191200057665507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,4,balanced,0.2286506692568461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,4,balanced,0.29314132531483966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,4,balanced,0.31940267483393353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,4,balanced,0.46005864938100177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,4,balanced,0.4705493450164795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,4,balanced,0.6731306711832682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,4,balanced,0.7602986494700114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,4,balanced,1.109946648279826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,4,balanced,1.372048060099284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,8,4,balanced,1.7486613591512044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,8,4,balanced,2.581914742787679
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,8,4,balanced,5.2093706130981445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,8,balanced,0.05012799799442291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,8,balanced,0.05223466455936432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,8,balanced,0.07267733414967854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,8,balanced,0.09915733337402344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,8,balanced,0.148117333650589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,8,balanced,0.24733867247899374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,8,balanced,0.3474613428115845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,8,balanced,0.3437013228734334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,8,balanced,0.34614400068918866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,8,balanced,0.3480106592178345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,8,balanced,0.34749333063761395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,8,balanced,0.3537280162175496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,8,balanced,0.35423465569814044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,8,balanced,0.3556906779607137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,8,balanced,0.36658668518066406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,8,balanced,0.36851731936136883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,8,balanced,0.3763786554336548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,8,balanced,0.38997864723205566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,8,balanced,0.40436267852783203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,8,balanced,0.4315679868062337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,8,balanced,0.4566506544748942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,8,balanced,0.5090186595916748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,8,balanced,0.5665119886398315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,8,balanced,0.6710879802703857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,8,balanced,0.8020373185475668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,8,balanced,0.9998986721038818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,8,balanced,1.290981372197469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,8,balanced,1.5780426661173503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,8,balanced,2.2972052892049155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,8,balanced,4.3872480392456055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,1,power_law_1.01,0.09267839789390564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,1,power_law_1.01,0.09555839896202087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,1,power_law_1.01,0.10524159669876099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,1,power_law_1.01,0.12709759473800658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,1,power_law_1.01,0.16819839477539061
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,1,power_law_1.01,0.20994560718536376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,1,power_law_1.01,0.2290816068649292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,1,power_law_1.01,0.2690943956375122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,1,power_law_1.01,0.2902400016784668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,1,power_law_1.01,0.29380478858947756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,1,power_law_1.01,0.3073472023010254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,1,power_law_1.01,0.3122944116592407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,1,power_law_1.01,0.3198784112930298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,1,power_law_1.01,0.33736319541931153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,1,power_law_1.01,0.35214080810546877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,1,power_law_1.01,0.36119680404663085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,1,power_law_1.01,0.3841279983520508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,1,power_law_1.01,0.4326655864715576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,1,power_law_1.01,0.4726975917816162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,1,power_law_1.01,0.5705984115600586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,1,power_law_1.01,0.6609024047851563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,1,power_law_1.01,0.8500991821289062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,1,power_law_1.01,1.0283583641052245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,1,power_law_1.01,1.405344009399414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,1,power_law_1.01,1.7700607299804687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,1,power_law_1.01,2.5123903274536135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,1,power_law_1.01,3.249740982055664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,1,power_law_1.01,3.998553466796875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,1,power_law_1.01,6.190848159790039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,1,power_law_1.01,11.99617919921875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,32,1,balanced,0.050666665037473045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,32,1,balanced,0.05470933516820272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,32,1,balanced,0.062447999914487205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,32,1,balanced,0.0918346643447876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,32,1,balanced,0.13371200362841287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,32,1,balanced,0.15428800384203592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,32,1,balanced,0.1571466624736786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,32,1,balanced,0.15968533356984457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,32,1,balanced,0.16076800227165222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,32,1,balanced,0.1629706621170044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,32,1,balanced,0.16569599509239197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,32,1,balanced,0.16976000865300497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,32,1,balanced,0.17072532574335733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,32,1,balanced,0.17734932899475098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,32,1,balanced,0.18617600202560425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,32,1,balanced,0.19097065925598145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,32,1,balanced,0.20171199242273966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,32,1,balanced,0.23983466625213623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,32,1,balanced,0.2621386647224426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,32,1,balanced,0.3699520031611125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,32,1,balanced,0.39692266782124835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,32,1,balanced,0.5849920113881429
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,32,1,balanced,0.6624000072479248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,32,1,balanced,0.981216033299764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,32,1,balanced,1.2037333647410076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,32,1,balanced,1.8260960578918457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,32,1,balanced,2.237936019897461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,32,1,balanced,2.6555360158284507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,32,1,balanced,4.2026933034261065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,32,1,balanced,8.102879842122396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,8,power_law_1.01,0.04901759922504425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,8,power_law_1.01,0.06903039813041686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,8,power_law_1.01,0.04991999864578247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,8,power_law_1.01,0.05687680244445801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,8,power_law_1.01,0.06272640228271484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,8,power_law_1.01,0.07519999742507935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,8,power_law_1.01,0.0790719985961914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,8,power_law_1.01,0.07858560085296631
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,8,power_law_1.01,0.08160640001296997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,8,power_law_1.01,0.0832256019115448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,8,power_law_1.01,0.08540800213813782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,8,power_law_1.01,0.0887935996055603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,8,power_law_1.01,0.08892800211906433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,8,power_law_1.01,0.09485440254211426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,8,power_law_1.01,0.10270719528198242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,8,power_law_1.01,0.10640000104904175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,8,power_law_1.01,0.10855679512023926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,8,power_law_1.01,0.12119040489196778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,8,power_law_1.01,0.13418879508972167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,8,power_law_1.01,0.16810879707336426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,8,power_law_1.01,0.19826560020446776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,8,power_law_1.01,0.2636159896850586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,8,power_law_1.01,0.3314687967300415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,8,power_law_1.01,0.4573184013366699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,8,power_law_1.01,0.5086016178131103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,8,power_law_1.01,0.7591807842254639
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,8,power_law_1.01,0.9829631805419922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,8,power_law_1.01,1.1951807975769042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,8,power_law_1.01,1.8053312301635742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,8,power_law_1.01,3.5779518127441405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,8,power_law_1.01,0.046214398741722104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,8,power_law_1.01,0.04976640045642853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,8,power_law_1.01,0.058233600854873654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,8,power_law_1.01,0.04912000000476837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,8,power_law_1.01,0.05305600166320801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,8,power_law_1.01,0.0592960000038147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,8,power_law_1.01,0.06195840239524841
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,8,power_law_1.01,0.06768640279769897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,8,power_law_1.01,0.0679040014743805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,8,power_law_1.01,0.07093120217323304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,8,power_law_1.01,0.07444480061531067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,8,power_law_1.01,0.08023679852485657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,8,power_law_1.01,0.08296959996223449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,8,power_law_1.01,0.08866559863090515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,8,power_law_1.01,0.09681919813156128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,8,power_law_1.01,0.09863680005073547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,8,power_law_1.01,0.10662399530410767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,8,power_law_1.01,0.12184319496154786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,8,power_law_1.01,0.1339840054512024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,8,power_law_1.01,0.16520960330963136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,8,power_law_1.01,0.19711359739303588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,8,power_law_1.01,0.262604808807373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,8,power_law_1.01,0.3336512088775635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,8,power_law_1.01,0.44615678787231444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,8,power_law_1.01,0.5574848175048828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,8,power_law_1.01,0.7977727890014649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,8,power_law_1.01,1.0148287773132325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,8,power_law_1.01,1.2717696189880372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,8,power_law_1.01,1.9548416137695312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,8,power_law_1.01,3.8420543670654297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,2,power_law_1.2,0.06394240260124207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,2,power_law_1.2,0.08962560296058655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,2,power_law_1.2,0.08331519961357117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,2,power_law_1.2,0.10190720558166504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,2,power_law_1.2,0.1336127996444702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,2,power_law_1.2,0.15911680459976196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,2,power_law_1.2,0.23316481113433837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,2,power_law_1.2,0.23960320949554442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,2,power_law_1.2,0.25511679649353025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,2,power_law_1.2,0.2598207950592041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,2,power_law_1.2,0.2674815893173218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,2,power_law_1.2,0.2814719915390015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,2,power_law_1.2,0.2879040002822876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,2,power_law_1.2,0.30524799823760984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,2,power_law_1.2,0.3241472005844116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,2,power_law_1.2,0.3319360017776489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,2,power_law_1.2,0.34792320728302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,2,power_law_1.2,0.39423360824584963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,2,power_law_1.2,0.4329792022705078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,2,power_law_1.2,0.5143167972564697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,2,power_law_1.2,0.5600063800811768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,2,power_law_1.2,0.6794047832489014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,2,power_law_1.2,0.7683775901794434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,2,power_law_1.2,1.0310015678405762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,2,power_law_1.2,1.3234880447387696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,2,power_law_1.2,1.8010623931884766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,4,power_law_1.2,0.03563520014286041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,4,power_law_1.2,0.029606398940086365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,4,power_law_1.2,0.029542401432991028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,2,power_law_1.2,2.270272064208984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,4,power_law_1.2,0.031231999397277832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,4,power_law_1.2,0.03308799862861633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,4,power_law_1.2,0.035545599460601804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,4,power_law_1.2,0.036025598645210266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,2,power_law_1.2,2.9427648544311524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,4,power_law_1.2,0.03653120100498199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,4,power_law_1.2,0.037561601400375365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,4,power_law_1.2,0.03782399892807007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,4,power_law_1.2,0.03989759981632233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,4,power_law_1.2,0.041606399416923526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,2,power_law_1.2,4.258092880249023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,4,power_law_1.2,0.04346239864826203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,4,power_law_1.2,0.04535039961338043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,4,power_law_1.2,0.049747198820114136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,4,power_law_1.2,0.053504002094268796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,4,power_law_1.2,0.05482239723205566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,4,power_law_1.2,0.06062080264091492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,4,power_law_1.2,0.07336959838867188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,4,power_law_1.2,0.08502399921417236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,2,power_law_1.2,7.9603523254394535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,4,power_law_1.2,0.09722880125045777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,4,power_law_1.2,0.12754559516906738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,4,power_law_1.2,0.17368320226669312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,4,power_law_1.2,0.25374081134796145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,4,power_law_1.2,0.30804479122161865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,4,power_law_1.2,0.4046720027923584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,4,power_law_1.2,0.6132927894592285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,4,power_law_1.2,0.7843711853027344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,4,power_law_1.2,1.3202879905700684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,4,power_law_1.2,2.6735040664672853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,16,1,power_law_1.01,0.030015999078750612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,16,1,power_law_1.01,0.031052801012992858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,16,1,power_law_1.01,0.031225600838661195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,16,1,power_law_1.01,0.03547520041465759
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,16,1,power_law_1.01,0.03523840010166168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,16,1,power_law_1.01,0.03640320003032684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,16,1,power_law_1.01,0.041356799006462094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,16,1,power_law_1.01,0.03842560052871704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,16,1,power_law_1.01,0.03930239975452423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,16,1,power_law_1.01,0.04129279851913452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,16,1,power_law_1.01,0.04405120015144348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,16,1,power_law_1.01,0.046854400634765626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,16,1,power_law_1.01,0.049158400297164916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,16,1,power_law_1.01,0.053881597518920896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,16,1,power_law_1.01,0.06234880089759827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,16,1,power_law_1.01,0.06753280162811279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,16,1,power_law_1.01,0.06886399984359741
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,16,1,power_law_1.01,0.085452800989151
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,16,1,power_law_1.01,0.09045119881629944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,16,1,power_law_1.01,0.11666560173034668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,16,1,power_law_1.01,0.13173120021820067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,16,1,power_law_1.01,0.1704319953918457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,16,1,power_law_1.01,0.21249918937683104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,16,1,power_law_1.01,0.2955199956893921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,16,1,power_law_1.01,0.38161919116973875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,16,1,power_law_1.01,0.5467455863952637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,16,1,power_law_1.01,0.7079360008239746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,16,1,power_law_1.01,0.8648384094238282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,16,1,power_law_1.01,1.3462719917297363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,16,1,power_law_1.01,2.628166389465332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,8,2,power_law_1.01,0.0383296012878418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,8,2,power_law_1.01,0.048563200235366824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,8,2,power_law_1.01,0.04590719938278198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,8,2,power_law_1.01,0.05368319749832153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,8,2,power_law_1.01,0.05514240264892578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,8,2,power_law_1.01,0.05496320128440857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,8,2,power_law_1.01,0.05727360248565674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,8,2,power_law_1.01,0.059436798095703125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,8,2,power_law_1.01,0.05934720039367676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,8,2,power_law_1.01,0.06285439729690552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,8,2,power_law_1.01,0.05931519865989685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,8,2,power_law_1.01,0.06572160124778748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,8,2,power_law_1.01,0.06609280109405517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,8,2,power_law_1.01,0.06686080098152161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,8,2,power_law_1.01,0.07304959893226623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,8,2,power_law_1.01,0.08609279990196228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,8,2,power_law_1.01,0.08005120158195496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,8,2,power_law_1.01,0.09743360280990601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,8,2,power_law_1.01,0.10383360385894776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,8,2,power_law_1.01,0.11928319931030273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,8,2,power_law_1.01,0.147052800655365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,8,2,power_law_1.01,0.17280639410018922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,8,2,power_law_1.01,0.23365120887756347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,8,2,power_law_1.01,0.2927295923233032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,8,2,power_law_1.01,0.3962111949920654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,8,2,power_law_1.01,0.5863359928131103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,8,2,power_law_1.01,0.7556608200073243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,8,2,power_law_1.01,0.8421952247619628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,8,2,power_law_1.01,1.4964799880981445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,8,2,power_law_1.01,2.2722944259643554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,4,power_law_1.01,0.05857279896736145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,4,power_law_1.01,0.06414080262184144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,4,power_law_1.01,0.05815039873123169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,4,power_law_1.01,0.06832640171051026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,4,power_law_1.01,0.08124160170555114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,4,power_law_1.01,0.0945792019367218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,4,power_law_1.01,0.09555839896202087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,4,power_law_1.01,0.10097919702529908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,4,power_law_1.01,0.10249600410461426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,4,power_law_1.01,0.10340479612350464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,4,power_law_1.01,0.10522240400314331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,4,power_law_1.01,0.10883840322494506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,4,power_law_1.01,0.11151360273361206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,4,power_law_1.01,0.11597440242767335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,4,power_law_1.01,0.12560640573501586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,4,power_law_1.01,0.12963199615478516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,4,power_law_1.01,0.13973759412765502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,4,power_law_1.01,0.15641599893569946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,4,power_law_1.01,0.16628479957580566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,4,power_law_1.01,0.19720319509506226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,4,power_law_1.01,0.22700159549713134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,4,power_law_1.01,0.28670079708099366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,4,power_law_1.01,0.34688639640808105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,4,power_law_1.01,0.5140351772308349
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,4,power_law_1.01,0.6274496078491211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,4,power_law_1.01,0.8668928146362305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,4,power_law_1.01,1.1108351707458497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,4,power_law_1.01,1.4056320190429688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,4,power_law_1.01,2.242291259765625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,4,power_law_1.01,4.401651382446289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,1,2,power_law_1.2,0.06835839748382569
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,1,2,power_law_1.2,0.0791167974472046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,1,2,power_law_1.2,0.07891200184822082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,1,2,power_law_1.2,0.10118399858474732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,1,2,power_law_1.2,0.09881600141525268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,1,2,power_law_1.2,0.10252159833908081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,1,2,power_law_1.2,0.10243840217590332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,1,2,power_law_1.2,0.10459519624710083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,1,2,power_law_1.2,0.11072640419006348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,1,2,power_law_1.2,0.11047680377960205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,1,2,power_law_1.2,0.11132160425186158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,1,2,power_law_1.2,0.11482880115509034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,1,2,power_law_1.2,0.11676160097122193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,1,2,power_law_1.2,0.12389119863510131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,1,2,power_law_1.2,0.13814400434494017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,1,2,power_law_1.2,0.14199039936065674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,1,2,power_law_1.2,0.15991679430007935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,1,2,power_law_1.2,0.18299520015716553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,1,2,power_law_1.2,0.21822719573974608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,1,2,power_law_1.2,0.27397119998931885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,1,2,power_law_1.2,0.3471168041229248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,1,2,power_law_1.2,0.4456768035888672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,1,2,power_law_1.2,0.4948927879333496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,1,2,power_law_1.2,0.9216704368591309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,1,2,power_law_1.2,1.218348789215088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,1,2,power_law_1.2,1.9161535263061524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,1,2,power_law_1.2,2.377952003479004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,1,2,power_law_1.2,3.028633689880371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,1,2,power_law_1.2,3.987161636352539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,1,2,power_law_1.2,7.788166046142578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,8,balanced,0.04574933151404063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,8,balanced,0.04573333263397217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,8,balanced,0.04971733192602793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,8,balanced,0.07340266803900401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,8,balanced,0.09880533814430237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,8,balanced,0.14899200201034546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,8,balanced,0.14946666359901428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,8,balanced,0.1520639955997467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,8,balanced,0.15433067083358765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,8,balanced,0.15382400155067444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,8,balanced,0.15441600481669107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,8,balanced,0.15615466237068176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,8,balanced,0.15668267011642456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,8,balanced,0.15837867061297098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,8,balanced,0.16729599237442017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,8,balanced,0.16795200109481812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,8,balanced,0.17432000239690146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,8,balanced,0.1876586675643921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,8,balanced,0.19594667355219522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,8,balanced,0.21866132815678915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,8,balanced,0.2541653315226237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,8,balanced,0.3125813404719035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,8,balanced,0.33850133419036865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,8,balanced,0.4453866481781006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,8,balanced,0.5386240084966024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,8,balanced,0.7394399642944336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,8,balanced,0.9294079939524332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,8,balanced,1.1670986811319988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,8,balanced,1.724474589029948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,8,balanced,3.3659467697143555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,32,1,power_law_1.2,0.032627201080322264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,32,1,power_law_1.2,0.03333759903907776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,32,1,power_law_1.2,0.03245440125465393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,32,1,power_law_1.2,0.03699199855327606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,32,1,power_law_1.2,0.03785600066184998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,32,1,power_law_1.2,0.0379584014415741
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,32,1,power_law_1.2,0.041068801283836366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,32,1,power_law_1.2,0.0409855991601944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,32,1,power_law_1.2,0.04445439875125885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,32,1,power_law_1.2,0.045177599787712096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,32,1,power_law_1.2,0.045228800177574156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,32,1,power_law_1.2,0.050387197732925416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,32,1,power_law_1.2,0.052070397138595584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,32,1,power_law_1.2,0.05310080051422119
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,32,1,power_law_1.2,0.06407039761543273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,32,1,power_law_1.2,0.06492800116539002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,32,1,power_law_1.2,0.07472000122070313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,32,1,power_law_1.2,0.08878719806671143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,32,1,power_law_1.2,0.10189440250396728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,32,1,power_law_1.2,0.13312000036239624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,32,1,power_law_1.2,0.14434560537338256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,32,1,power_law_1.2,0.20295679569244385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,32,1,power_law_1.2,0.24184958934783934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,32,1,power_law_1.2,0.3433727979660034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,32,1,power_law_1.2,0.4422719955444336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,32,1,power_law_1.2,0.638099193572998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,32,1,power_law_1.2,0.8198335647583008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,32,1,power_law_1.2,1.0163007736206056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,32,1,power_law_1.2,1.585318374633789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,32,1,power_law_1.2,3.114521598815918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,32,power_law_1.01,0.03389439880847931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,32,power_law_1.01,0.036396801471710205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,32,power_law_1.01,0.03377920091152191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,32,power_law_1.01,0.03203839957714081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,32,power_law_1.01,0.033267199993133545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,32,power_law_1.01,0.03335039913654327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,32,power_law_1.01,0.034169599413871765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,32,power_law_1.01,0.03450239896774292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,32,power_law_1.01,0.0353408008813858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,32,power_law_1.01,0.035622400045394895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,32,power_law_1.01,0.03713279962539673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,16,4,power_law_1.01,0.04292480051517487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,16,4,power_law_1.01,0.03456639945507049
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,32,power_law_1.01,0.03809280097484589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,16,4,power_law_1.01,0.032595199346542356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,16,4,power_law_1.01,0.03277440071105957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,16,4,power_law_1.01,0.033478400111198424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,16,4,power_law_1.01,0.03403519988059998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,16,4,power_law_1.01,0.03519999980926514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,32,power_law_1.01,0.03914240002632141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,16,4,power_law_1.01,0.0371071994304657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,16,4,power_law_1.01,0.0400191992521286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,16,4,power_law_1.01,0.04044159948825836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,16,4,power_law_1.01,0.04142079949378967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,32,power_law_1.01,0.04081920087337494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,16,4,power_law_1.01,0.04585599899291992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,16,4,power_law_1.01,0.045900800824165346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,16,4,power_law_1.01,0.04728319942951202
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,16,4,power_law_1.01,0.05735039710998535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,32,power_law_1.01,0.04577920138835907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,16,4,power_law_1.01,0.06292480230331421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,16,4,power_law_1.01,0.06778240203857422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,16,4,power_law_1.01,0.07687039971351624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,16,4,power_law_1.01,0.08526080250740051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,32,power_law_1.01,0.046239998936653134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,16,4,power_law_1.01,0.11075839996337891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,16,4,power_law_1.01,0.13771519660949708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,16,4,power_law_1.01,0.15598080158233643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,32,power_law_1.01,0.04859519898891449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,16,4,power_law_1.01,0.18703999519348144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,16,4,power_law_1.01,0.29670400619506837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,16,4,power_law_1.01,0.3638592004776001
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,32,power_law_1.01,0.05483520030975342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,16,4,power_law_1.01,0.523795223236084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,16,4,power_law_1.01,0.608076810836792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,32,power_law_1.01,0.06119040250778198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,16,4,power_law_1.01,0.7678016185760498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,32,power_law_1.01,0.07085440158843995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,16,4,power_law_1.01,1.3617600440979003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,32,power_law_1.01,0.09605119824409485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,16,4,power_law_1.01,2.6428800582885743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,32,power_law_1.01,0.11990400552749633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,32,power_law_1.01,0.1377087950706482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,32,power_law_1.01,0.1970944046974182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,32,power_law_1.01,0.2412031888961792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,32,power_law_1.01,0.38273279666900634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,32,power_law_1.01,0.49658880233764646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,32,power_law_1.01,0.5321407794952393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,32,power_law_1.01,0.870969581604004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,32,power_law_1.01,1.7982143402099608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,8,balanced,0.03142400085926056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,8,balanced,0.03188266605138779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,8,balanced,0.031386665999889374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,8,balanced,0.03405333310365677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,8,balanced,0.037808001041412354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,8,balanced,0.03972800076007843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,8,balanced,0.03963200002908707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,8,balanced,0.03967999915281931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,8,balanced,0.0414986660083135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,8,balanced,0.039546666045983635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,8,balanced,0.04153066625197729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,8,balanced,0.04164266586303711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,8,balanced,0.042863999803860985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,8,balanced,0.042170668641726174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,8,balanced,0.04808533191680908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,8,balanced,0.04790399968624115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,8,balanced,0.05007466673851013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,8,balanced,0.05186133086681366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,8,balanced,0.06214400132497152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,8,balanced,0.07041066884994507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,8,balanced,0.08285333216190338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,8,balanced,0.11337600151697795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,8,balanced,0.14433599511782327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,8,balanced,0.18567466735839844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,8,balanced,0.2315679987271627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,8,balanced,0.32434133688608807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,8,balanced,0.41874667008717853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,8,balanced,0.5150986512502035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,8,balanced,0.79147736231486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,8,balanced,1.5262986818949382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,64,power_law_1.2,0.06303359866142273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,4,power_law_1.01,0.046316799521446225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,64,power_law_1.2,0.05958399772644043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,4,power_law_1.01,0.051577597856521606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,64,power_law_1.2,0.056518399715423585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,4,power_law_1.01,0.05847679972648621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,64,power_law_1.2,0.05808640122413635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,4,power_law_1.01,0.05483520030975342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,64,power_law_1.2,0.058316802978515624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,4,power_law_1.01,0.06213759779930115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,64,power_law_1.2,0.05905280113220215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,4,power_law_1.01,0.07549440264701843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,4,power_law_1.01,0.08009600043296813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,64,power_law_1.2,0.06080639958381653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,4,power_law_1.01,0.08984959721565247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,64,power_law_1.2,0.06117759943008423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,4,power_law_1.01,0.09716479778289795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,64,power_law_1.2,0.06263039708137512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,4,power_law_1.01,0.09761279821395874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,64,power_law_1.2,0.06302080154418946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,4,power_law_1.01,0.1002303957939148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,64,power_law_1.2,0.06279039978981019
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,4,power_law_1.01,0.10673279762268066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,64,power_law_1.2,0.06331520080566407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,4,power_law_1.01,0.11034879684448243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,64,power_law_1.2,0.0641215980052948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,64,power_law_1.2,0.06590080261230469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,4,power_law_1.01,0.11783679723739623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,64,power_law_1.2,0.07123839855194092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,4,power_law_1.01,0.13032959699630736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,64,power_law_1.2,0.07148799896240235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,4,power_law_1.01,0.13376640081405639
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,64,power_law_1.2,0.07571840286254883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,4,power_law_1.01,0.147161602973938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,64,power_law_1.2,0.08863999843597412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,4,power_law_1.01,0.1769215941429138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,64,power_law_1.2,0.09964159727096558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,4,power_law_1.01,0.17839360237121582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,64,power_law_1.2,0.1275264024734497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,4,power_law_1.01,0.21715199947357178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,4,power_law_1.01,0.2538815975189209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,64,power_law_1.2,0.14558720588684082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,64,power_law_1.2,0.1992576003074646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,4,power_law_1.01,0.3317951917648315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,64,power_law_1.2,0.24460160732269287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,4,power_law_1.01,0.4124608039855957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,64,power_law_1.2,0.3326591968536377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,4,power_law_1.01,0.5677184104919434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,64,power_law_1.2,0.43050241470336914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,4,power_law_1.01,0.7160639762878418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,64,power_law_1.2,0.6382207870483398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,4,power_law_1.01,1.0316160202026368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,64,power_law_1.2,0.8133312225341797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,4,power_law_1.01,1.3293760299682618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,64,power_law_1.2,0.9948160171508789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,4,power_law_1.01,1.631532859802246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,64,power_law_1.2,1.7408000946044921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,4,power_law_1.01,2.5067392349243165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,64,power_law_1.2,3.3638912200927735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,4,power_law_1.01,4.98853759765625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,4,4,power_law_1.01,0.034771201014518735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,4,4,power_law_1.01,0.036006399989128114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,4,4,power_law_1.01,0.03423359990119934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,4,4,power_law_1.01,0.03468799889087677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,4,4,power_law_1.01,0.03454720079898834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,4,4,power_law_1.01,0.03804160058498383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,4,4,power_law_1.01,0.041440001130104064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,4,4,power_law_1.01,0.0440447986125946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,4,4,power_law_1.01,0.04504320025444031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,64,power_law_1.01,0.10147839784622192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,4,4,power_law_1.01,0.04689280092716217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,4,4,power_law_1.01,0.047968000173568726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,4,4,power_law_1.01,0.05192959904670715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,4,4,power_law_1.01,0.05406079888343811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,4,4,power_law_1.01,0.055251199007034305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,4,4,power_law_1.01,0.06726400256156921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,4,4,power_law_1.01,0.06812160015106201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,4,4,power_law_1.01,0.07615360021591186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,4,4,power_law_1.01,0.09131519794464112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,4,4,power_law_1.01,0.09230719804763794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,4,4,power_law_1.01,0.1385599970817566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,4,4,power_law_1.01,0.14095360040664673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,64,power_law_1.01,0.09790080189704894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,4,4,power_law_1.01,0.1679808020591736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,4,4,power_law_1.01,0.2588992118835449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,4,4,power_law_1.01,0.35418241024017333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,4,4,power_law_1.01,0.5390336036682128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,4,4,power_law_1.01,0.6850944042205811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,4,4,power_law_1.01,0.8192319869995117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,64,power_law_1.01,0.09639040231704712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,4,4,power_law_1.01,1.0957568168640137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,4,4,power_law_1.01,1.4988287925720214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,64,power_law_1.01,0.07201279997825623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,4,4,power_law_1.01,3.493983840942383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,64,power_law_1.01,0.07756800055503846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,64,power_law_1.01,0.08515200018882751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,64,power_law_1.01,0.08789759874343872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,64,power_law_1.01,0.09375360012054443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,64,power_law_1.01,0.09340800046920776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,64,power_law_1.01,0.09423360228538513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,64,power_law_1.01,0.0905023992061615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,64,power_law_1.01,0.09630720019340515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,64,power_law_1.01,0.09624959826469422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,64,power_law_1.01,0.09764479994773864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,64,power_law_1.01,0.10769920349121094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,64,power_law_1.01,0.10310399532318115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,64,power_law_1.01,0.1126911997795105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,64,power_law_1.01,0.11497600078582763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,64,power_law_1.01,0.12700799703598023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,64,power_law_1.01,0.14744319915771484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,64,power_law_1.01,0.1687999963760376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,64,power_law_1.01,0.2220479965209961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,64,power_law_1.01,0.25525760650634766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,64,power_law_1.01,0.3494335889816284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,64,power_law_1.01,0.4156479835510254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,64,power_law_1.01,0.6233727931976318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,64,power_law_1.01,0.7867455959320069
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,64,power_law_1.01,0.9373503684997558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,64,power_law_1.01,1.5362175941467284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,64,power_law_1.01,2.8263423919677733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,1,balanced,0.057189335425694786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,1,balanced,0.06533866624037425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,1,balanced,0.08364799618721008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,1,balanced,0.1156213382879893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,1,balanced,0.17574934164683023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,1,balanced,0.28573866685231525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,1,balanced,0.3871626853942871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,1,balanced,0.3898293177286784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,1,balanced,0.3914453188578288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,1,balanced,0.39422933260599774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,1,balanced,0.3983199993769328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,1,balanced,0.4068746566772461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,1,balanced,0.4089759985605876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,1,balanced,0.4188586473464966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,1,balanced,0.4299519856770833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,1,balanced,0.43983999888102215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,1,balanced,0.46159998575846356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,1,balanced,0.5160426696141561
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,1,balanced,0.566975990931193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,1,balanced,0.6483253240585327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,1,balanced,0.7136693000793457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,1,balanced,0.8738719622294108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,1,balanced,1.072266658147176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,1,balanced,1.3842026392618816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,1,balanced,1.8135786056518555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,1,balanced,2.4421866734822593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,1,balanced,3.2907253901163735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,1,balanced,4.069567998250325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,1,balanced,6.111637115478516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,1,balanced,11.702208201090494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,8,balanced,0.0462773342927297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,8,balanced,0.048010667165120445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,8,balanced,0.05301866432030996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,8,balanced,0.0714026689529419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,8,balanced,0.09883733590443929
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,8,balanced,0.14896000425020853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,8,balanced,0.2007466753323873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,8,balanced,0.20203733444213867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,8,balanced,0.20056533813476562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,8,balanced,0.20148799816767374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,8,balanced,0.20189867417017618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,8,balanced,0.20895999670028687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,8,balanced,0.2067413330078125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,8,balanced,0.21101333697636923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,8,balanced,0.2163146734237671
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,8,balanced,0.21733333667119345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,8,balanced,0.2243946592013041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,8,balanced,0.2335253357887268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,8,balanced,0.24087466796239218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,8,balanced,0.260368009408315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,8,balanced,0.2788426677385966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,8,balanced,0.3228213389714559
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,8,balanced,0.36213334401448566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,8,balanced,0.43164265155792236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,8,balanced,0.5180639823277792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,8,balanced,0.6661653518676758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,8,balanced,0.8456106980641683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,8,balanced,1.042629321416219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,8,balanced,1.5074987411499023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,8,balanced,2.8940372467041016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,16,2,power_law_1.2,0.03699199855327606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,16,2,power_law_1.2,0.03650560081005096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,16,2,power_law_1.2,0.035673600435256955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,16,2,power_law_1.2,0.037171199917793274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,16,2,power_law_1.2,0.03607679903507233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,16,2,power_law_1.2,0.03665919899940491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,16,2,power_law_1.2,0.037964800000190736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,16,2,power_law_1.2,0.037836799025535585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,16,2,power_law_1.2,0.03863039910793305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,16,2,power_law_1.2,0.040627199411392215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,16,2,power_law_1.2,0.03925119936466217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,16,2,power_law_1.2,0.040345600247383116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,16,2,power_law_1.2,0.041382399201393125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,16,2,power_law_1.2,0.04322560131549835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,16,2,power_law_1.2,0.049772799015045166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,16,2,power_law_1.2,0.05130239725112915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,16,2,power_law_1.2,0.0562175989151001
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,16,2,power_law_1.2,0.06321920156478882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,16,2,power_law_1.2,0.06609920263290406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,16,2,power_law_1.2,0.08326399922370911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,16,2,power_law_1.2,0.10280959606170655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,16,2,power_law_1.2,0.1356927990913391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,16,2,power_law_1.2,0.14908159971237184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,16,2,power_law_1.2,0.2171776056289673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,16,2,power_law_1.2,0.26362879276275636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,16,2,power_law_1.2,0.42658557891845705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,16,2,power_law_1.2,0.49336957931518555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,16,2,power_law_1.2,0.6094719886779785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,16,2,power_law_1.2,0.9784640312194824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,16,2,power_law_1.2,2.14017276763916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,64,power_law_1.2,0.0680895984172821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,64,power_law_1.2,0.059033602476119995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,64,power_law_1.2,0.05069440007209778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,64,power_law_1.2,0.05422080159187317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,64,power_law_1.2,0.055846399068832396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,64,power_law_1.2,0.057548797130584715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,64,power_law_1.2,0.0585536003112793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,64,power_law_1.2,0.056550401449203494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,64,power_law_1.2,0.0564736008644104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,64,power_law_1.2,0.057785600423812866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,64,power_law_1.2,0.06085759997367859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,64,power_law_1.2,0.06453760266304016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,64,power_law_1.2,0.0655680000782013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,64,power_law_1.2,0.06938880085945129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,64,power_law_1.2,0.07777919769287109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,64,power_law_1.2,0.08044800162315369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,64,power_law_1.2,0.08524159789085388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,64,power_law_1.2,0.10131839513778687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,64,power_law_1.2,0.10547840595245361
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,64,power_law_1.2,0.13379839658737183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,64,power_law_1.2,0.16219520568847656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,64,power_law_1.2,0.1950592041015625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,64,power_law_1.2,0.20328960418701172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,64,power_law_1.2,0.2616703987121582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,64,power_law_1.2,0.33162240982055663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,64,power_law_1.2,0.48705282211303713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,64,power_law_1.2,0.5464831829071045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,4,64,power_law_1.2,0.7204480171203613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,4,64,power_law_1.2,1.1224767684936523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,16,power_law_1.2,0.038406398892402646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,16,power_law_1.2,0.030508801341056824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,4,64,power_law_1.2,2.365964889526367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,16,power_law_1.2,0.030432000756263733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,16,power_law_1.2,0.031385600566864014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,16,power_law_1.2,0.03253760039806366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,16,power_law_1.2,0.03411200046539307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,16,power_law_1.2,0.03475199937820435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,16,power_law_1.2,0.037427198886871335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,16,power_law_1.2,0.03722879886627197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,16,power_law_1.2,0.03961600065231323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,16,power_law_1.2,0.04275839924812317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,16,power_law_1.2,0.044531199336051944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,16,power_law_1.2,0.04736000001430511
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,16,power_law_1.2,0.049132800102233885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,16,power_law_1.2,0.05475199818611145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,16,power_law_1.2,0.05739520192146301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,16,power_law_1.2,0.06032000184059143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,16,power_law_1.2,0.06746240258216858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,16,power_law_1.2,0.08187519907951354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,16,power_law_1.2,0.10126080513000488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,16,power_law_1.2,0.11064319610595703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,16,power_law_1.2,0.1429311990737915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,16,power_law_1.2,0.19931520223617555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,16,power_law_1.2,0.26634879112243653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,2,balanced,0.05248000224431356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,2,balanced,0.08298666775226593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,2,balanced,0.10078400373458862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,2,balanced,0.14858667055765787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,2,balanced,0.24726933240890503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,16,power_law_1.2,0.3702016115188599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,2,balanced,0.4440319935480754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,2,balanced,0.44545066356658936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,2,balanced,0.4481866757074992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,2,balanced,0.4498773415883382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,2,balanced,0.4536373217900594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,2,balanced,0.45578666528066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,16,power_law_1.2,0.5388224124908447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,2,balanced,0.46345067024230957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,2,balanced,0.4630560080210368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,2,balanced,0.471343994140625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,2,balanced,0.48415998617808026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,2,balanced,0.4891626834869385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,2,balanced,0.5015466610590616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,2,balanced,0.5259519815444946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,16,power_law_1.2,0.6614208221435547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,2,balanced,0.5467306772867838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,2,balanced,0.5961546500523885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,2,balanced,0.6539573272069296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,2,balanced,0.7509493033091227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,2,balanced,0.8447360197703043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,16,power_law_1.2,0.9493696212768554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,2,balanced,1.11627197265625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,2,balanced,1.2697866757710774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,2,balanced,1.7988373438517253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,2,balanced,2.2255093256632485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,16,power_law_1.2,1.6284351348876953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,2,balanced,2.803098678588867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,2,balanced,4.139424006144206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,2,balanced,8.020730972290039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,16,power_law_1.2,3.2119873046875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,32,8,power_law_1.01,0.02969599962234497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,32,8,power_law_1.01,0.029465600848197937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,32,8,power_law_1.01,0.028627198934555054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,32,8,power_law_1.01,0.02739199995994568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,32,8,power_law_1.01,0.028019198775291444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,32,8,power_law_1.01,0.02927359938621521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,32,8,power_law_1.01,0.029721599817276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,32,8,power_law_1.01,0.029708799719810487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,32,8,power_law_1.01,0.029542401432991028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,32,8,power_law_1.01,0.032569599151611325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,64,power_law_1.2,0.06888319849967957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,32,8,power_law_1.01,0.030022400617599487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,32,8,power_law_1.01,0.030451199412345885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,32,8,power_law_1.01,0.03134720027446747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,32,8,power_law_1.01,0.031327998638153075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,32,8,power_law_1.01,0.03464959859848023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,32,8,power_law_1.01,0.0363072007894516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,32,8,power_law_1.01,0.03927040100097656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,32,8,power_law_1.01,0.04542720019817352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,32,8,power_law_1.01,0.04929920136928558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,32,8,power_law_1.01,0.05568000078201294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,32,8,power_law_1.01,0.06679040193557739
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,32,8,power_law_1.01,0.078847998380661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,32,8,power_law_1.01,0.10145280361175538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,64,power_law_1.2,0.05443199872970581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,32,8,power_law_1.01,0.13886719942092896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,32,8,power_law_1.01,0.1776128053665161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,32,8,power_law_1.01,0.24574720859527588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,32,8,power_law_1.01,0.3070784091949463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,32,8,power_law_1.01,0.3592447996139526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,32,8,power_law_1.01,0.4717376232147217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,64,power_law_1.2,0.04707840085029602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,32,8,power_law_1.01,1.0237440109252929
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,64,power_law_1.2,0.04692479968070984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,64,power_law_1.2,0.048537600040435794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,64,power_law_1.2,0.049158400297164916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,64,power_law_1.2,0.04984959959983826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,64,power_law_1.2,0.0505407989025116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,64,power_law_1.2,0.051622402667999265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,32,power_law_1.2,0.1083456039428711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,64,power_law_1.2,0.052121597528457644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,32,power_law_1.2,0.09750400185585022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,64,power_law_1.2,0.052665597200393675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,32,power_law_1.2,0.06214399933815003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,64,power_law_1.2,0.056550401449203494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,32,power_law_1.2,0.06780160069465638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,64,power_law_1.2,0.056364798545837404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,32,power_law_1.2,0.07729920148849487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,64,power_law_1.2,0.06106240153312683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,32,power_law_1.2,0.09523839950561523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,64,power_law_1.2,0.06580479741096497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,32,power_law_1.2,0.09771519899368286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,64,power_law_1.2,0.06893439888954163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,32,power_law_1.2,0.10114560127258301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,64,power_law_1.2,0.08080000281333924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,32,power_law_1.2,0.09985920190811157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,32,power_law_1.2,0.10257920026779174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,64,power_law_1.2,0.09471359848976135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,32,power_law_1.2,0.1029312014579773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,64,power_law_1.2,0.10377600193023681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,32,power_law_1.2,0.10339200496673584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,64,power_law_1.2,0.1333631992340088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,32,power_law_1.2,0.11033600568771362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,64,power_law_1.2,0.15178879499435424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,32,power_law_1.2,0.11324800252914428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,64,power_law_1.2,0.2017535924911499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,32,power_law_1.2,0.1192255973815918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,64,power_law_1.2,0.2504319906234741
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,32,power_law_1.2,0.12403199672698975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,64,power_law_1.2,0.34603519439697267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,32,power_law_1.2,0.13825279474258423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,32,power_law_1.2,0.1622015953063965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,64,power_law_1.2,0.4383872032165527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,32,power_law_1.2,0.1661247968673706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,64,power_law_1.2,0.6624512195587158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,32,power_law_1.2,0.2194751977920532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,64,power_law_1.2,0.8169856071472168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,32,power_law_1.2,0.2375488042831421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,32,power_law_1.2,0.31639680862426756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,64,power_law_1.2,1.054086399078369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,32,power_law_1.2,0.3333695888519287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,64,power_law_1.2,1.6152320861816407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,32,power_law_1.2,0.4376704216003418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,32,power_law_1.2,0.576800012588501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,64,power_law_1.2,3.319091033935547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,32,power_law_1.2,0.7860928058624268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,32,power_law_1.2,0.8758784294128418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,1,32,power_law_1.2,1.2058176040649413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,4,balanced,0.032613334556420646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,4,balanced,0.03403199960788091
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,4,balanced,0.03409600009520849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,4,balanced,0.04018666595220566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,4,balanced,0.04598933458328247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,4,balanced,0.04499199986457825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,4,balanced,0.04629333317279816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,4,balanced,0.047024001677831016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,4,balanced,0.046384001771608986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,4,balanced,0.047509332497914634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,4,balanced,0.04641599953174591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,4,balanced,0.04938133557637533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,4,balanced,0.05049066742261251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,4,balanced,0.053802669048309326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,4,balanced,0.06392533580462138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,1,32,power_law_1.2,1.7813568115234375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,4,balanced,0.0639573335647583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,4,balanced,0.06565333406130473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,4,balanced,0.07210666437943776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,4,balanced,0.08323733508586884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,4,balanced,0.10942400495211284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,4,balanced,0.12286399801572163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,4,balanced,0.16214932998021445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,4,balanced,0.19717333714167276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,4,balanced,0.2709333300590515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,4,balanced,0.34169598420461017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,4,balanced,0.49003732204437256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,4,balanced,0.6421226660410563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,4,balanced,0.795530637105306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,4,balanced,1.2322346369425456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,4,balanced,2.379397392272949
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,1,32,power_law_1.2,3.989932632446289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,8,power_law_1.01,0.09583359956741333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,8,power_law_1.01,0.14627840518951415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,8,power_law_1.01,0.10338560342788697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,8,power_law_1.01,0.09262080192565918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,8,power_law_1.01,0.1338047981262207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,8,power_law_1.01,0.15084799528121948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,8,power_law_1.01,0.1873792052268982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,8,power_law_1.01,0.2037247896194458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,8,power_law_1.01,0.19942400455474854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,8,power_law_1.01,0.21354238986968993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,8,power_law_1.01,0.21188480854034425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,8,power_law_1.01,0.22607998847961425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,8,power_law_1.01,0.22249600887298585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,8,power_law_1.01,0.23795840740203858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,8,power_law_1.01,0.25528318881988527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,8,power_law_1.01,0.2568831920623779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,8,power_law_1.01,0.2709376096725464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,8,power_law_1.01,0.298035192489624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,8,power_law_1.01,0.2982271909713745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,8,power_law_1.01,0.3590336084365845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,8,power_law_1.01,0.36138238906860354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,8,power_law_1.01,0.44161281585693357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,8,power_law_1.01,0.5341375827789306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,8,power_law_1.01,0.6854656219482422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,8,power_law_1.01,0.8315072059631348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,8,power_law_1.01,1.1329407691955566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,8,power_law_1.01,1.4745792388916015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,8,power_law_1.01,1.6820863723754882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,8,power_law_1.01,2.6946943283081053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,8,power_law_1.01,4.9113727569580075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,2,balanced,0.056320001681645714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,2,balanced,0.08698667089144389
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,2,balanced,0.09899733463923137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,2,balanced,0.14940266807874045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,2,balanced,0.24599466721216837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,2,balanced,0.4416319926579793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,2,balanced,0.6404426495234171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,2,balanced,0.6389439900716146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,2,balanced,0.6405119895935059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,2,balanced,0.6414719820022583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,2,balanced,0.6462986469268799
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,2,balanced,0.649072011311849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,2,balanced,0.6524373292922974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,2,balanced,0.6588106552759806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,2,balanced,0.6716319719950358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,2,balanced,0.6750933329264323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,2,balanced,0.6890026728312174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,2,balanced,0.7121600310007731
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,2,balanced,0.7364799976348877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,2,balanced,0.7833066781361898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,2,balanced,0.8161599636077881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,2,balanced,0.9172693093617758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,2,balanced,0.9856159687042236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,2,balanced,1.1536693572998047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,2,balanced,1.4114774068196614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,2,balanced,1.8041653633117676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,2,balanced,2.284026622772217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,2,balanced,2.890538533528646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,2,balanced,4.059263865152995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,2,balanced,7.549674352010091
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,16,balanced,0.0533493310213089
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,16,balanced,0.04364266494909922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,16,balanced,0.045594667394955955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,16,balanced,0.047728002071380615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,16,balanced,0.05032533407211304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,16,balanced,0.07390933235486348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,16,balanced,0.07468266785144806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,16,balanced,0.07541866600513458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,16,balanced,0.07494399944941203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,16,balanced,0.07649066547552745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,16,balanced,0.07693333427111308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,16,balanced,0.07840533554553986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,16,balanced,0.07745600243409474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,16,balanced,0.0801333338022232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,16,balanced,0.08347200353940327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,16,balanced,0.08488000432650249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,16,balanced,0.0885759989420573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,16,balanced,0.09319999814033508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,16,balanced,0.09817066788673401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,16,balanced,0.10915199915568034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,16,balanced,0.12803199887275696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,16,balanced,0.14909866452217102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,16,balanced,0.16879467169443765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,16,balanced,0.21467200915018717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,16,balanced,0.2667679985364278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,16,balanced,0.3556533257166545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,16,balanced,0.4443306525548299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,16,balanced,0.5646880070368449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,16,balanced,0.8003679911295573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,16,balanced,1.547653357187907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,64,power_law_1.2,0.032569599151611325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,64,power_law_1.2,0.03314560055732727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,64,power_law_1.2,0.0348800003528595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,64,power_law_1.2,0.03294079899787903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,32,power_law_1.01,0.04668799936771393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,64,power_law_1.2,0.03405439853668213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,32,power_law_1.01,0.04586879909038544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,64,power_law_1.2,0.034860798716545106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,32,power_law_1.01,0.043084800243377686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,64,power_law_1.2,0.03488639891147614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,32,power_law_1.01,0.043884798884391785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,64,power_law_1.2,0.03529599905014038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,32,power_law_1.01,0.044300800561904906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,64,power_law_1.2,0.03555839955806732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,32,power_law_1.01,0.04576640129089356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,32,power_law_1.01,0.04679679870605469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,64,power_law_1.2,0.03580160140991211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,32,power_law_1.01,0.047737601399421695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,64,power_law_1.2,0.03849599957466125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,32,power_law_1.01,0.04721280038356781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,64,power_law_1.2,0.0390720009803772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,32,power_law_1.01,0.04888960123062134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,32,power_law_1.01,0.050310397148132326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,64,power_law_1.2,0.0406143993139267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,32,power_law_1.01,0.05073919892311096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,64,power_law_1.2,0.043059200048446655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,32,power_law_1.01,0.05145599842071533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,64,power_law_1.2,0.049107199907302855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,32,power_law_1.01,0.05432320237159729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,64,power_law_1.2,0.050419199466705325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,32,power_law_1.01,0.05934079885482788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,64,power_law_1.2,0.05297279953956604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,32,power_law_1.01,0.061452800035476686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,32,power_law_1.01,0.06170240044593811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,64,power_law_1.2,0.059487998485565186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,32,power_law_1.01,0.07004799842834472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,64,power_law_1.2,0.06993920207023621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,32,power_law_1.01,0.07662720084190369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,64,power_law_1.2,0.09605119824409485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,32,power_law_1.01,0.10342400074005127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,64,power_law_1.2,0.11178879737854004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,32,power_law_1.01,0.12468479871749878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,64,power_law_1.2,0.1505344033241272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,32,power_law_1.01,0.1628543972969055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,64,power_law_1.2,0.16826239824295045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,32,power_law_1.01,0.20225920677185058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,64,power_law_1.2,0.28771200180053713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,32,power_law_1.01,0.2798912048339844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,64,power_law_1.2,0.3609600067138672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,32,power_law_1.01,0.3474112033843994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,32,power_law_1.01,0.5008831977844238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,64,power_law_1.2,0.49776639938354494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,32,power_law_1.01,0.644985580444336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,64,power_law_1.2,0.7373248100280761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,32,power_law_1.01,0.7640768051147461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,64,power_law_1.2,0.840390396118164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,32,power_law_1.01,1.1971839904785155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,64,power_law_1.2,1.357855987548828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,32,power_law_1.01,2.249017524719238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,64,power_law_1.2,2.5825599670410155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,32,4,balanced,0.03916800022125244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,32,4,balanced,0.037776000797748566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,32,4,balanced,0.03775466730197271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,32,4,balanced,0.037418665985266365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,32,4,balanced,0.037503999968369804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,32,4,balanced,0.037717332442601524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,32,4,balanced,0.03770666569471359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,32,4,balanced,0.03797333439191183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,32,4,balanced,0.03774400055408478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,32,4,balanced,0.039434666434923805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,32,4,balanced,0.03813866774241129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,32,4,balanced,0.03977599988381068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,32,4,balanced,0.04137066751718521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,32,4,balanced,0.04165866722663244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,32,4,balanced,0.4551200071970622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,32,4,balanced,0.04563199977080027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,2,2,power_law_1.01,0.11917439699172974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,32,4,balanced,0.047695999344189964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,32,4,balanced,0.05229333539803823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,32,4,balanced,0.053802669048309326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,32,4,balanced,0.06044266621271769
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,32,4,balanced,0.07052800059318542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,32,4,balanced,0.088319996992747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,32,4,balanced,0.10498133301734924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,2,2,power_law_1.01,0.12119040489196778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,32,4,balanced,0.13638400038083395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,32,4,balanced,0.16530133287111917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,32,4,balanced,0.22874132792154947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,32,4,balanced,0.2828106681505839
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,32,4,balanced,0.3477226495742798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,2,2,power_law_1.01,0.16792320013046264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,32,4,balanced,0.523365338643392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,32,4,balanced,0.9904266993204752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,2,2,power_law_1.01,0.21690878868103028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,2,2,power_law_1.01,0.2112191915512085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,2,2,power_law_1.01,0.21093759536743165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,2,2,power_law_1.01,0.21697280406951905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,2,2,power_law_1.01,0.22177278995513916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,2,2,power_law_1.01,0.23740799427032472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,2,2,power_law_1.01,0.23522560596466063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,2,2,power_law_1.01,0.23182079792022706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,2,2,power_law_1.01,0.23635199069976806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,2,2,power_law_1.01,0.26143360137939453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,2,2,power_law_1.01,0.2475584030151367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,2,2,power_law_1.01,0.2667455911636353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,2,2,power_law_1.01,0.29626240730285647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,2,2,power_law_1.01,0.29317760467529297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,2,2,power_law_1.01,0.3552767992019653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,2,2,power_law_1.01,0.3995520114898682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,2,2,power_law_1.01,0.48944640159606934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,2,2,power_law_1.01,0.6589888095855713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,2,2,power_law_1.01,0.8605952262878418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,2,2,power_law_1.01,1.0111295700073242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,2,2,power_law_1.01,1.3404352188110351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,2,2,power_law_1.01,1.9722303390502929
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,2,2,power_law_1.01,3.1500223159790037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,2,2,power_law_1.01,3.488934326171875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,2,2,power_law_1.01,4.252460861206055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,32,4,power_law_1.01,0.0456063985824585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,32,4,power_law_1.01,0.05048959851264954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,2,2,power_law_1.01,7.839020538330078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,32,4,power_law_1.01,0.06392959952354431
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,32,4,power_law_1.01,0.05438079833984375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,2,2,power_law_1.01,13.950099182128906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,32,4,power_law_1.01,0.060601598024368285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,32,4,power_law_1.01,0.07304959893226623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,32,4,power_law_1.01,0.07719680070877075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,32,4,power_law_1.01,0.0926144003868103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,32,4,power_law_1.01,0.0942463994026184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,32,4,power_law_1.01,0.09711359739303589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,32,4,power_law_1.01,0.10194560289382934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,32,4,power_law_1.01,0.10516480207443238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,32,4,power_law_1.01,0.11029119491577148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,32,4,power_law_1.01,0.11248639822006226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,2,power_law_1.2,0.10124160051345825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,2,power_law_1.2,0.08791040182113648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,32,4,power_law_1.01,0.12049920558929443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,2,power_law_1.2,0.1218176007270813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,32,4,power_law_1.01,0.1260543942451477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,2,power_law_1.2,0.17484159469604493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,32,4,power_law_1.01,0.12872960567474365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,2,power_law_1.2,0.2519999980926514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,32,4,power_law_1.01,0.1457535982131958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,2,power_law_1.2,0.4432640075683594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,32,4,power_law_1.01,0.15707520246505738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,2,power_law_1.2,0.470854377746582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,32,4,power_law_1.01,0.20145280361175538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,2,power_law_1.2,0.5
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,32,4,power_law_1.01,0.22188799381256102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,2,power_law_1.2,0.5111999988555909
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,32,4,power_law_1.01,0.3023616075515747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,2,power_law_1.2,0.5281983852386475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,2,power_law_1.2,0.5304959774017334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,32,4,power_law_1.01,0.37922561168670654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,2,power_law_1.2,0.5354047775268554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,32,4,power_law_1.01,0.5032127857208252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,2,power_law_1.2,0.5504576206207276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,2,power_law_1.2,0.5780608177185058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,32,4,power_law_1.01,0.6220223903656006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,2,power_law_1.2,0.5979584217071533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,32,4,power_law_1.01,0.8449536323547363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,2,power_law_1.2,0.6150911808013916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,2,power_law_1.2,0.5996799945831299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,32,4,power_law_1.01,1.1680000305175782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,2,power_law_1.2,0.6969151973724366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,2,power_law_1.2,0.734822416305542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,32,4,power_law_1.01,1.4377984046936034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,2,power_law_1.2,0.8751680374145507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,2,power_law_1.2,0.9060352325439454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,32,4,power_law_1.01,2.063871955871582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,2,power_law_1.2,1.1794752120971679
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,2,power_law_1.2,1.1910464286804199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,2,power_law_1.2,1.4995136260986328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,32,4,power_law_1.01,4.363033676147461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,2,power_law_1.2,1.973721694946289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,2,power_law_1.2,2.5201791763305663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,2,power_law_1.2,3.150783920288086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,2,power_law_1.2,3.6485057830810548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,2,power_law_1.2,5.743289566040039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,8,2,power_law_1.2,0.035174399614334106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,8,2,power_law_1.2,0.034406399726867674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,8,2,power_law_1.2,0.03497599959373474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,8,2,power_law_1.2,0.036723199486732486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,8,2,power_law_1.2,0.037171199917793274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,8,2,power_law_1.2,0.03735679984092712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,8,2,power_law_1.2,0.037484800815582274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,8,2,power_law_1.2,0.03820799887180328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,8,2,power_law_1.2,0.03847039937973022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,2,power_law_1.2,10.04400634765625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,8,2,power_law_1.2,0.03948799967765808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,8,2,power_law_1.2,0.04111360013484955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,8,2,power_law_1.2,0.0441536009311676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,8,2,power_law_1.2,0.04392319917678833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,8,2,power_law_1.2,0.04782719910144806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,8,2,power_law_1.2,0.05723519921302796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,8,2,power_law_1.2,0.05728639960289002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,8,2,power_law_1.2,0.05734400153160095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,8,2,power_law_1.2,0.06918399930000305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,8,2,power_law_1.2,0.0825984001159668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,8,2,power_law_1.2,0.09594879746437072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,8,2,power_law_1.2,0.11862399578094482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,8,2,power_law_1.2,0.1605631947517395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,8,2,power_law_1.2,0.1736448049545288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,8,2,power_law_1.2,0.2374272108078003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,8,2,power_law_1.2,0.2812544107437134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,8,2,power_law_1.2,0.43112959861755373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,8,2,power_law_1.2,0.6235583782196045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,8,2,power_law_1.2,0.7932096004486084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,8,2,power_law_1.2,1.0710847854614258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,4,balanced,0.05000533163547516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,4,balanced,0.07541333138942719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,4,balanced,0.10174399614334106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,4,balanced,0.15727999806404114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,4,balanced,0.2552693287531535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,4,balanced,0.31126399834950763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,4,balanced,0.30905600388844806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,4,balanced,0.3128640055656433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,4,balanced,0.31383466720581055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,4,balanced,0.314191997051239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,4,balanced,0.3139999906222026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,8,2,power_law_1.2,2.087455940246582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,4,balanced,0.3169706662495931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,4,balanced,0.31804800033569336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,4,balanced,0.32157333691914874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,4,balanced,0.32947200536727905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,4,balanced,0.33075199524561566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,4,balanced,0.33628801504770917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,4,balanced,0.35946134726206463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,4,balanced,0.36002667744954425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,4,balanced,0.44469865163167316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,4,balanced,0.4050240119298299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,4,balanced,0.6452266772588094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,4,balanced,0.5175093412399292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,4,balanced,0.760042667388916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,4,balanced,0.7489813168843588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,4,balanced,1.0265333652496338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,4,balanced,1.3044906457265217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,4,balanced,1.492645263671875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,4,balanced,2.324965318044027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,4,balanced,4.367445309956868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,4,balanced,0.05845866600672404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,4,balanced,0.06648533542950948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,4,balanced,0.08657067020734151
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,4,balanced,0.11989866693814595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,4,balanced,0.17883199453353882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,4,balanced,0.2975040078163147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,4,balanced,0.29820799827575684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,4,balanced,0.3003466725349426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,4,balanced,0.30138667424519855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,4,balanced,0.3012320001920064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,4,balanced,0.3044266700744629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,4,balanced,0.3059733311335246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,4,balanced,0.3089119990666707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,4,balanced,0.31334932645161945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,4,balanced,0.3203146656354268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,4,balanced,0.32201600074768066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,4,balanced,0.32842133442560834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,4,balanced,0.339626669883728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,4,balanced,0.35542933146158856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,4,balanced,0.3835200071334839
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,4,balanced,0.41706665356953937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,4,balanced,0.47441065311431885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,4,balanced,0.5291359821955363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,4,balanced,0.7013546625773112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,4,balanced,0.8059679667154948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,4,balanced,1.1292800108591716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,4,balanced,1.3902559280395508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,4,balanced,1.7285119692484539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,4,balanced,2.578752040863037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,4,balanced,4.861850738525391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,16,balanced,0.049653331438700356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,16,balanced,0.052426666021347046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,16,balanced,0.0577706644932429
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,16,balanced,0.08281599978605907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,16,balanced,0.09813867012659709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,16,balanced,0.10872000455856323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,16,balanced,0.11004799604415894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,16,balanced,0.10985599954922994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,16,balanced,0.11008000373840332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,16,balanced,0.10987200339635213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,16,balanced,0.11055999994277954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,16,balanced,0.11443733175595601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,16,balanced,0.11469333370526631
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,16,balanced,0.11756267150243123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,16,balanced,0.12255466977755229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,16,balanced,0.12458133697509766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,16,balanced,0.12917866309483847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,16,balanced,0.14179199934005737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,16,balanced,0.15012799700101218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,16,balanced,0.18596800168355307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,16,balanced,0.20047465960184732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,16,balanced,0.281605343023936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,16,balanced,0.26239999135335285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,16,balanced,0.3558186690012614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,16,balanced,0.41621867815653485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,16,balanced,0.5821386575698853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,16,balanced,0.6896853446960449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,2,16,balanced,0.8508000373840332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,2,16,balanced,1.2509067058563232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,2,16,balanced,2.553215980529785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,4,power_law_1.01,0.05383679866790771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,4,power_law_1.01,0.08016639947891235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,4,power_law_1.01,0.05939199924468994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,4,power_law_1.01,0.06696959733963012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,4,power_law_1.01,0.08455680012702942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,4,power_law_1.01,0.10677119493484497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,4,power_law_1.01,0.10687359571456909
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,4,power_law_1.01,0.11141120195388794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,4,power_law_1.01,0.1137984037399292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,4,power_law_1.01,0.11588480472564697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,4,power_law_1.01,0.11816960573196411
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,4,power_law_1.01,0.12655999660491943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,4,power_law_1.01,0.1260543942451477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,4,power_law_1.01,0.13319679498672485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,4,power_law_1.01,0.13815679550170898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,4,power_law_1.01,0.14342399835586547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,4,power_law_1.01,0.15631359815597534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,4,power_law_1.01,0.17817599773406984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,4,power_law_1.01,0.18516479730606078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,4,power_law_1.01,0.22632319927215577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,4,power_law_1.01,0.2666944026947021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,4,power_law_1.01,0.3344255924224854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,4,power_law_1.01,0.406444787979126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,4,power_law_1.01,0.5578495979309082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,4,power_law_1.01,0.7078591823577881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,4,power_law_1.01,0.9292032241821289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,4,power_law_1.01,1.2684927940368653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,4,power_law_1.01,1.5920319557189941
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,4,power_law_1.01,2.484217643737793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,4,power_law_1.01,4.728927993774414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,16,power_law_1.2,0.0484607994556427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,16,power_law_1.2,0.033036801218986514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,16,power_law_1.2,0.030931198596954347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,16,power_law_1.2,0.03059200048446655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,16,power_law_1.2,0.031167998909950256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,16,power_law_1.2,0.033478400111198424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,16,power_law_1.2,0.03418880105018616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,16,power_law_1.2,0.03480319976806641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,16,power_law_1.2,0.0358271986246109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,16,power_law_1.2,0.0355648010969162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,16,power_law_1.2,0.03633280098438263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,16,power_law_1.2,0.040038400888442995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,16,power_law_1.2,0.04157440066337585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,16,power_law_1.2,0.044223999977111815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,16,power_law_1.2,0.0490880012512207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,16,power_law_1.2,0.05422080159187317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,16,power_law_1.2,0.0546176016330719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,16,power_law_1.2,0.06368640065193176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,16,power_law_1.2,0.06386560201644897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,16,power_law_1.2,0.07228800058364868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,16,power_law_1.2,0.08504319787025452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,16,power_law_1.2,0.10163840055465698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,16,power_law_1.2,0.12934399843215943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,16,power_law_1.2,0.14635519981384276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,16,power_law_1.2,0.19529600143432618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,16,power_law_1.2,0.29504640102386476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,16,power_law_1.2,0.33276801109313964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,1,16,power_law_1.2,0.3949631929397583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,1,16,power_law_1.2,0.6429376125335693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,1,16,power_law_1.2,1.484921646118164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,4,balanced,0.045498669147491455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,4,balanced,0.04764799773693085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,4,balanced,0.05198933184146881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,4,balanced,0.0746666689713796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,4,balanced,0.10422933101654053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,4,balanced,0.1537866691748301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,4,balanced,0.15412799517313638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,4,balanced,0.15870400269826254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,4,balanced,0.158053328593572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,4,balanced,0.15820266803105673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,4,balanced,0.1609440048535665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,4,balanced,0.1639306644598643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,4,balanced,0.16544000307718912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,4,balanced,0.1688800056775411
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,4,balanced,0.17922665675481161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,4,balanced,0.18045334021250406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,4,balanced,0.1897333264350891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,4,balanced,0.20695465803146362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,4,balanced,0.22114666302998862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,4,balanced,0.2494666576385498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,4,balanced,0.28437866767247516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,4,balanced,0.3526986837387085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,4,balanced,0.4137333234151204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,4,balanced,0.5708426634470621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,4,balanced,0.6827200253804525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,4,balanced,0.9581867059071859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,4,balanced,1.2091786861419678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,4,balanced,1.5115520159403484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,4,balanced,2.3061493237813315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,4,balanced,4.485573450724284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,4,power_law_1.2,0.06887680292129517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,4,power_law_1.2,0.11635199785232545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,4,power_law_1.2,0.0669376015663147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,4,power_law_1.2,0.08799359798431397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,4,power_law_1.2,0.1097983956336975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,4,power_law_1.2,0.1384063959121704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,4,power_law_1.2,0.18056319952011107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,4,power_law_1.2,0.19041919708251953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,4,power_law_1.2,0.19342080354690552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,4,power_law_1.2,0.1993600010871887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,4,power_law_1.2,0.22085120677947997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,4,power_law_1.2,0.2245568037033081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,4,power_law_1.2,0.22673280239105226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,4,power_law_1.2,0.24300799369812012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,4,power_law_1.2,0.2560256004333496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,4,power_law_1.2,0.25932159423828127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,4,power_law_1.2,0.27414400577545167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,4,power_law_1.2,0.3130111932754517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,4,power_law_1.2,0.32213759422302246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,4,power_law_1.2,0.4033088207244873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,4,power_law_1.2,0.42218880653381347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,4,power_law_1.2,0.5309120178222656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,4,power_law_1.2,0.6544703960418701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,4,power_law_1.2,0.8237248420715332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,4,power_law_1.2,0.9534784317016601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,4,power_law_1.2,1.4182592391967774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,4,power_law_1.2,1.8762943267822265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,4,power_law_1.2,2.24267520904541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,4,power_law_1.2,3.668025588989258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,4,power_law_1.2,6.521875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,8,balanced,0.029050665597120922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,8,balanced,0.030938667555650074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,8,balanced,0.02951466788848241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,8,balanced,0.03350933392842611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,8,balanced,0.035530666510264076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,8,balanced,0.03748266647259394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,8,balanced,0.037776000797748566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,8,balanced,0.03961066653331121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,8,balanced,0.03739733248949051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,8,balanced,0.03774933268626531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,8,balanced,0.03754133234421412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,8,balanced,0.03948266555865606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,8,balanced,0.03956266740957896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,8,balanced,0.03953066716591517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,8,balanced,0.043680002291997276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,8,balanced,0.0498879998922348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,8,balanced,0.048325334986050926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,8,balanced,0.054005334774653115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,8,balanced,0.05226666728655497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,8,balanced,0.06741333504517873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,8,balanced,0.07249600191911061
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,8,balanced,0.0906933347384135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,8,balanced,0.100490669409434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,8,balanced,0.13384000460306802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,8,balanced,0.15528000394503275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,8,balanced,0.2114773392677307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,8,balanced,0.24874667326609293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,2,8,balanced,0.30688534180323285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,2,8,balanced,0.45243199666341144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,2,8,balanced,0.8478666941324869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,4,2,balanced,0.025439999997615814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,4,2,balanced,0.027215999861558277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,4,2,balanced,0.027477333943049114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,4,2,balanced,0.03352533280849457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,4,2,balanced,0.06453866759936015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,4,2,balanced,0.0529013325770696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,4,2,balanced,0.05372266471385956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,4,2,balanced,0.05518933137257894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,4,2,balanced,0.05599466462930044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,4,2,balanced,0.056074668963750206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,4,2,balanced,0.05605866511662801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,4,2,balanced,0.058245331048965454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,4,2,balanced,0.05610666672388712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,4,2,balanced,0.06442666550477345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,4,2,balanced,0.07045333087444305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,4,2,balanced,0.0716480016708374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,4,2,balanced,0.07158400118350983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,4,2,balanced,0.08165333171685536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,2,balanced,0.08571733037630717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,2,balanced,0.10992532968521118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,2,balanced,0.11205333471298218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,2,balanced,0.14405866463979086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,2,balanced,0.16620266437530518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,2,balanced,0.24034667015075684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,2,balanced,0.3054879903793335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,2,balanced,0.4355039993921916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,2,balanced,0.5648426612218221
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,2,balanced,0.7003680070241293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,2,balanced,1.091589371363322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,2,balanced,2.039365291595459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,8,2,power_law_1.01,0.041536000370979306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,8,2,power_law_1.01,0.042656001448631284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,8,2,power_law_1.01,0.044614401459693906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,8,2,power_law_1.01,0.05173119902610779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,8,2,power_law_1.01,0.05711359977722168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,8,2,power_law_1.01,0.07057279944419861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,8,2,power_law_1.01,0.07391999959945679
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,8,2,power_law_1.01,0.07518079876899719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,8,2,power_law_1.01,0.07482879757881164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,8,2,power_law_1.01,0.0773311972618103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,8,2,power_law_1.01,0.07848320007324219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,8,2,power_law_1.01,0.0828224003314972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,8,2,power_law_1.01,0.08447999954223633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,8,2,power_law_1.01,0.08835840225219727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,8,2,power_law_1.01,0.09522560238838196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,8,2,power_law_1.01,0.10015360116958619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,8,2,power_law_1.01,0.10714240074157715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,8,2,power_law_1.01,0.12282240390777588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,8,2,power_law_1.01,0.14404480457305907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,8,2,power_law_1.01,0.1700160026550293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,8,2,power_law_1.01,0.20794880390167236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,8,2,power_law_1.01,0.2740351915359497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,8,2,power_law_1.01,0.3447808027267456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,8,2,power_law_1.01,0.5060416221618652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,8,2,power_law_1.01,0.6499839782714844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,8,2,power_law_1.01,0.9072832107543946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,8,2,power_law_1.01,1.1823424339294433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,8,2,power_law_1.01,1.5242431640625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,8,2,power_law_1.01,2.321107292175293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,8,2,power_law_1.01,4.613196945190429
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,1,power_law_1.01,0.06970239877700805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,1,power_law_1.01,0.09765120148658753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,1,power_law_1.01,0.11050879955291748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,1,power_law_1.01,0.18584320545196534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,1,power_law_1.01,0.2784127950668335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,1,power_law_1.01,0.4005824089050293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,1,power_law_1.01,0.5823872089385986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,1,power_law_1.01,0.6095424175262452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,1,power_law_1.01,0.6203711986541748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,1,power_law_1.01,0.6464896202087402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,1,power_law_1.01,0.6952127933502197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,1,power_law_1.01,0.6959167957305908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,1,power_law_1.01,0.7184512138366699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,1,power_law_1.01,0.7455039978027344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,1,power_law_1.01,0.7628416061401367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,1,power_law_1.01,0.7883840084075928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,1,power_law_1.01,0.8270079612731933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,1,power_law_1.01,0.9117695808410644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,1,power_law_1.01,0.9340224266052246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,1,power_law_1.01,1.0927103996276855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,1,power_law_1.01,1.1935999870300293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,1,power_law_1.01,1.399129581451416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,1,power_law_1.01,1.4914112091064453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,1,power_law_1.01,1.8856000900268555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,1,power_law_1.01,2.2747776031494142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,1,power_law_1.01,3.0795135498046875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,1,power_law_1.01,3.8876094818115234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,1,power_law_1.01,4.663872146606446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,1,power_law_1.01,6.755007934570313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,1,power_law_1.01,12.784537506103515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,1,4,balanced,0.09473066528638203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,1,4,balanced,0.12691733241081238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,1,4,balanced,0.1470186710357666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,1,4,balanced,0.1351040005683899
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,1,4,balanced,0.13709333539009094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,1,4,balanced,0.13821867108345032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,1,4,balanced,0.1379200021425883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,1,4,balanced,0.13857600092887878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,1,4,balanced,0.1474453310171763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,1,4,balanced,0.14735999703407288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,1,4,balanced,0.14430933197339377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,1,4,balanced,0.1443893313407898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,1,4,balanced,0.14390933513641357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,1,4,balanced,0.1511893371740977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,1,4,balanced,0.16174399852752686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,1,4,balanced,0.16190399726231894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,1,4,balanced,0.16859199603398642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,1,4,balanced,0.189082662264506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,1,4,balanced,0.2069173256556193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,1,4,balanced,0.2555039922396342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,1,4,balanced,0.26710933446884155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,1,4,balanced,0.35441064834594727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,1,4,balanced,0.4253386656443278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,1,4,balanced,0.609226663907369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,1,4,balanced,0.7954666614532471
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,1,4,balanced,1.162618637084961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,1,4,balanced,1.4736693700154622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,1,4,balanced,1.8359360694885254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,1,4,balanced,2.913130760192871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,1,4,balanced,5.754074732462565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,8,power_law_1.01,0.050348800420761106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,8,power_law_1.01,0.0722495973110199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,8,power_law_1.01,0.09755520224571228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,8,power_law_1.01,0.05294719934463501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,8,power_law_1.01,0.06811519861221313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,8,power_law_1.01,0.08352640271186829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,8,power_law_1.01,0.09153280258178711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,8,power_law_1.01,0.10205440521240235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,8,power_law_1.01,0.10535039901733398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,8,power_law_1.01,0.10380799770355224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,8,power_law_1.01,0.10830719470977783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,8,power_law_1.01,0.11451519727706909
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,8,power_law_1.01,0.11599359512329102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,8,power_law_1.01,0.12491519451141357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,8,power_law_1.01,0.13381760120391845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,8,power_law_1.01,0.13330559730529784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,8,power_law_1.01,0.13845759630203247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,8,power_law_1.01,0.15484800338745117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,8,power_law_1.01,0.1708351969718933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,8,power_law_1.01,0.19553920030593872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,8,power_law_1.01,0.2096127986907959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,8,power_law_1.01,0.2671168088912964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,8,power_law_1.01,0.32880640029907227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,8,power_law_1.01,0.4105408191680908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,8,power_law_1.01,0.5503551959991455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,8,power_law_1.01,0.7522816181182861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,8,power_law_1.01,0.8760448455810547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,8,power_law_1.01,1.1199999809265138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,8,power_law_1.01,1.590054416656494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,8,power_law_1.01,3.207846450805664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,1,balanced,0.0879253347714742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,1,balanced,0.12748799721399942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,1,balanced,0.19146132469177246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,1,balanced,0.33054399490356445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,1,balanced,0.5969280004501343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,1,balanced,0.8530666828155518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,1,balanced,0.8650453090667725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,1,balanced,0.8620213667551676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,1,balanced,0.8675839900970459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,1,balanced,0.871450662612915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,1,balanced,0.8795626958211263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,1,balanced,0.8824640115102133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,1,balanced,0.8899573485056559
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,1,balanced,0.9009386698404948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,1,balanced,0.916053295135498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,1,balanced,0.9281013011932373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,1,balanced,0.9547413190205892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,1,balanced,1.008245309193929
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,1,balanced,1.0611146291097004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,1,balanced,1.359055995941162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,1,balanced,1.246677319208781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,1,balanced,1.84389861424764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,1,balanced,1.716383934020996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,1,balanced,2.601802666982015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,1,balanced,2.8081067403157554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,1,balanced,4.372005462646484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,1,balanced,4.940634727478027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,2,1,balanced,6.488757451375325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,2,1,balanced,9.72985585530599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,2,1,balanced,18.625386555989582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,16,balanced,0.031530665854612984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,16,balanced,0.03153600047032038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,16,balanced,0.03149333347876867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,16,balanced,0.03374933451414108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,16,balanced,0.05570133527119955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,16,balanced,0.05460800230503082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,16,balanced,0.05681600173314413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,16,balanced,0.05557866891225179
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,16,balanced,0.05606933434804281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,16,balanced,0.056741332014401756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,16,balanced,0.056101332108179726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,16,balanced,0.056330665946006775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,16,balanced,0.058335999647776283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,16,balanced,0.05840533475081126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,16,balanced,0.0628053347269694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,16,balanced,0.06208533545335134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,16,balanced,0.06398400167624156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,16,balanced,0.06418666740258534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,16,balanced,0.06659199794133504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,16,balanced,0.07647466659545898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,16,balanced,0.0787306676308314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,16,balanced,0.09034132957458496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,16,balanced,0.10145599643389384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,16,balanced,0.12113599975903828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,16,balanced,0.15050666530927023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,16,balanced,0.1893813411394755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,16,balanced,0.2188106576601664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,16,balanced,0.265664001305898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,16,balanced,0.3804853359858195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,16,balanced,0.6905173460642496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,32,power_law_1.2,0.14911359548568726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,32,power_law_1.2,0.13456000089645387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,32,power_law_1.2,0.0765504002571106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,32,power_law_1.2,0.09155200123786926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,32,power_law_1.2,0.11463040113449097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,32,power_law_1.2,0.13889919519424437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,32,power_law_1.2,0.1396607995033264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,32,power_law_1.2,0.1412927985191345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,32,power_law_1.2,0.14599679708480834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,32,power_law_1.2,0.14094719886779786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,32,power_law_1.2,0.1456447958946228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,32,power_law_1.2,0.15493119955062867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,32,power_law_1.2,0.15242880582809448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,32,power_law_1.2,0.1588863968849182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,32,power_law_1.2,0.16599680185317994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,32,power_law_1.2,0.17694720029830932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,32,power_law_1.2,0.18624000549316405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,32,power_law_1.2,0.21653759479522705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,32,balanced,0.06744533280531566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,32,balanced,0.05229333539803823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,32,balanced,0.05178666611512502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,32,balanced,0.0705386648575465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,32,balanced,0.09521599610646565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,32,balanced,0.14627200365066528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,32,balanced,0.14763200283050537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,32,balanced,0.14644799629847208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,32,balanced,0.1476586659749349
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,32,balanced,0.1504853367805481
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,32,balanced,0.14846400419871011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,32,balanced,0.1497119963169098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,32,power_law_1.2,0.22919681072235107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,32,balanced,0.1514133314291636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,32,balanced,0.1514293352762858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,32,balanced,0.15702933073043823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,32,balanced,0.15781866510709128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,32,balanced,0.16127467155456543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,32,balanced,0.16577066977818808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,32,balanced,0.17023466030756632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,32,balanced,0.18042665719985962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,32,balanced,0.19363733132680258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,32,balanced,0.2116480072339376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,32,balanced,0.2294399936993917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,32,balanced,0.28893866141637164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,32,balanced,0.32130134105682373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,32,balanced,0.43195199966430664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,32,balanced,0.4965173403422038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,32,balanced,0.630944013595581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,32,power_law_1.2,0.29224319458007814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,32,balanced,0.8575519720713297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,32,balanced,1.664170742034912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,32,power_law_1.2,0.33391358852386477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,32,power_law_1.2,0.433465576171875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,32,power_law_1.2,0.423967981338501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,32,power_law_1.2,0.5285183906555175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,32,power_law_1.2,0.6749311923980713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,32,power_law_1.2,0.9671039581298828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,16,power_law_1.01,0.09663360118865967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,16,power_law_1.01,0.1506816029548645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,16,power_law_1.01,0.09296640157699584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,32,power_law_1.2,1.165113639831543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,16,power_law_1.01,0.07710080146789551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,16,power_law_1.01,0.10227199792861938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,16,power_law_1.01,0.10773119926452637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,16,power_law_1.01,0.1238592028617859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,1,32,power_law_1.2,1.4647040367126465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,16,power_law_1.01,0.12973439693450928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,16,power_law_1.01,0.14097919464111328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,16,power_law_1.01,0.13539199829101561
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,16,power_law_1.01,0.13697279691696168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,16,power_law_1.01,0.1408128023147583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,1,32,power_law_1.2,2.3500864028930666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,16,power_law_1.01,0.1448639988899231
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,16,power_law_1.01,0.14342399835586547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,16,power_law_1.01,0.15520000457763672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,16,power_law_1.01,0.16302080154418946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,16,power_law_1.01,0.1699072003364563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,16,power_law_1.01,0.19023360013961793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,16,power_law_1.01,0.20093441009521484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,16,power_law_1.01,0.24005119800567626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,1,32,power_law_1.2,5.27836799621582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,16,power_law_1.01,0.2539263963699341
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,16,power_law_1.01,0.321996808052063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,16,power_law_1.01,0.3845952033996582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,16,power_law_1.01,0.47887358665466306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,16,power_law_1.01,0.6293824195861817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,16,power_law_1.01,0.8577088356018067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,16,power_law_1.01,1.0301376342773438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,16,power_law_1.01,1.251699161529541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,16,power_law_1.01,2.1472959518432617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,128,power_law_1.2,0.035104000568389894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,128,power_law_1.2,0.0355648010969162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,128,power_law_1.2,0.03455359935760498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,128,power_law_1.2,0.036051198840141296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,128,power_law_1.2,0.03612160086631775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,128,power_law_1.2,0.03648639917373657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,16,power_law_1.01,3.8022720336914064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,128,power_law_1.2,0.03681280016899109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,128,power_law_1.2,0.03624959886074066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,128,power_law_1.2,0.036211198568344115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,128,power_law_1.2,0.036364799737930296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,128,power_law_1.2,0.03731200098991394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,128,power_law_1.2,0.037555199861526486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,128,power_law_1.2,0.038361600041389464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,128,power_law_1.2,0.03852159976959228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,128,power_law_1.2,0.04131839871406555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,128,power_law_1.2,0.04131200015544891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,128,power_law_1.2,0.043321600556373595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,128,power_law_1.2,0.045516800880432126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,128,power_law_1.2,0.04876160025596619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,128,power_law_1.2,0.05384320020675659
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,128,power_law_1.2,0.06121600270271301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,128,power_law_1.2,0.07269120216369629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,128,power_law_1.2,0.08677759766578674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,128,power_law_1.2,0.11095679998397827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,128,power_law_1.2,0.1367616057395935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,128,power_law_1.2,0.19480960369110106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,32,power_law_1.01,0.10977280139923096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,128,power_law_1.2,0.24056320190429686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,128,power_law_1.2,0.3089215993881226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,32,power_law_1.01,0.13455359935760497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,128,power_law_1.2,0.5005184173583984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,32,power_law_1.01,0.13514879941940308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,32,power_law_1.01,0.07785599827766418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,128,power_law_1.2,0.9680255889892578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,32,power_law_1.01,0.09381120204925537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,32,power_law_1.01,0.09991040229797363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,32,power_law_1.01,0.10364799499511719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,32,power_law_1.01,0.11940480470657348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,32,power_law_1.01,0.11287039518356323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,32,power_law_1.01,0.12507519721984864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,32,power_law_1.01,0.11905920505523682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,32,power_law_1.01,0.12226560115814208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,32,power_law_1.01,0.12319999933242798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,8,4,power_law_1.2,0.049772799015045166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,8,4,power_law_1.2,0.06354560256004334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,8,4,power_law_1.2,0.045433598756790164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,32,power_law_1.01,0.1241919994354248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,8,4,power_law_1.2,0.051667201519012454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,8,4,power_law_1.2,0.05232639908790589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,8,4,power_law_1.2,0.05334399938583374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,8,4,power_law_1.2,0.05589119791984558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,8,4,power_law_1.2,0.05680000185966492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,8,4,power_law_1.2,0.05792639851570129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,8,4,power_law_1.2,0.0624064028263092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,32,power_law_1.01,0.12798719406127929
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,8,4,power_law_1.2,0.0617792010307312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,8,4,power_law_1.2,0.06435199975967407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,8,4,power_law_1.2,0.0661184012889862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,8,4,power_law_1.2,0.06743680238723755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,8,4,power_law_1.2,0.08067200183868409
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,8,4,power_law_1.2,0.08001279830932617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,8,4,power_law_1.2,0.08487679958343505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,32,power_law_1.01,0.13225599527359008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,8,4,power_law_1.2,0.10586880445480347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,8,4,power_law_1.2,0.11269760131835938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,8,4,power_law_1.2,0.1522879958152771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,8,4,power_law_1.2,0.17549439668655395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,8,4,power_law_1.2,0.23080320358276368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,8,4,power_law_1.2,0.2809983968734741
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,32,power_law_1.01,0.1348863959312439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,8,4,power_law_1.2,0.36472320556640625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,8,4,power_law_1.2,0.5444159984588623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,8,4,power_law_1.2,0.7024767875671387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,32,power_law_1.01,0.14592000246047973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,8,4,power_law_1.2,1.108454418182373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,8,4,power_law_1.2,1.5002495765686035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,32,power_law_1.01,0.15797120332717896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,8,4,power_law_1.2,2.1376319885253907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,32,power_law_1.01,0.1723199963569641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,8,4,power_law_1.2,3.881836700439453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,32,power_law_1.01,0.20629758834838868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,32,power_law_1.01,0.2592384099960327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,32,power_law_1.01,0.2990272045135498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,32,power_law_1.01,0.41249918937683105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,32,power_law_1.01,0.5037951946258545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,32,power_law_1.01,0.6794432163238525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,32,power_law_1.01,0.8849344253540039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,64,power_law_1.2,0.04359680116176605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,32,power_law_1.01,1.0949503898620605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,64,power_law_1.2,0.0316864013671875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,64,power_law_1.2,0.027590399980545043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,64,power_law_1.2,0.02771199941635132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,64,power_law_1.2,0.028870400786399842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,32,power_law_1.01,1.7731647491455078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,64,power_law_1.2,0.02924799919128418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,64,power_law_1.2,0.029452800750732422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,64,power_law_1.2,0.029465600848197937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,64,power_law_1.2,0.029497599601745604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,64,power_law_1.2,0.029977598786354066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,64,power_law_1.2,0.029798400402069092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,64,power_law_1.2,0.030880001187324525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,32,power_law_1.01,3.2614784240722656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,64,power_law_1.2,0.031020799279212953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,64,power_law_1.2,0.03188480138778686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,64,power_law_1.2,0.03490560054779053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,64,power_law_1.2,0.03649280071258545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,64,power_law_1.2,0.036671999096870425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,64,power_law_1.2,0.038924801349639895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,64,power_law_1.2,0.04371840059757233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,64,power_law_1.2,0.048230400681495665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,64,power_law_1.2,0.04848000109195709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,64,power_law_1.2,0.05959039926528931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,64,power_law_1.2,0.07524480223655701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,64,power_law_1.2,0.09571840167045594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,64,power_law_1.2,0.12604160308837892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,64,power_law_1.2,0.18040319681167602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,64,power_law_1.2,0.19625600576400756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,4,64,power_law_1.2,0.2855871915817261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,4,64,power_law_1.2,0.39612159729003904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,1,4,power_law_1.01,0.030969598889350893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,1,4,power_law_1.01,0.038924801349639895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,4,64,power_law_1.2,0.8754303932189942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,1,4,power_law_1.01,0.03765760064125061
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,1,4,power_law_1.01,0.04395520091056824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,1,4,power_law_1.01,0.04844799935817719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,1,4,power_law_1.01,0.06798080205917359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,1,4,power_law_1.01,0.07121919989585876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,1,4,power_law_1.01,0.07111679911613464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,1,4,power_law_1.01,0.07535359859466553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,1,4,power_law_1.01,0.07182719707489013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,1,4,power_law_1.01,0.07575680017471313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,1,4,power_law_1.01,0.07466880083084107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,1,4,power_law_1.01,0.07502719759941101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,1,4,power_law_1.01,0.078847998380661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,1,4,power_law_1.01,0.08522239923477173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,1,4,power_law_1.01,0.08629119992256165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,1,4,power_law_1.01,0.08837760090827942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,1,4,power_law_1.01,0.09715840220451355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,4,power_law_1.01,0.103603196144104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,4,power_law_1.01,0.11755520105361938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,4,power_law_1.01,0.13211519718170167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,4,power_law_1.01,0.15777920484542846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,4,2,balanced,0.03345600018898646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,4,2,balanced,0.039706667264302574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,4,2,balanced,0.059621334075927734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,4,2,balanced,0.05832533538341522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,4,power_law_1.01,0.19202560186386108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,4,2,balanced,0.05849599838256836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,4,2,balanced,0.06108800073464712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,4,2,balanced,0.060640002290407814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,4,2,balanced,0.06241600215435028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,4,2,balanced,0.06253333389759064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,4,2,balanced,0.06376533210277557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,4,2,balanced,0.06222933530807495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,4,2,balanced,0.06420266628265381
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,4,2,balanced,0.06609066824118297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,4,2,balanced,0.06630399823188782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,4,2,balanced,0.07959466675917308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,4,2,balanced,0.0787360022465388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,4,2,balanced,0.08061866462230682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,4,2,balanced,0.08874133229255676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,4,power_law_1.01,0.24892799854278563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,4,2,balanced,0.09710400303204854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,4,2,balanced,0.11823466420173645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,4,2,balanced,0.13245333234469095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,4,2,balanced,0.17268800735473633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,4,2,balanced,0.21340266863505045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,4,2,balanced,0.29231999317804974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,4,2,balanced,0.3633973201115926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,4,2,balanced,0.5190293391545614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,4,power_law_1.01,0.29548799991607666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,4,2,balanced,0.6670133272806803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,4,2,balanced,0.8174826304117838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,4,2,balanced,1.2594292958577473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,4,power_law_1.01,0.41283841133117677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,4,2,balanced,2.445845286051432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,4,power_law_1.01,0.5075200080871582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,4,power_law_1.01,0.6362112045288086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,4,power_law_1.01,0.9961152076721191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,4,power_law_1.01,1.9486207962036133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,2,128,balanced,0.03368533402681351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,2,128,balanced,0.03397866586844126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,2,128,balanced,0.03345066557327906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,2,128,balanced,0.03366400053103765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,2,128,balanced,0.035258665680885315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,2,128,balanced,0.035642666121323906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,2,128,balanced,0.03537066777547201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,2,128,balanced,0.035631999373435974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,2,128,balanced,0.03563733398914337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,2,128,balanced,0.03557866563399633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,2,128,balanced,0.037717332442601524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,2,128,balanced,0.03554133325815201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,2,128,balanced,0.03544000039498011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,2,128,balanced,0.037685332198937736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,2,128,balanced,0.03955200066169103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,2,128,balanced,0.0396373321612676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,2,128,balanced,0.03962666789690653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,2,128,balanced,0.03977599988381068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,128,balanced,0.04200533529122671
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,128,balanced,0.041690667470296226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,128,balanced,0.043552001317342125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,128,balanced,0.04577066500981649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,128,balanced,0.04994666576385498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,128,balanced,0.054101333022117615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,128,balanced,0.06038400034109751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,128,balanced,0.07218133409818013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,128,balanced,0.08488532900810242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,128,balanced,0.10101866722106934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,128,balanced,0.1399786671002706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,128,balanced,0.2453440030415853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,64,power_law_1.2,0.06568319797515869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,64,power_law_1.2,0.06815360188484192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,64,power_law_1.2,0.05573760271072388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,64,power_law_1.2,0.06142079830169678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,64,power_law_1.2,0.06053760051727295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,64,power_law_1.2,0.06289280056953431
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,8,power_law_1.01,0.037567999958992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,64,power_law_1.2,0.06919040083885193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,8,power_law_1.01,0.02993920147418976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,64,power_law_1.2,0.07100160121917724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,8,power_law_1.01,0.03088639974594116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,8,power_law_1.01,0.030937600135803222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,64,power_law_1.2,0.06970239877700805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,8,power_law_1.01,0.03289600014686585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,8,power_law_1.01,0.03624320030212402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,64,power_law_1.2,0.07189120054244995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,8,power_law_1.01,0.03651840090751648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,64,power_law_1.2,0.07069439888000488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,8,power_law_1.01,0.03695360124111176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,8,power_law_1.01,0.038073599338531494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,64,power_law_1.2,0.07279360294342041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,8,power_law_1.01,0.037785598635673524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,8,power_law_1.01,0.04002560079097748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,64,power_law_1.2,0.07339519858360291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,8,power_law_1.01,0.04455040097236633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,8,power_law_1.01,0.04748159945011139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,64,power_law_1.2,0.07641599774360656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,8,power_law_1.01,0.049721598625183105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,64,power_law_1.2,0.08214399814605713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,8,power_law_1.01,0.05436800122261047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,8,power_law_1.01,0.056460797786712646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,64,power_law_1.2,0.08215680122375488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,8,power_law_1.01,0.05559039711952209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,8,power_law_1.01,0.06053760051727295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,64,power_law_1.2,0.08886399865150452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,8,power_law_1.01,0.0698751986026764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,64,power_law_1.2,0.10030720233917237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,8,power_law_1.01,0.08413439989089966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,8,power_law_1.01,0.09567999839782715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,64,power_law_1.2,0.11118719577789307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,8,power_law_1.01,0.11827199459075928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,64,power_law_1.2,0.1412992000579834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,8,power_law_1.01,0.144051194190979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,8,power_law_1.01,0.19690239429473877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,64,power_law_1.2,0.16913919448852538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,8,power_law_1.01,0.24519679546356202
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,64,power_law_1.2,0.2062079906463623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,8,power_law_1.01,0.3274240016937256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,64,power_law_1.2,0.25523838996887205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,8,power_law_1.01,0.42970881462097166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,64,power_law_1.2,0.3694591999053955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,8,power_law_1.01,0.5910975933074951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,64,power_law_1.2,0.46652798652648925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,8,power_law_1.01,0.7740543842315674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,64,power_law_1.2,0.7026432037353516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,64,power_law_1.2,0.9546048164367675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,8,power_law_1.01,1.6028032302856445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,64,power_law_1.2,1.157369613647461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,64,power_law_1.2,1.8788671493530273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,64,power_law_1.2,3.9291007995605467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,16,balanced,0.04364266494909922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,16,balanced,0.04371733466784159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,16,balanced,0.04366933306058248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,16,balanced,0.0459146648645401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,16,balanced,0.051818668842315674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,16,balanced,0.062261333068211876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,16,balanced,0.062208001812299095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,16,balanced,0.06470400094985962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,16,balanced,0.06589866677920024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,16,balanced,0.06628266473611195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,16,balanced,0.06585066517194112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,16,balanced,0.06798933446407318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,16,balanced,0.06784533460934956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,16,balanced,0.0687306672334671
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,16,balanced,0.07235200206438701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,16,balanced,0.07445866862932841
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,16,balanced,0.07433066765467326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,16,balanced,0.08853866656621297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,16,balanced,0.0846613347530365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,16,balanced,0.09984532992045085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,16,balanced,0.1051093339920044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,16,balanced,0.1302773356437683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,16,balanced,0.14075199762980142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,16,balanced,0.20082134008407593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,16,balanced,0.22163732846577963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,16,balanced,0.31457600990931195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,16,balanced,0.37918933232625324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,16,balanced,0.4336746533711751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,16,balanced,0.6550133228302002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,16,balanced,1.2419573465983074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,4,balanced,0.031930667658646904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,4,balanced,0.031680000325044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,4,balanced,0.03154666721820831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,4,balanced,0.03373866776625315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,4,balanced,0.03974399964014689
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,4,balanced,0.040789333482583366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,4,balanced,0.04142399877309799
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,4,balanced,0.04001066585381826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,4,balanced,0.04154666761557261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,4,balanced,0.04167466859022776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,4,balanced,0.041834667325019836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,4,balanced,0.044010668992996216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,4,balanced,0.043525333205858864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,4,balanced,0.04515733321507772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,4,balanced,0.05208533505598704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,4,balanced,0.051407997806866966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,4,balanced,0.053743998209635414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,4,balanced,0.06038933495680491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,4,balanced,0.0643093337615331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,4,balanced,0.08718400200208028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,4,balanced,0.09754666686058044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,4,balanced,0.12566933035850525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,4,balanced,0.15019733707110086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,4,balanced,0.2027733325958252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,4,balanced,0.24536534150441489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,4,balanced,0.3559360106786092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,4,balanced,0.46001601219177246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,4,balanced,0.5675253470738729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,4,balanced,0.8890079657236735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,4,balanced,1.71998929977417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,16,balanced,0.045935998360315956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,16,balanced,0.0487413356701533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,16,balanced,0.050245334704717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,16,balanced,0.07252266506354015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,16,balanced,0.0979146659374237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,16,balanced,0.14827733238538107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,16,balanced,0.1997013290723165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,16,balanced,0.1994240085283915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,16,balanced,0.19967466592788696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,16,balanced,0.20087466637293497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,16,balanced,0.20282665888468424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,16,balanced,0.20545067389806113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,16,balanced,0.20562134186426798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,16,balanced,0.207914670308431
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,16,balanced,0.21465599536895752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,16,balanced,0.21724800268809
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,16,balanced,0.22420799732208252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,16,balanced,0.23572800556818643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,16,balanced,0.24648000796635947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,16,balanced,0.26734934250513714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,16,balanced,0.2911146680514018
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,16,balanced,0.33484800656636554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,16,balanced,0.380351980527242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,16,balanced,0.4634026686350505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,16,balanced,0.5743360122044882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,16,balanced,0.7398186524709066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,64,power_law_1.01,0.11084159612655639
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,16,balanced,0.9448053042093912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,16,balanced,1.1561706860860188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,16,balanced,1.7008800506591797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,16,balanced,3.287285486857096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,64,power_law_1.01,0.11003520488739013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,8,4,balanced,0.04163199911514918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,8,4,balanced,0.03835733234882355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,64,power_law_1.01,0.10549119710922242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,8,4,balanced,0.03991466760635376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,8,4,balanced,0.03988266736268997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,8,4,balanced,0.03996799886226654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,8,4,balanced,0.040234667559464775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,8,4,balanced,0.038293334345022835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,8,4,balanced,0.0401653324564298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,8,4,balanced,0.04025600105524063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,8,4,balanced,0.039664000272750854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,8,4,balanced,0.03999999910593033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,8,4,balanced,0.03997866561015447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,8,4,balanced,0.04181866844495138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,8,4,balanced,0.04207466542720795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,8,4,balanced,0.05240533252557119
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,8,4,balanced,0.05204799771308899
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,8,4,balanced,0.05194666484991709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,8,4,balanced,0.05845333139101664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,8,4,balanced,0.06620266536871593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,8,4,balanced,0.0803466687599818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,8,4,balanced,0.08896000186602275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,8,4,balanced,0.10900266965230306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,8,4,balanced,0.1226026713848114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,8,4,balanced,0.15919466813405356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,8,4,balanced,0.19554666678110758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,8,4,balanced,0.26952000459035236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,64,power_law_1.01,0.07809919714927674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,8,4,balanced,0.3372906843821208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,8,4,balanced,0.4095253149668376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,8,4,balanced,0.6154613494873047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,8,4,balanced,1.1736053625742595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,64,power_law_1.01,0.09003520011901855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,64,power_law_1.01,0.0939136028289795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,64,power_law_1.01,0.09749760031700135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,64,power_law_1.01,0.10181759595870972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,64,power_law_1.01,0.1028864026069641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,64,power_law_1.01,0.09928320050239563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,64,power_law_1.01,0.10472320318222046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,64,power_law_1.01,0.10518399477005005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,64,power_law_1.01,0.1015936017036438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,32,power_law_1.01,0.03983359932899475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,64,power_law_1.01,0.10527360439300537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,32,power_law_1.01,0.03624320030212402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,32,power_law_1.01,0.03470720052719116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,64,power_law_1.01,0.11370240449905396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,32,power_law_1.01,0.034201601147651674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,32,power_law_1.01,0.03498879969120026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,64,power_law_1.01,0.11480319499969482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,32,power_law_1.01,0.036057600378990175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,32,power_law_1.01,0.036294400691986084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,64,power_law_1.01,0.11590399742126464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,32,power_law_1.01,0.036108800768852235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,32,power_law_1.01,0.03680639863014221
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,32,power_law_1.01,0.036057600378990175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,64,power_law_1.01,0.12063360214233398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,32,power_law_1.01,0.03739520013332367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,32,power_law_1.01,0.038515201210975646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,64,power_law_1.01,0.1324031949043274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,32,power_law_1.01,0.03888640105724335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,32,power_law_1.01,0.042028799653053284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,64,power_law_1.01,0.1550976037979126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,32,power_law_1.01,0.04499199986457825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,32,power_law_1.01,0.045491200685501096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,64,power_law_1.01,0.17559679746627807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,32,power_law_1.01,0.04818559885025024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,32,power_law_1.01,0.0523904025554657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,32,power_law_1.01,0.05429760217666626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,64,power_law_1.01,0.2194688081741333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,32,power_law_1.01,0.06212480068206787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,32,power_law_1.01,0.07109119892120361
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,64,power_law_1.01,0.24515841007232667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,32,power_law_1.01,0.08975359797477722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,32,power_law_1.01,0.11084799766540528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,64,power_law_1.01,0.33686399459838867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,32,power_law_1.01,0.1497599959373474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,32,power_law_1.01,0.19283839464187622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,64,power_law_1.01,0.4175551891326904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,32,power_law_1.01,0.28136959075927737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,32,power_law_1.01,0.32315518856048586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,64,power_law_1.01,0.5865024089813232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,32,power_law_1.01,0.4054272174835205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,64,power_law_1.01,0.7320064067840576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,32,power_law_1.01,0.646611213684082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,64,power_law_1.01,0.9064127922058105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,32,power_law_1.01,1.1675968170166016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,64,power_law_1.01,1.395263957977295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,64,power_law_1.01,2.7620607376098634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,8,power_law_1.2,0.03488639891147614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,8,power_law_1.2,0.031795200705528257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,8,power_law_1.2,0.030457600951194763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,8,power_law_1.2,0.031948798894882204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,8,power_law_1.2,0.03367680013179779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,8,power_law_1.2,0.03660799860954285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,8,power_law_1.2,0.03731839954853058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,8,power_law_1.2,0.0379967987537384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,8,power_law_1.2,0.038771200180053714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,8,power_law_1.2,0.03917439877986908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,8,power_law_1.2,0.04193280041217804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,8,power_law_1.2,0.04251520037651062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,8,power_law_1.2,0.045491200685501096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,8,power_law_1.2,0.043884798884391785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,8,power_law_1.2,0.04834559857845307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,8,power_law_1.2,0.050387197732925416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,8,power_law_1.2,0.054451197385787964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,8,power_law_1.2,0.06454399824142457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,8,power_law_1.2,0.07895039916038513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,8,power_law_1.2,0.09431040287017822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,8,power_law_1.2,0.11438720226287842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,8,power_law_1.2,0.17420799732208253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,8,power_law_1.2,0.18458880186080934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,8,power_law_1.2,0.27560958862304685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,8,power_law_1.2,0.3543231964111328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,8,power_law_1.2,0.4901887893676758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,8,power_law_1.2,0.6912447929382324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,8,power_law_1.2,0.8198911666870117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,8,power_law_1.2,1.7024896621704102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,8,power_law_1.2,2.932153511047363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,2,power_law_1.2,0.05270400047302246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,2,power_law_1.2,0.07258880138397217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,2,power_law_1.2,0.07498239874839782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,2,power_law_1.2,0.08498560190200806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,2,power_law_1.2,0.12215039730072022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,2,power_law_1.2,0.14866559505462645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,2,power_law_1.2,0.17610880136489868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,2,power_law_1.2,0.22851200103759767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,2,power_law_1.2,0.24494080543518065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,2,power_law_1.2,0.25188479423522947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,2,power_law_1.2,0.26572160720825194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,2,power_law_1.2,0.2854079961776733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,2,power_law_1.2,0.28373119831085203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,2,power_law_1.2,0.2987776041030884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,2,power_law_1.2,0.32126080989837646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,2,power_law_1.2,0.33228800296783445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,2,power_law_1.2,0.35552639961242677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,2,power_law_1.2,0.3799040079116821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,2,power_law_1.2,0.4151167869567871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,2,power_law_1.2,0.4965951919555664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,2,power_law_1.2,0.5457087993621826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,2,power_law_1.2,0.6636672019958496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,2,power_law_1.2,0.8337599754333496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,2,power_law_1.2,1.0392319679260253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,2,power_law_1.2,1.352883243560791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,2,power_law_1.2,1.7840831756591797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,2,power_law_1.2,2.1263168334960936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,2,power_law_1.2,2.6602943420410154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,2,power_law_1.2,3.907392120361328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,2,power_law_1.2,7.2501884460449215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,32,1,power_law_1.01,0.03118079900741577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,32,1,power_law_1.01,0.03391999900341034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,32,1,power_law_1.01,0.03772160112857818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,32,1,power_law_1.01,0.0459199994802475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,32,1,power_law_1.01,0.054079997539520266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,32,1,power_law_1.01,0.07319039702415467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,32,1,power_law_1.01,0.07722240090370178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,32,1,power_law_1.01,0.08056960105895997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,32,1,power_law_1.01,0.08104320168495179
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,32,1,power_law_1.01,0.08184959888458251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,32,1,power_law_1.01,0.08531839847564697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,32,1,power_law_1.01,0.08954880237579346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,32,1,power_law_1.01,0.09307519793510437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,32,1,power_law_1.01,0.0991487979888916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,32,1,power_law_1.01,0.1071936011314392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,32,1,power_law_1.01,0.11348479986190796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,32,1,power_law_1.01,0.12065919637680053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,32,1,power_law_1.01,0.14277759790420533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,32,1,power_law_1.01,0.16870399713516235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,32,1,power_law_1.01,0.20679678916931152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,32,1,power_law_1.01,0.25493760108947755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,8,power_law_1.2,0.10373120307922364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,32,1,power_law_1.01,0.3530944108963013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,32,1,power_law_1.01,0.41660161018371583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,8,power_law_1.2,0.0874239981174469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,32,1,power_law_1.01,0.5858943939208985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,32,1,power_law_1.01,0.7475840091705322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,8,power_law_1.2,0.0784704029560089
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,32,1,power_law_1.01,1.059398365020752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,8,power_law_1.2,0.09589120149612426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,32,1,power_law_1.01,1.3981439590454101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,32,1,power_law_1.01,1.7292160034179687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,8,power_law_1.2,0.11054079532623291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,32,1,power_law_1.01,2.709984016418457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,8,power_law_1.2,0.15559680461883546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,8,power_law_1.2,0.16403199434280397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,8,power_law_1.2,0.1708224058151245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,32,1,power_law_1.01,5.243929672241211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,8,power_law_1.2,0.16847360134124756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,8,power_law_1.2,0.17831039428710938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,8,power_law_1.2,0.1818112015724182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,8,power_law_1.2,0.18842240571975707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,8,power_law_1.2,0.19364479780197144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,8,power_law_1.2,0.20031359195709228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,8,power_law_1.2,0.2124095916748047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,8,power_law_1.2,0.21135358810424804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,8,power_law_1.2,0.22773759365081786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,8,power_law_1.2,0.2603327989578247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,8,power_law_1.2,0.2871551990509033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,8,power_law_1.2,0.3423680067062378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,8,power_law_1.2,0.42578558921813964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,8,power_law_1.2,0.5180416107177734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,8,power_law_1.2,0.5815680027008057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,8,power_law_1.2,0.7883264064788819
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,8,power_law_1.2,1.0194944381713866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,8,power_law_1.2,1.4081088066101075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,8,power_law_1.2,1.8242879867553712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,1,8,power_law_1.2,2.050182342529297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,1,8,power_law_1.2,3.466969680786133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,1,8,power_law_1.2,7.050739288330078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,2,balanced,0.07700266440709432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,2,balanced,0.1269706686337789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,2,balanced,0.15288000305493674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,2,balanced,0.25760533412297565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,2,balanced,0.4610666831334432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,2,balanced,0.5694293181101481
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,2,balanced,0.5723839998245239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,2,balanced,0.5750666856765747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,2,balanced,0.5766239960988363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,2,balanced,0.5796960194905599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,2,balanced,0.5802239974339803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,2,balanced,0.5832159916559855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,2,balanced,0.5865066846211752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,2,balanced,0.5931626558303833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,2,balanced,0.6036159992218018
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,2,balanced,0.606549342473348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,2,balanced,0.6152106523513794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,2,balanced,0.6579466660817465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,2,balanced,0.6516746679941813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,2,balanced,0.7830399672190348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,2,balanced,0.7392106850941976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,2,balanced,1.0636266867319744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,2,balanced,0.9068266550699869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,2,balanced,1.3281439940134685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,2,balanced,1.3565066655476887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,2,balanced,2.2022825876871743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,2,balanced,2.404090722401937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,2,balanced,2.8519252141316733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,2,balanced,4.324597358703613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,2,balanced,8.127215703328451
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,32,balanced,0.07718400160471599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,32,balanced,0.08423999945322673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,32,balanced,0.07844266792138417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,32,balanced,0.09645866354306538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,32,balanced,0.12044266859690349
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,32,balanced,0.19340799252192178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,32,balanced,0.24174400170644125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,32,balanced,0.2472533384958903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,32,balanced,0.25282132625579834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,32,balanced,0.24901866912841797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,32,balanced,0.26808534065882367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,32,balanced,0.25361599524815875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,32,balanced,0.2629599968592326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,32,balanced,0.27932266394297284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,32,balanced,0.25712533791859943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,32,balanced,0.26149866978327435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,32,balanced,0.2706399957338969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,32,balanced,0.26474666595458984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,32,balanced,0.27742934226989746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,32,balanced,0.28310932715733844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,32,balanced,0.29286400477091473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,32,balanced,0.3146773378054301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,32,balanced,0.3384053309758504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,32,balanced,0.39420799414316815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,32,balanced,0.5478773514429728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,32,balanced,0.6222079992294312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,32,balanced,0.808240016301473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,32,balanced,1.085477352142334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,32,balanced,1.5782507260640461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,32,balanced,2.773791948954264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,8,power_law_1.2,0.09476479887962341
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,8,power_law_1.2,0.16123520135879515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,8,power_law_1.2,0.07316480278968811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,8,power_law_1.2,0.089683198928833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,8,power_law_1.2,0.10453120470046998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,8,power_law_1.2,0.14565119743347169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,8,power_law_1.2,0.18119679689407348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,8,power_law_1.2,0.2041408061981201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,8,power_law_1.2,0.1999168038368225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,8,power_law_1.2,0.20383999347686768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,8,power_law_1.2,0.20327680110931395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,8,power_law_1.2,0.218176007270813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,8,power_law_1.2,0.22685439586639405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,8,power_law_1.2,0.2242432117462158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,8,power_law_1.2,0.23228158950805664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,8,power_law_1.2,0.25217280387878416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,8,power_law_1.2,0.27164158821105955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,8,power_law_1.2,0.3038975954055786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,8,power_law_1.2,0.3006144046783447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,8,power_law_1.2,0.3529279947280884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,8,power_law_1.2,0.3636352062225342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,8,power_law_1.2,0.4665664196014404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,8,power_law_1.2,0.5232704162597657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,8,power_law_1.2,0.643558406829834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,8,power_law_1.2,0.7795711994171143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,8,power_law_1.2,1.1691776275634767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,8,power_law_1.2,1.3001279830932617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,8,power_law_1.2,1.5438464164733887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,8,power_law_1.2,2.342176055908203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,8,power_law_1.2,5.184735870361328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,16,balanced,0.04612799982229868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,16,balanced,0.04584533472855886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,16,balanced,0.05012799799442291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,16,balanced,0.07664533456166585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,16,balanced,0.10332799951235454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,16,balanced,0.11596266428629558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,16,balanced,0.11703466375668843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,16,balanced,0.11737599968910217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,16,balanced,0.11946133772532146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,16,balanced,0.11918933192888896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,16,balanced,0.11803199847539265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,16,balanced,0.1223520040512085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,16,balanced,0.12160000205039978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,16,balanced,0.12543466687202454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,16,balanced,0.13099199533462524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,16,balanced,0.1341493328412374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,16,balanced,0.13517866532007852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,16,balanced,0.15109866857528687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,16,balanced,0.1508746643861135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,16,balanced,0.19060800472895303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,16,balanced,0.18403732776641846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,16,balanced,0.23706134160359701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,16,balanced,0.25195199251174927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,16,balanced,0.3503626585006714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,16,balanced,0.40717331568400067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,16,balanced,0.5496106545130411
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,16,balanced,0.7056852976481119
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,16,balanced,0.8158986568450928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,16,balanced,1.2753012975056965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,16,balanced,2.434234619140625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,64,power_law_1.01,0.031711998581886294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,64,power_law_1.01,0.03165439963340759
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,64,power_law_1.01,0.031327998638153075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,64,power_law_1.01,0.03255679905414581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,64,power_law_1.01,0.03240320086479187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,64,power_law_1.01,0.03320960104465485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,64,power_law_1.01,0.03363839983940124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,64,power_law_1.01,0.03387520015239716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,64,power_law_1.01,0.03399679958820343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,64,power_law_1.01,0.03468160033226013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,64,power_law_1.01,0.03669120073318481
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,64,power_law_1.01,0.03743999898433685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,64,power_law_1.01,0.039084801077842714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,64,power_law_1.01,0.041510400176048276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,64,power_law_1.01,0.04598399996757507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,64,power_law_1.01,0.048256000876426695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,64,power_law_1.01,0.047993600368499756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,64,power_law_1.01,0.0554751992225647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,64,power_law_1.01,0.06159359812736511
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,64,power_law_1.01,0.07705600261688232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,64,power_law_1.01,0.08988159894943237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,64,power_law_1.01,0.12104959487915039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,64,power_law_1.01,0.13819520473480223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,64,power_law_1.01,0.19741439819335938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,64,power_law_1.01,0.24311680793762208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,64,power_law_1.01,0.3363071918487549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,64,power_law_1.01,0.41960320472717283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,64,power_law_1.01,0.5520895957946778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,64,power_law_1.01,0.7790527820587159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,64,power_law_1.01,1.6556735992431642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,128,power_law_1.2,0.03830400109291077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,128,power_law_1.2,0.039289599657058714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,128,power_law_1.2,0.03952000141143799
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,128,power_law_1.2,0.040006399154663086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,128,power_law_1.2,0.039827200770378116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,128,power_law_1.2,0.040191999077796935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,128,power_law_1.2,0.04049279987812042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,128,power_law_1.2,0.04046719968318939
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,128,power_law_1.2,0.04051199853420258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,128,power_law_1.2,0.04056960046291351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,128,power_law_1.2,0.04095360040664673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,128,power_law_1.2,0.04085119962692261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,128,power_law_1.2,0.041433599591255185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,128,power_law_1.2,0.042342400550842284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,128,power_law_1.2,0.04565759897232056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,128,power_law_1.2,0.04519680142402649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,128,power_law_1.2,0.04848639965057373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,128,power_law_1.2,0.05121279954910278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,128,power_law_1.2,0.056480002403259275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,128,power_law_1.2,0.06159999966621399
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,128,power_law_1.2,0.08038399815559387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,128,power_law_1.2,0.10056320428848267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,128,power_law_1.2,0.12321280241012574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,128,power_law_1.2,0.16977280378341675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,128,power_law_1.2,0.22225921154022216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,128,power_law_1.2,0.3121984004974365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,128,power_law_1.2,0.41042561531066896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,128,power_law_1.2,0.48993921279907227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,128,power_law_1.2,0.7792064189910889
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,128,power_law_1.2,1.5446720123291016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,16,power_law_1.01,0.1875712037086487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,16,power_law_1.01,0.249452805519104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,16,power_law_1.01,0.08545920252799988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,16,power_law_1.01,0.1285823941230774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,16,power_law_1.01,0.14362879991531372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,16,power_law_1.01,0.20718080997467042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,16,power_law_1.01,0.2211967945098877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,16,power_law_1.01,0.21931519508361816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,16,power_law_1.01,0.2262592077255249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,16,power_law_1.01,0.23119359016418456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,16,power_law_1.01,0.23117439746856688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,16,power_law_1.01,0.23866240978240966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,16,power_law_1.01,0.23999359607696533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,16,power_law_1.01,0.24924800395965577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,16,power_law_1.01,0.25635199546813964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,16,power_law_1.01,0.25615999698638914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,16,power_law_1.01,0.27156479358673097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,16,power_law_1.01,0.32140159606933594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,16,power_law_1.01,0.3211008071899414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,16,power_law_1.01,0.3994944095611572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,16,power_law_1.01,0.4311552047729492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,16,power_law_1.01,0.5244607925415039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,16,power_law_1.01,0.5549888134002685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,16,power_law_1.01,0.7145792007446289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,16,power_law_1.01,0.8204863548278809
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,16,power_law_1.01,1.177286434173584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,16,power_law_1.01,1.345759963989258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,1,16,power_law_1.01,1.6169727325439454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,1,16,power_law_1.01,2.583091163635254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,1,16,power_law_1.01,4.573100662231445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,8,balanced,0.05808533231417338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,8,balanced,0.06004266440868378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,8,balanced,0.07026666899522145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,32,balanced,0.0545653353134791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,8,balanced,0.09290666381518047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,8,balanced,0.11599999666213989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,32,balanced,0.04640533526738485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,8,balanced,0.1360106666882833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,32,balanced,0.046096002062161766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,8,balanced,0.13645866513252258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,32,balanced,0.04641599953174591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,8,balanced,0.13635200262069702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,32,balanced,0.05080533524354299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,8,balanced,0.1381439963976542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,32,balanced,0.07277333239714305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,8,balanced,0.13794133067131042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,32,balanced,0.07348266740640004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,8,balanced,0.14260266224543253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,8,balanced,0.14221333463986716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,32,balanced,0.07399466633796692
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,8,balanced,0.1450399955113729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,32,balanced,0.07478933533032735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,8,balanced,0.14642133315404257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,32,balanced,0.07351466516653697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,8,balanced,0.1534773310025533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,32,balanced,0.07540800174077351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,8,balanced,0.155349334081014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,32,balanced,0.07522666454315186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,8,balanced,0.1630773345629374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,32,balanced,0.07620266576608022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,8,balanced,0.18214933077494302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,32,balanced,0.07709333300590515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,8,balanced,0.19713600476582846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,32,balanced,0.08347733815511067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,8,balanced,0.27499733368555707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,32,balanced,0.08124266564846039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,32,balanced,0.08551999926567078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,8,balanced,0.2714826663335164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,32,balanced,0.0900266667207082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,8,balanced,0.3739466667175293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,32,balanced,0.09198400378227234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,8,balanced,0.39058132966359455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,32,balanced,0.10473066568374634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,32,balanced,0.11215999722480774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,8,balanced,0.5453973213831583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,32,balanced,0.13025599718093872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,32,balanced,0.1492800017197927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,8,balanced,0.636410673459371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,32,balanced,0.18823466698328653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,32,balanced,0.2251466711362203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,8,balanced,0.9239733219146729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,32,balanced,0.30531734228134155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,8,balanced,1.1287413438161213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,32,balanced,0.3768213192621867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,32,balanced,0.46433067321777344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,2,8,balanced,1.4000639915466309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,32,balanced,0.6597493489583334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,32,balanced,1.2742186387379963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,2,8,balanced,2.1338507334391275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,2,8,balanced,4.261322657267253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,16,1,power_law_1.01,0.03578239977359772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,16,1,power_law_1.01,0.03774079978466034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,16,1,power_law_1.01,0.03800959885120392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,16,1,power_law_1.01,0.04271360039710999
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,16,1,power_law_1.01,0.04347519874572754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,16,1,power_law_1.01,0.04333440065383911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,16,1,power_law_1.01,0.046112000942230225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,16,1,power_law_1.01,0.0451200008392334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,16,1,power_law_1.01,0.04527359902858734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,16,1,power_law_1.01,0.049619200825691226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,16,1,power_law_1.01,0.049395200610160825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,16,1,power_law_1.01,0.055270397663116456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,16,1,power_law_1.01,0.05620480179786682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,16,1,power_law_1.01,0.061312001943588254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,16,1,power_law_1.01,0.0704576015472412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,16,1,power_law_1.01,0.07369599938392639
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,16,1,power_law_1.01,0.08213120102882385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,16,1,power_law_1.01,0.09742720127105713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,16,1,power_law_1.01,0.10400639772415161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,16,1,power_law_1.01,0.12373759746551513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,16,1,power_law_1.01,0.13805439472198486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,16,1,power_law_1.01,0.18067840337753296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,16,1,power_law_1.01,0.22004480361938478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,16,1,power_law_1.01,0.30709760189056395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,16,1,power_law_1.01,0.3891072034835815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,16,1,power_law_1.01,0.5552512168884277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,64,power_law_1.2,0.036051198840141296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,16,1,power_law_1.01,0.7166592121124268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,16,1,power_law_1.01,0.880025577545166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,64,power_law_1.2,0.028729599714279175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,16,1,power_law_1.01,1.372332763671875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,64,power_law_1.2,0.02887679934501648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,16,1,power_law_1.01,2.67142391204834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,64,power_law_1.2,0.029849600791931153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,64,power_law_1.2,0.02983039915561676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,64,power_law_1.2,0.031711998581886294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,64,power_law_1.2,0.030713599920272828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,64,power_law_1.2,0.030611199140548707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,64,power_law_1.2,0.031040000915527343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,64,power_law_1.2,0.032128000259399415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,64,power_law_1.2,0.03223679959774017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,64,power_law_1.2,0.032927998900413515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,64,power_law_1.2,0.03326080143451691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,64,power_law_1.2,0.03402880132198334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,64,power_law_1.2,0.03733760118484497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,64,power_law_1.2,0.038047999143600464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,64,power_law_1.2,0.03985919952392578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,64,power_law_1.2,0.044838398694992065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,64,power_law_1.2,0.044972801208496095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,64,power_law_1.2,0.053600001335144046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,32,power_law_1.2,0.08413439989089966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,64,power_law_1.2,0.058899199962615965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,64,power_law_1.2,0.07057920098304749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,32,power_law_1.2,0.08086400032043457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,64,power_law_1.2,0.08523520231246948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,32,power_law_1.2,0.05375999808311462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,64,power_law_1.2,0.1309056043624878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,64,power_law_1.2,0.18178559541702272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,32,power_law_1.2,0.05011199712753296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,64,power_law_1.2,0.2341439962387085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,32,power_law_1.2,0.051846402883529666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,64,power_law_1.2,0.3416064023971558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,32,power_law_1.2,0.0531711995601654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,64,power_law_1.2,0.42810878753662107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,32,power_law_1.2,0.056415998935699464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,64,power_law_1.2,0.6833407878875732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,32,power_law_1.2,0.0587007999420166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,32,power_law_1.2,0.06467840075492859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,64,power_law_1.2,1.4409919738769532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,32,power_law_1.2,0.0602944016456604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,32,power_law_1.2,0.06069120168685913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,32,power_law_1.2,0.0662015974521637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,32,power_law_1.2,0.06580479741096497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,32,power_law_1.2,0.06893439888954163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,32,power_law_1.2,0.07560319900512695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,32,power_law_1.2,0.07879679799079894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,32,power_law_1.2,0.08416640162467956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,32,power_law_1.2,0.09234560132026673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,8,balanced,0.052469333012898765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,8,balanced,0.05382933219273885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,8,balanced,0.0633653352657954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,8,balanced,0.0876639982064565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,8,balanced,0.11819733182589214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,32,power_law_1.2,0.10222079753875732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,8,balanced,0.134853333234787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,8,balanced,0.1378506620724996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,8,balanced,0.1360586682955424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,8,balanced,0.13691733280817667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,8,balanced,0.13760000467300415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,8,balanced,0.13846932848294577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,8,balanced,0.13885333140691122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,8,balanced,0.1418719987074534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,8,balanced,0.14520532886187235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,8,balanced,0.1514293352762858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,8,balanced,0.1509813368320465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,8,balanced,0.1569546659787496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,32,power_law_1.2,0.12918399572372435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,8,balanced,0.16285866498947144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,8,balanced,0.17289066314697266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,8,balanced,0.2085813283920288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,8,balanced,0.20731200774510702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,8,balanced,0.2661813298861186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,8,balanced,0.2855093280474345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,8,balanced,0.4042346477508545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,8,balanced,0.4549866517384847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,8,balanced,0.6594026486078898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,32,power_law_1.2,0.13693439960479736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,8,balanced,0.7899946371714274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,8,balanced,0.9171040058135986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,8,balanced,1.4403467178344727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,32,power_law_1.2,0.17571200132369996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,8,balanced,2.7128480275472007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,32,power_law_1.2,0.2133631944656372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,32,power_law_1.2,0.29205760955810545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,32,power_law_1.2,0.3512831926345825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,32,power_law_1.2,0.5525568008422852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,32,power_law_1.2,0.6554624080657959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,32,power_law_1.2,0.7996416091918945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,1,balanced,0.10296533505121867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,1,balanced,0.2005386749903361
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,1,balanced,0.2578773299853007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,1,balanced,0.46379733085632324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,1,balanced,0.879594643910726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,32,power_law_1.2,1.3131839752197265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,1,balanced,1.0887786547342937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,1,balanced,1.0947733720143635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,1,balanced,1.1004213492075603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,1,balanced,1.1060693264007568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,1,balanced,1.1135839621225994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,1,balanced,1.1165599822998047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,1,balanced,1.1224853197733562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,1,balanced,1.127520004908244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,1,balanced,1.1404213110605876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,1,balanced,1.1556426684061687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,1,balanced,1.158255974451701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,1,balanced,1.1798240343729656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,1,balanced,1.2587680021921794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,1,balanced,1.252394676208496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,32,power_law_1.2,2.5833152770996093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,1,balanced,1.4907573064168294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,1,balanced,1.3978986740112305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,1,balanced,2.031856060028076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,1,balanced,1.6955679257710774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,1,balanced,2.5577492713928223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,1,balanced,2.5482613245646157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,1,balanced,4.223685264587402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,1,balanced,4.5469919840494795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,1,balanced,5.5704695383707685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,1,balanced,8.293647766113281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,1,balanced,15.551279703776041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,1,power_law_1.01,0.19946240186691283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,1,power_law_1.01,0.3216383934020996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,1,power_law_1.01,0.4014336109161377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,1,power_law_1.01,0.7956352233886719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,1,power_law_1.01,1.261081600189209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,1,power_law_1.01,2.2085952758789062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,1,power_law_1.01,2.4164287567138674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,1,power_law_1.01,2.503071975708008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,1,power_law_1.01,2.5735231399536134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,1,power_law_1.01,2.6543296813964843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,1,power_law_1.01,2.7342912673950197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,1,power_law_1.01,2.8754112243652346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,1,power_law_1.01,2.847295951843262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,1,power_law_1.01,2.975980758666992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,1,power_law_1.01,3.0692672729492188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,1,power_law_1.01,3.1769855499267576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,1,power_law_1.01,3.338777542114258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,1,power_law_1.01,3.6657665252685545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,1,power_law_1.01,3.6771968841552733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,1,power_law_1.01,4.159231948852539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,1,power_law_1.01,4.4310142517089846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,1,power_law_1.01,5.239910507202149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,1,power_law_1.01,5.226419067382812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,1,power_law_1.01,6.324198532104492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,1,power_law_1.01,8.491187286376952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,1,power_law_1.01,10.967680358886719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,1,power_law_1.01,11.949388885498047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,1,1,power_law_1.01,14.284275817871094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,1,1,power_law_1.01,21.211085510253906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,1,1,power_law_1.01,39.088916015625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,2,balanced,0.05609600245952606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,2,balanced,0.06031466523806254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,2,balanced,0.07470400134722392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,2,balanced,0.09888533751169841
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,2,balanced,0.13809067010879517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,2,balanced,0.1586026648680369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,2,balanced,0.16126400232315063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,2,balanced,0.1630773345629374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,2,balanced,0.16564266880353293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,2,balanced,0.1666719913482666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,2,balanced,0.16833599408467612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,2,balanced,0.17080533504486084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,2,balanced,0.17421332995096842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,2,balanced,0.1793760061264038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,2,balanced,0.18942934274673462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,2,balanced,0.19118932882944742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,2,balanced,0.1964319944381714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,2,balanced,0.21942400932312012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,2,balanced,0.23696533838907877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,2,balanced,0.28195200363794964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,2,balanced,0.3111039996147156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,2,balanced,0.43192001183827716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,2,balanced,0.5005280176798502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,2,balanced,0.7272960344950358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,2,balanced,0.890618642171224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,2,balanced,1.306991974512736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,2,balanced,1.6386879285176594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,2,balanced,1.9414666493733723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,2,balanced,3.046794573465983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,2,balanced,5.924832026163737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,8,power_law_1.2,0.050342398881912234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,8,power_law_1.2,0.04791040122509003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,8,power_law_1.2,0.04830079972743988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,8,power_law_1.2,0.05241600275039673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,8,power_law_1.2,0.06231679916381836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,8,power_law_1.2,0.07404159903526306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,8,power_law_1.2,0.07984640002250672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,8,power_law_1.2,0.08232319951057435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,8,power_law_1.2,0.08228480219841003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,8,power_law_1.2,0.08284800052642823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,8,power_law_1.2,0.08651520013809204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,8,power_law_1.2,0.08828160166740417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,8,power_law_1.2,0.09205759763717651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,8,power_law_1.2,0.09668480157852173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,8,power_law_1.2,0.10153599977493286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,8,power_law_1.2,0.10677759647369385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,8,power_law_1.2,0.11177599430084229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,8,power_law_1.2,0.12873599529266358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,8,power_law_1.2,0.13689600229263305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,8,power_law_1.2,0.17818880081176758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,8,power_law_1.2,0.21461760997772217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,8,power_law_1.2,0.2671168088912964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,8,power_law_1.2,0.349619197845459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,8,power_law_1.2,0.4786816120147705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,8,power_law_1.2,0.5499328136444092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,8,power_law_1.2,0.8582400321960449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,8,power_law_1.2,0.9659839630126953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,8,power_law_1.2,1.3803135871887207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,8,power_law_1.2,2.1132671356201174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,8,power_law_1.2,5.734643173217774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,8,balanced,0.036176001032193504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,8,balanced,0.03595199932654699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,8,balanced,0.03612266729275385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,8,balanced,0.04009066770474116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,8,balanced,0.044112001856168113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,8,balanced,0.04419200122356415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,8,balanced,0.04599999884764353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,8,balanced,0.04608533283074697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,8,balanced,0.04637866715590159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,8,balanced,0.04542933404445648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,8,balanced,0.04612799982229868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,8,balanced,0.04807466765244802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,8,balanced,0.04780266682306925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,8,balanced,0.05026133358478546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,8,balanced,0.05194666484991709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,8,balanced,0.05401599903901418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,8,balanced,0.056330665946006775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,8,balanced,0.06028800209363302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,8,balanced,0.06648000081380208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,8,balanced,0.07684266567230225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,8,balanced,0.0827466646830241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,8,balanced,0.11158933242162068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,8,balanced,0.13597866892814636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,8,balanced,0.17376534144083658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,8,balanced,0.20981866121292114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,8,balanced,0.290608008702596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,8,balanced,0.37009600798288983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,8,balanced,0.4501760005950928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,8,balanced,0.6903306643168131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,8,balanced,1.3207039833068848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,2,balanced,0.05468800167242686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,2,balanced,0.05845866600672404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,2,balanced,0.07349333167076111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,2,balanced,0.10090667009353638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,2,balanced,0.15308266878128052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,2,balanced,0.2500426570574443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,2,balanced,0.35414934158325195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,2,balanced,0.3540106614430745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,2,balanced,0.3601280053456624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,2,balanced,0.35710398356119794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,2,balanced,0.3613813320795695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,2,balanced,0.3733760118484497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,2,balanced,0.3739200035730998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,2,balanced,0.37763198216756183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,2,balanced,0.3838133414586385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,2,balanced,0.3856000105539958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,2,balanced,0.4047413269678752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,2,balanced,0.4230186541875203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,2,balanced,0.44680531819661456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,2,balanced,0.49159467220306396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,2,balanced,0.5275573333104452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,2,balanced,0.606607993443807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,2,balanced,0.7130453586578369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,2,balanced,0.8928000132242838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,2,balanced,1.1245546340942383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,2,balanced,1.4500586191813152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,2,balanced,1.9162294069925945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,2,balanced,2.404362678527832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,2,balanced,3.4414027531941733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,2,balanced,6.5153656005859375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,32,power_law_1.2,0.05504000186920166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,32,power_law_1.2,0.05301759839057922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,32,power_law_1.2,0.043609601259231565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,32,power_law_1.2,0.046342399716377256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,32,power_law_1.2,0.048742398619651794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,32,power_law_1.2,0.04725759923458099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,32,power_law_1.2,0.05119360089302063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,32,power_law_1.2,0.0531328022480011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,32,power_law_1.2,0.05380480289459229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,32,power_law_1.2,0.05580800175666809
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,32,power_law_1.2,0.05681920051574707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,32,power_law_1.2,0.06115840077400207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,32,power_law_1.2,0.06218879818916321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,32,power_law_1.2,0.06583679914474487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,32,power_law_1.2,0.07163519859313965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,32,power_law_1.2,0.07594239711761475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,32,power_law_1.2,0.08536319732666016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,32,power_law_1.2,0.09863680005073547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,32,power_law_1.2,0.10882560014724732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,32,power_law_1.2,0.12912000417709352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,32,power_law_1.2,0.15121920108795167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,32,power_law_1.2,0.2007040023803711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,32,power_law_1.2,0.24726400375366211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,32,power_law_1.2,0.3320512056350708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,32,power_law_1.2,0.47258877754211426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,32,power_law_1.2,0.6853248119354248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,32,power_law_1.2,0.8843647956848144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,32,power_law_1.2,0.9494720458984375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,32,power_law_1.2,1.6724159240722656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,32,power_law_1.2,2.9909055709838865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,1,power_law_1.2,0.048198398947715757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,1,power_law_1.2,0.07548800110816956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,1,power_law_1.2,0.081523197889328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,1,power_law_1.2,0.11018240451812744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,1,power_law_1.2,0.14871679544448851
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,1,power_law_1.2,0.2535871982574463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,1,power_law_1.2,0.2630784034729004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,1,power_law_1.2,0.2742016077041626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,1,power_law_1.2,0.28789119720458983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,1,power_law_1.2,0.2907776117324829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,1,power_law_1.2,0.3030143976211548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,1,power_law_1.2,0.3092479944229126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,1,power_law_1.2,0.3150399923324585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,1,power_law_1.2,0.33345279693603513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,1,power_law_1.2,0.34815359115600586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,1,power_law_1.2,0.35797119140625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,1,power_law_1.2,0.38353919982910156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,1,power_law_1.2,0.4312960147857666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,1,power_law_1.2,0.4654719829559326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,1,power_law_1.2,0.5549503803253174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,1,power_law_1.2,0.669209623336792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,1,power_law_1.2,0.8218496322631836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,1,power_law_1.2,1.0521727561950684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,1,power_law_1.2,1.3460543632507325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,1,power_law_1.2,1.6956352233886718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,1,power_law_1.2,2.2700544357299806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,1,power_law_1.2,2.774060821533203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,1,power_law_1.2,3.356390380859375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,1,power_law_1.2,5.572409439086914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,1,power_law_1.2,9.499314880371093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,4,power_law_1.2,0.04062080085277557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,4,power_law_1.2,0.03384959995746613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,4,power_law_1.2,0.0351936012506485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,4,power_law_1.2,0.03845120072364807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,4,power_law_1.2,0.04149119853973389
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,4,power_law_1.2,0.0497408002614975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,4,power_law_1.2,0.0492031991481781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,4,power_law_1.2,0.05069440007209778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,4,power_law_1.2,0.05061119794845581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,4,power_law_1.2,0.05074560046195984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,4,power_law_1.2,0.051948797702789304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,4,power_law_1.2,0.05348479747772217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,4,power_law_1.2,0.05633280277252197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,4,power_law_1.2,0.057843202352523805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,4,power_law_1.2,0.06365439891815186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,4,power_law_1.2,0.06562560200691223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,4,power_law_1.2,0.06872959733009339
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,4,power_law_1.2,0.07932159900665284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,4,power_law_1.2,0.0917568027973175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,4,power_law_1.2,0.10774400234222412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,4,power_law_1.2,0.12632319927215577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,4,power_law_1.2,0.16238720417022706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,4,power_law_1.2,0.18926719427108765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,4,power_law_1.2,0.2715712070465088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,4,power_law_1.2,0.3760319948196411
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,4,power_law_1.2,0.5515520095825195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,4,power_law_1.2,0.7034495830535888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,4,power_law_1.2,0.9261504173278808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,4,power_law_1.2,1.6213567733764649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,4,power_law_1.2,2.998905563354492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,32,1,power_law_1.2,0.04916479885578155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,32,1,power_law_1.2,0.05141760110855102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,32,1,power_law_1.2,0.059462398290634155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,32,1,power_law_1.2,0.07339519858360291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,32,1,power_law_1.2,0.08943359851837158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,32,1,power_law_1.2,0.1096384048461914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,2,power_law_1.2,0.03288320004940033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,2,power_law_1.2,0.040992000699043275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,32,1,power_law_1.2,0.1401087999343872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,2,power_law_1.2,0.04164479970932007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,2,power_law_1.2,0.0548799991607666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,32,1,power_law_1.2,0.14764800071716308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,2,power_law_1.2,0.06559360027313232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,32,1,power_law_1.2,0.1519487977027893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,2,power_law_1.2,0.0987712025642395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,2,power_law_1.2,0.1096384048461914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,32,1,power_law_1.2,0.1575551986694336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,2,power_law_1.2,0.11114879846572875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,2,power_law_1.2,0.11530879735946656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,32,1,power_law_1.2,0.16892800331115723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,2,power_law_1.2,0.11546239852905274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,2,power_law_1.2,0.11866240501403809
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,32,1,power_law_1.2,0.18526079654693603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,2,power_law_1.2,0.12347520589828491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,32,1,power_law_1.2,0.19230079650878906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,2,power_law_1.2,0.12288639545440674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,2,power_law_1.2,0.13228800296783447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,32,1,power_law_1.2,0.2092479944229126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,2,power_law_1.2,0.13703680038452148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,32,1,power_law_1.2,0.22736001014709473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,2,power_law_1.2,0.14895999431610107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,2,power_law_1.2,0.14599039554595947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,8,balanced,0.04391466577847799
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,8,balanced,0.04598933458328247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,32,1,power_law_1.2,0.23853440284729005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,8,balanced,0.0470719983180364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,8,balanced,0.05142400165398916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,8,balanced,0.0724533349275589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,8,balanced,0.09981866677602132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,2,power_law_1.2,0.1531008005142212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,8,balanced,0.10201600193977356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,8,balanced,0.10371733705202739
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,8,balanced,0.10609599947929382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,8,balanced,0.1056106686592102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,8,balanced,0.10563733180363973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,8,balanced,0.10573333501815796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,8,balanced,0.10942932963371277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,8,balanced,0.10753066341082256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,8,balanced,0.11474666992823283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,8,balanced,0.1158026655515035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,2,power_law_1.2,0.17222399711608888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,8,balanced,0.11868266264597575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,32,1,power_law_1.2,0.23877758979797364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,8,balanced,0.12818132837613425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,8,balanced,0.13683733344078064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,8,balanced,0.15264000495274863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,8,balanced,0.1698240041732788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,8,balanced,0.20188266038894653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,8,balanced,0.23453332980473837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,2,power_law_1.2,0.19640320539474487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,8,balanced,0.299125333627065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,8,balanced,0.3678026596705119
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,8,balanced,0.5051413377126058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,8,balanced,0.6306080023447672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,32,1,power_law_1.2,0.28976640701293943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,2,power_law_1.2,0.20864639282226563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,8,balanced,0.7636853059132894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,8,balanced,1.1423306465148926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,2,power_law_1.2,0.25271680355072024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,8,balanced,2.188389301300049
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,32,1,power_law_1.2,0.3332864046096802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,2,power_law_1.2,0.2891455888748169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,32,1,power_law_1.2,0.4193215847015381
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,2,power_law_1.2,0.37073919773101804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,2,power_law_1.2,0.44878082275390624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,32,1,power_law_1.2,0.49646720886230467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,2,power_law_1.2,0.6008255958557129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,32,1,power_law_1.2,0.6565311908721924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,2,power_law_1.2,0.7810751914978027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,32,1,power_law_1.2,0.8245823860168457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,1,2,power_law_1.2,0.9284416198730469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,32,1,power_law_1.2,1.137228775024414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,1,2,power_law_1.2,1.5163328170776367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,32,1,power_law_1.2,1.449184036254883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,32,1,power_law_1.2,2.067545509338379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,1,2,power_law_1.2,2.633644866943359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,32,1,power_law_1.2,2.720403289794922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,32,1,power_law_1.2,3.35533447265625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,4,power_law_1.2,0.10337920188903808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,4,power_law_1.2,0.10131839513778687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,4,power_law_1.2,0.09513599872589111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,4,power_law_1.2,0.14005119800567628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,32,1,power_law_1.2,5.238022232055664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,4,power_law_1.2,0.14759680032730102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,4,power_law_1.2,0.24534399509429933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,4,power_law_1.2,0.2640127897262573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,4,power_law_1.2,0.26897919178009033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,4,power_law_1.2,0.2799936056137085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,4,power_law_1.2,0.28494079113006593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,4,power_law_1.2,0.30405120849609374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,32,1,power_law_1.2,10.24642562866211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,4,power_law_1.2,0.30154879093170167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,4,power_law_1.2,0.3147711992263794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,8,power_law_1.01,0.09627519845962525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,4,power_law_1.2,0.30599040985107423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,4,power_law_1.2,0.30971519947052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,8,power_law_1.01,0.14495999813079835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,4,power_law_1.2,0.32895359992980955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,4,power_law_1.2,0.34911360740661623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,8,power_law_1.01,0.24242560863494872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,4,power_law_1.2,0.3944704055786133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,8,power_law_1.01,0.09697279930114747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,4,power_law_1.2,0.4352255821228027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,4,power_law_1.2,0.5040768146514892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,8,power_law_1.01,0.13611520528793336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,4,power_law_1.2,0.5602560043334961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,4,power_law_1.2,0.6978367805480957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,8,power_law_1.01,0.1871999979019165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,4,power_law_1.2,0.687065601348877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,8,power_law_1.01,0.191539204120636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,4,power_law_1.2,0.8894911766052246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,8,power_law_1.01,0.23281280994415282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,4,power_law_1.2,1.0353407859802246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,4,balanced,0.031514666974544525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,4,balanced,0.03200533241033554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,4,balanced,0.03349333256483078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,4,balanced,0.037717332442601524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,4,balanced,0.05941333373387655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,8,power_law_1.01,0.24876799583435058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,4,balanced,0.059061333537101746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,4,power_law_1.2,1.5007616043090821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,4,balanced,0.059989333152770996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,4,balanced,0.06046399970849355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,4,balanced,0.058650667468706764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,4,balanced,0.06007466713587443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,4,balanced,0.0610346645116806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,4,balanced,0.061717331409454346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,4,balanced,0.06282666822274525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,4,balanced,0.06454933186372121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,4,balanced,0.07026133437951405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,4,balanced,0.0730453332265218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,4,balanced,0.07366399963696797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,4,balanced,0.07923733194669087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,4,balanced,0.08470933636029561
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,4,balanced,0.10234133402506511
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,4,balanced,0.10869866609573364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,4,balanced,0.13729600111643472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,4,balanced,0.16674133141835532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,4,balanced,0.23014400402704874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,4,balanced,0.27504533529281616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,4,balanced,0.38334401448567706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,8,power_law_1.01,0.2749376058578491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,4,balanced,0.49677332242329914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,4,balanced,0.6031733353932699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,4,power_law_1.2,2.1725696563720702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,4,balanced,0.9094026883443197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,4,balanced,1.7237226168314617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,8,power_law_1.01,0.2679296016693115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,4,power_law_1.2,2.5977151870727537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,8,power_law_1.01,0.27174398899078367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,8,power_law_1.01,0.28378241062164306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,4,power_law_1.2,3.8716289520263674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,8,power_law_1.01,0.29091200828552244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,8,power_law_1.01,0.31997439861297605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,8,power_law_1.01,0.31088640689849856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,4,power_law_1.2,5.5512641906738285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,8,power_law_1.01,0.3138495922088623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,8,power_law_1.01,0.36326398849487307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,8,power_law_1.01,0.3940160036087036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,8,power_law_1.01,0.44430079460144045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,1,power_law_1.2,0.05229439735412598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,8,power_law_1.01,0.47363839149475095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,1,power_law_1.2,0.05808640122413635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,1,power_law_1.2,0.07130240201950074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,8,power_law_1.01,0.5558335781097412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,1,power_law_1.2,0.09336959719657897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,8,power_law_1.01,0.6732800006866455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,1,power_law_1.2,0.13141119480133057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,8,power_law_1.01,0.8269120216369629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,1,power_law_1.2,0.17285120487213135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,8,power_law_1.01,0.9235199928283692
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,1,power_law_1.2,0.19567359685897828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,1,power_law_1.2,0.25340158939361573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,8,power_law_1.01,1.2275775909423827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,1,power_law_1.2,0.27013120651245115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,8,power_law_1.01,1.551974391937256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,1,power_law_1.2,0.2742144107818604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,1,power_law_1.2,0.29656960964202883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,8,power_law_1.01,1.932863998413086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,1,power_law_1.2,0.3137792110443115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,1,power_law_1.2,0.32478721141815187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,8,power_law_1.01,2.8821695327758787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,1,power_law_1.2,0.35350399017333983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,1,power_law_1.2,0.3747136116027832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,1,power_law_1.2,0.3920703887939453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,8,power_law_1.01,5.375455856323242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,1,power_law_1.2,0.43637762069702146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,1,power_law_1.2,0.5245440006256104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,1,power_law_1.2,0.5322495937347412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,1,power_law_1.2,0.64334077835083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,1,power_law_1.2,0.7327231884002685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,1,power_law_1.2,0.9279040336608887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,1,power_law_1.2,1.1350272178649903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,1,power_law_1.2,1.5040575981140136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,1,power_law_1.2,1.9399423599243164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,1,power_law_1.2,0.09964159727096558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,1,power_law_1.2,0.13571200370788575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,1,power_law_1.2,0.18903039693832396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,1,power_law_1.2,2.624095916748047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,1,power_law_1.2,0.29173760414123534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,1,power_law_1.2,0.40019841194152833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,1,power_law_1.2,0.8051072120666504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,1,power_law_1.2,0.8627584457397461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,1,power_law_1.2,3.209408187866211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,1,power_law_1.2,0.931942367553711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,1,power_law_1.2,0.9697343826293945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,1,power_law_1.2,0.9901568412780761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,1,power_law_1.2,3.9209598541259765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,1,power_law_1.2,1.0033599853515625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,1,power_law_1.2,1.051846408843994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,1,power_law_1.2,1.068832015991211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,1,power_law_1.2,1.0912639617919921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,1,power_law_1.2,6.217536163330078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,1,power_law_1.2,1.1407423973083497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,1,power_law_1.2,1.1715007781982423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,1,power_law_1.2,1.205254364013672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,1,power_law_1.2,1.3185664176940919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,1,power_law_1.2,1.3599488258361816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,1,power_law_1.2,1.5343232154846191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,1,power_law_1.2,11.806047821044922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,1,power_law_1.2,1.730508804321289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,1,power_law_1.2,2.073036766052246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,1,power_law_1.2,2.122444725036621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,1,power_law_1.2,2.5817407608032226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,1,power_law_1.2,3.413792037963867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,1,power_law_1.2,4.277228927612304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,1,power_law_1.2,5.1875457763671875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,64,balanced,0.03252266595760981
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,64,balanced,0.032069332897663116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,1,power_law_1.2,6.1954303741455075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,64,balanced,0.03155199935038885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,64,balanced,0.03250666707754135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,64,balanced,0.033557333052158356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,64,balanced,0.03369600077470144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,64,balanced,0.033973333736260734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,64,balanced,0.03385599950949351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,64,balanced,0.03588266670703888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,64,balanced,0.03435199956099192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,64,balanced,0.035775999228159584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,64,balanced,0.035818666219711304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,64,balanced,0.03759466608365377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,64,balanced,0.03589866558710734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,64,balanced,0.04176533222198486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,64,balanced,0.041509332756201424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,64,balanced,0.04177066683769226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,64,balanced,0.04401599864164988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,64,balanced,0.045754666129748024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,64,balanced,0.04816000163555145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,64,balanced,0.05249066650867462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,64,balanced,0.06863999863465627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,64,balanced,0.0865760048230489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,64,balanced,0.11663466691970825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,64,balanced,0.14012266198794046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,64,balanced,0.1884373426437378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,64,balanced,0.24021865924199423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,64,balanced,0.29285333553949994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,64,balanced,0.42079468568166095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,64,balanced,0.7895466486612955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,1,power_law_1.2,8.89599380493164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,32,balanced,0.04635733366012573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,32,balanced,0.05194133520126343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,32,balanced,0.05067199965318044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,32,balanced,0.056186666091283165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,32,balanced,0.06052266558011373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,32,balanced,0.060906668504079185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,32,balanced,0.06136000156402588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,32,balanced,0.062463998794555664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,32,balanced,0.062447999914487205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,32,balanced,0.06253866851329803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,32,balanced,0.06196799874305725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,32,balanced,0.06533333162466685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,32,balanced,0.0647680014371872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,32,balanced,0.06564799944559734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,32,balanced,0.0691840002934138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,32,balanced,0.0691893349091212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,32,balanced,0.07551999886830647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,32,balanced,0.09380267063776652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,32,balanced,0.09161600470542908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,32,balanced,0.12145599722862244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,32,balanced,0.1362986663977305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,32,balanced,0.1812266707420349
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,32,balanced,0.16800000270207724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,32,balanced,0.21501866976420084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,32,balanced,0.24943999449412027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,1,power_law_1.2,16.26666259765625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,32,balanced,0.3338079849878947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,32,balanced,0.3943946758906047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,8,32,balanced,0.4840159813563029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,8,32,balanced,0.696938673655192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,8,32,balanced,1.49346129099528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,8,balanced,0.04801600178082784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,8,balanced,0.060191998879114784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,8,balanced,0.07044800122578938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,8,balanced,0.09619200229644775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,8,balanced,0.12930666406949362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,8,balanced,0.15457066893577576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,8,balanced,0.15493333339691162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,8,balanced,0.1588266690572103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,8,balanced,0.1588640014330546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,8,balanced,0.16024000446001688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,8,balanced,0.1601706643899282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,8,balanced,0.16113600134849548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,8,balanced,0.16287466883659363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,8,balanced,0.16485333442687988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,8,balanced,0.167738676071167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,8,balanced,0.17270400126775107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,8,balanced,0.17850132783253989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,8,balanced,0.19800533850987753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,8,balanced,0.2042133410771688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,8,balanced,0.2672106623649597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,8,balanced,0.26530667146046955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,8,balanced,0.3454986810684204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,8,balanced,0.3454666535059611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,8,balanced,0.47755734125773114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,8,balanced,0.5354719956715902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,8,balanced,0.7650079727172852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,8,balanced,0.887989362080892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,4,8,balanced,1.1218773523966472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,4,8,balanced,1.6699306170145671
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,4,8,balanced,3.250981330871582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,32,1,power_law_1.01,0.04730879962444305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,32,1,power_law_1.01,0.05111680030822754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,1,power_law_1.01,0.04992640018463135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,32,1,power_law_1.01,0.05774719715118408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,32,1,power_law_1.01,0.07305600047111512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,1,power_law_1.01,0.05251839756965637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,32,1,power_law_1.01,0.09704959988594056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,1,power_law_1.01,0.06314240097999572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,32,1,power_law_1.01,0.11361919641494751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,1,power_law_1.01,0.07511680126190186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,32,1,power_law_1.01,0.1404863953590393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,1,power_law_1.01,0.08804479837417603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,32,1,power_law_1.01,0.14670079946517944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,1,power_law_1.01,0.1209663987159729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,32,1,power_law_1.01,0.15241600275039674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,1,power_law_1.01,0.13185280561447144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,32,1,power_law_1.01,0.1578879952430725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,1,power_law_1.01,0.1354367971420288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,128,power_law_1.01,0.0661184012889862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,32,1,power_law_1.01,0.1697216033935547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,1,power_law_1.01,0.1382591962814331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,32,1,power_law_1.01,0.17970559597015381
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,1,power_law_1.01,0.13901439905166627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,128,power_law_1.01,0.06992639899253845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,32,1,power_law_1.01,0.19022079706192016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,1,power_law_1.01,0.1428607940673828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,32,1,power_law_1.01,0.20470399856567384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,1,power_law_1.01,0.14559359550476075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,128,power_law_1.01,0.05546240210533142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,32,1,power_law_1.01,0.22283520698547363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,1,power_law_1.01,0.14872959852218628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,32,1,power_law_1.01,0.23594241142272948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,1,power_law_1.01,0.15665919780731202
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,128,power_law_1.01,0.05839999914169312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,1,power_law_1.01,0.16899839639663697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,32,1,power_law_1.01,0.23104639053344728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,1,power_law_1.01,0.17235200405120848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,32,1,power_law_1.01,0.28361599445343016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,128,power_law_1.01,0.055769598484039305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,1,power_law_1.01,0.18769279718399048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,32,1,power_law_1.01,0.31848320960998533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,1,power_law_1.01,0.22387840747833251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,128,power_law_1.01,0.057657599449157715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,32,1,power_law_1.01,0.41245441436767577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,1,power_law_1.01,0.2539072036743164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,32,1,power_law_1.01,0.48492159843444826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,128,power_law_1.01,0.0582144021987915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,1,power_law_1.01,0.31969280242919923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,32,1,power_law_1.01,0.6543360233306885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,1,power_law_1.01,0.3781888008117676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,128,power_law_1.01,0.05829120278358459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,1,power_law_1.01,0.5148543834686279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,32,1,power_law_1.01,0.8197504043579101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,1,power_law_1.01,0.6440896034240723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,128,power_law_1.01,0.05878400206565857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,32,1,power_law_1.01,1.136691188812256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,1,power_law_1.01,0.9050111770629883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,32,1,power_law_1.01,1.4453184127807617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,128,power_law_1.01,0.059520000219345094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,1,power_law_1.01,1.1558783531188965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,32,1,power_law_1.01,2.0681472778320313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,128,power_law_1.01,0.0593280017375946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,1,power_law_1.01,1.6700864791870118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,128,power_law_1.01,0.0627135992050171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,32,1,power_law_1.01,2.717356872558594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,1,power_law_1.01,2.1748672485351563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,128,power_law_1.01,0.06294400095939637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,32,1,power_law_1.01,3.344678497314453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,1,power_law_1.01,2.6985471725463865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,128,power_law_1.01,0.06345599889755249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,128,power_law_1.01,0.0699455976486206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,1,power_law_1.01,4.167647933959961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,32,1,power_law_1.01,5.237510299682617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,128,power_law_1.01,0.07297279834747314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,128,power_law_1.01,0.0789247989654541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,1,power_law_1.01,8.202559661865234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,128,power_law_1.01,0.09303680062294006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,32,1,power_law_1.01,10.232115173339844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,128,power_law_1.01,0.10366719961166382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,128,power_law_1.01,0.13121919631958007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,128,power_law_1.01,0.14848639965057372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,128,power_law_1.01,0.19470080137252807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,128,power_law_1.01,0.1934847950935364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,128,power_law_1.01,0.2605695962905884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,128,power_law_1.2,0.07731199860572815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,128,power_law_1.01,0.3074687957763672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,8,power_law_1.01,0.11947519779205322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,128,power_law_1.2,0.0744704008102417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,8,power_law_1.01,0.178547203540802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,128,power_law_1.01,0.457747220993042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,128,power_law_1.2,0.06870399713516236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,8,power_law_1.01,0.0819711983203888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,128,power_law_1.2,0.06629120111465454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,128,power_law_1.01,0.5486464023590087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,8,power_law_1.01,0.11028480529785156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,128,power_law_1.2,0.06938880085945129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,8,power_law_1.01,0.13038079738616942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,128,power_law_1.2,0.07396479845046997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,8,power_law_1.01,0.19482879638671874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,1,128,power_law_1.01,0.6669951915740967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,128,power_law_1.2,0.07324159741401673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,8,power_law_1.01,0.2065216064453125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,128,power_law_1.2,0.07239680290222168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,8,power_law_1.01,0.2198335886001587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,128,power_law_1.2,0.07232000231742859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,1,128,power_law_1.01,1.0331520080566405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,8,power_law_1.01,0.22657279968261718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,128,power_law_1.2,0.07474560141563416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,8,power_law_1.01,0.22542719841003417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,128,power_law_1.2,0.07393919825553893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,8,power_law_1.01,0.22597761154174806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,128,power_law_1.2,0.0738048017024994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,8,power_law_1.01,0.23423359394073487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,128,power_law_1.2,0.07749760150909424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,128,power_law_1.2,0.0765504002571106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,8,power_law_1.01,0.23268480300903321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,1,128,power_law_1.01,2.11144962310791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,128,power_law_1.2,0.08300160169601441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,8,power_law_1.01,0.24600958824157715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,128,power_law_1.2,0.08392959833145142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,8,power_law_1.01,0.25540480613708494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,128,power_law_1.2,0.0907904028892517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,8,power_law_1.01,0.262393593788147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,128,power_law_1.2,0.10533119440078735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,8,power_law_1.01,0.2829440116882324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,128,power_law_1.2,0.125164794921875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,128,power_law_1.2,0.15030399560928345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,8,power_law_1.01,0.3229504108428955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,128,power_law_1.2,0.1823359966278076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,8,power_law_1.01,0.34443519115447996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,128,power_law_1.2,0.2267008066177368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,8,power_law_1.01,0.3990976095199585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,128,power_law_1.2,0.2817280054092407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,8,power_law_1.01,0.4467904090881348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,128,power_law_1.2,0.4097343921661377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,128,power_law_1.2,0.5185855865478516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,8,power_law_1.01,0.5685247898101806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,16,8,power_law_1.2,0.03081600069999695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,16,8,power_law_1.2,0.029843199253082275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,16,8,power_law_1.2,0.029420799016952513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,16,8,power_law_1.2,0.029625600576400755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,128,power_law_1.2,0.8112447738647461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,16,8,power_law_1.2,0.02969599962234497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,16,8,power_law_1.2,0.030976000428199767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,16,8,power_law_1.2,0.031513598561286923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,8,power_law_1.01,0.6187839984893799
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,16,8,power_law_1.2,0.031001600623130798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,16,8,power_law_1.2,0.03192319869995117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,16,8,power_law_1.2,0.03185279965400696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,16,8,power_law_1.2,0.03389439880847931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,16,8,power_law_1.2,0.035174399614334106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,16,8,power_law_1.2,0.036051198840141296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,16,8,power_law_1.2,0.035148799419403076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,128,power_law_1.2,1.0335040092468262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,16,8,power_law_1.2,0.04276480078697205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,16,8,power_law_1.2,0.045203199982643126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,16,8,power_law_1.2,0.046803200244903566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,16,8,power_law_1.2,0.053651201725006106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,8,power_law_1.01,0.7691391944885254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,16,8,power_law_1.2,0.05774719715118408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,16,8,power_law_1.2,0.06975359916687011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,16,8,power_law_1.2,0.08319360017776489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,16,8,power_law_1.2,0.12151679992675782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,16,8,power_law_1.2,0.13156479597091675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,128,power_law_1.2,1.2498944282531739
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,16,8,power_law_1.2,0.20129919052124023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,16,8,power_law_1.2,0.24982399940490724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,16,8,power_law_1.2,0.3603264093399048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,16,8,power_law_1.2,0.46124801635742185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,8,power_law_1.01,0.8717887878417969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,16,8,power_law_1.2,0.5704448223114014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,16,8,power_law_1.2,0.8508352279663086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,128,power_law_1.2,2.242214393615723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,16,8,power_law_1.2,1.6401599884033202
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,8,power_law_1.01,1.1693504333496094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,128,power_law_1.2,4.523852920532226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,8,power_law_1.01,1.4070400238037108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,2,8,power_law_1.01,1.7011903762817382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,2,balanced,0.03653866549332937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,2,balanced,0.037952000896135964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,2,balanced,0.058330665032068886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,2,balanced,0.08066133161385854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,2,balanced,0.126309335231781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,2,balanced,0.12865066528320312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,2,balanced,0.12664000193277994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,2,balanced,0.12879467010498047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,2,balanced,0.12870933612187704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,2,balanced,0.13040000200271606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,2,balanced,0.13315199812253317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,2,balanced,0.13473066687583923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,2,balanced,0.13691733280817667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,2,balanced,0.13845866918563843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,2,balanced,0.1479680041472117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,2,balanced,0.1525173286596934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,2,balanced,0.15921066204706827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,2,balanced,0.1738026738166809
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,2,8,power_law_1.01,2.4744192123413087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,2,balanced,0.1893226703008016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,2,balanced,0.2321173350016276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,2,balanced,0.2516160011291504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,2,balanced,0.33030933141708374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,2,balanced,0.39048532644907635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,2,balanced,0.5528426567713419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,2,balanced,0.687455972035726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,2,balanced,0.9781813621520996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,2,balanced,1.2654773394266765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,2,balanced,1.5596639315287273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,2,balanced,2.364581267038981
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,2,balanced,4.617018699645996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,32,balanced,0.052704001466433205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,32,balanced,0.05005866785844167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,32,balanced,0.04985066751639048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,32,balanced,0.04974933465321859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,32,balanced,0.05209066470464071
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,32,balanced,0.053823997577031456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,32,balanced,0.0547626664241155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,32,balanced,0.05574400226275126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,32,balanced,0.0563679983218511
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,32,balanced,0.054192001620928444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,32,balanced,0.057861333092053734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,32,balanced,0.05796800057093302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,32,balanced,0.05750933289527893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,32,balanced,0.06044266621271769
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,32,balanced,0.0643039991458257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,32,balanced,0.064560001095136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,32,balanced,0.06635199983914693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,32,balanced,0.07064533233642578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,32,balanced,0.07795199751853943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,32,balanced,0.09874666730562846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,32,balanced,0.1090773344039917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,2,8,power_law_1.01,4.9777473449707035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,32,balanced,0.13500799735387167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,32,balanced,0.14730667074521384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,32,balanced,0.18881599108378092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,32,balanced,0.21772799889246622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,32,balanced,0.29476267099380493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,32,balanced,0.3662399848302205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,32,balanced,0.4328266779581706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,32,balanced,0.6525919834772745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,32,balanced,1.2525333563486736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,128,power_law_1.01,0.03573119938373566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,128,power_law_1.01,0.0354559987783432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,128,power_law_1.01,0.034771201014518735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,128,power_law_1.01,0.035174399614334106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,128,power_law_1.01,0.035820800065994265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,128,power_law_1.01,0.036364799737930296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,128,power_law_1.01,0.03681280016899109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,128,power_law_1.01,0.03648639917373657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,128,power_law_1.01,0.03731200098991394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,128,power_law_1.01,0.0369024008512497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,128,power_law_1.01,0.037062400579452516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,128,power_law_1.01,0.03771519958972931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,128,power_law_1.01,0.037785598635673524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,8,power_law_1.2,0.085452800989151
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,128,power_law_1.01,0.038649600744247434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,1,power_law_1.2,0.09152640104293823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,128,power_law_1.01,0.04149119853973389
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,8,power_law_1.2,0.1092352032661438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,1,power_law_1.2,0.11573760509490967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,128,power_law_1.01,0.041280001401901245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,128,power_law_1.01,0.04142079949378967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,1,power_law_1.2,0.17056000232696533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,8,power_law_1.2,0.06403840184211732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,128,power_law_1.01,0.044275200366973876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,1,power_law_1.2,0.2376255989074707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,8,power_law_1.2,0.07828480005264282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,128,power_law_1.01,0.04645119905471802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,1,power_law_1.2,0.3242624044418335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,128,power_law_1.01,0.049958398938179015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,1,power_law_1.2,0.631328010559082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,8,power_law_1.2,0.09805439710617066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,128,power_law_1.01,0.05413119792938233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,1,power_law_1.2,0.6803647994995117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,128,power_law_1.01,0.06232960224151611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,8,power_law_1.2,0.10568959712982177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,1,power_law_1.2,0.7108223915100098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,128,power_law_1.01,0.07099519968032837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,8,power_law_1.2,0.13886719942092896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,1,power_law_1.2,0.7177919864654541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,128,power_law_1.01,0.09293439984321594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,1,power_law_1.2,0.7269055843353271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,8,power_law_1.2,0.14666880369186402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,128,power_law_1.01,0.11104639768600463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,1,power_law_1.2,0.7411136150360107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,8,power_law_1.2,0.15215359926223754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,128,power_law_1.01,0.149510395526886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,1,power_law_1.2,0.7512639999389649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,8,power_law_1.2,0.15782400369644164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,1,power_law_1.2,0.7626688003540039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,128,power_law_1.01,0.17785600423812867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,1,power_law_1.2,0.7569920063018799
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,8,power_law_1.2,0.16005120277404786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,128,power_law_1.01,0.22139520645141603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,1,power_law_1.2,0.8037887573242187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,8,power_law_1.2,0.15960320234298705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,1,power_law_1.2,0.8244735717773437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,8,power_law_1.2,0.1692415952682495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,1,power_law_1.2,0.8295231819152832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,128,power_law_1.01,0.34469759464263916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,1,power_law_1.2,0.9123071670532227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,8,power_law_1.2,0.16543359756469728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,1,power_law_1.2,0.9168255805969239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,8,power_law_1.2,0.17649919986724855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,1,power_law_1.2,1.0377216339111328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,128,power_law_1.01,0.7525440216064453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,8,power_law_1.2,0.1835968017578125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,1,power_law_1.2,1.1665920257568358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,8,power_law_1.2,0.18932479619979858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,1,power_law_1.2,1.3911999702453612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,8,power_law_1.2,0.21288321018218995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,1,power_law_1.2,1.5866815567016601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,8,power_law_1.2,0.2316351890563965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,1,power_law_1.2,2.0195327758789063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,8,power_law_1.2,0.267903995513916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,1,power_law_1.2,2.4202495574951173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,8,power_law_1.2,0.3043584108352661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,8,power_law_1.2,0.355731201171875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,1,power_law_1.2,3.3389633178710936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,8,power_law_1.2,0.4287360191345215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,1,power_law_1.2,4.2221118927001955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,8,power_law_1.2,0.5520383834838867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,8,power_law_1.2,0.6948031902313232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,1,power_law_1.2,5.120102310180664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,8,power_law_1.2,0.9832703590393066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,32,power_law_1.01,0.03226880133152008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,32,power_law_1.01,0.03195520043373108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,8,power_law_1.2,1.2320192337036133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,32,power_law_1.01,0.030003198981285097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,32,power_law_1.01,0.03023360073566437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,1,power_law_1.2,9.854822540283203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,32,power_law_1.01,0.030316799879074097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,8,power_law_1.2,1.810099220275879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,32,power_law_1.01,0.03144960105419159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,32,power_law_1.01,0.031999999284744264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,32,power_law_1.01,0.032671999931335446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,32,power_law_1.01,0.032915198802948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,8,power_law_1.2,2.422412872314453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,32,power_law_1.01,0.03386240005493164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,32,power_law_1.01,0.03540480136871338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,32,power_law_1.01,0.03610239923000336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,32,power_law_1.01,0.03685120046138764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,32,power_law_1.01,0.037254399061203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,32,power_law_1.01,0.040057599544525146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,32,power_law_1.01,0.04183039963245392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,1,power_law_1.2,19.338540649414064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,8,power_law_1.2,4.4766590118408205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,32,power_law_1.01,0.043347200751304625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,32,power_law_1.01,0.047891199588775635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,32,power_law_1.01,0.051507198810577394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,32,power_law_1.01,0.059571200609207155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,32,power_law_1.01,0.06853119730949402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,32,power_law_1.01,0.10058879852294922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,32,power_law_1.01,0.12677119970321654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,32,power_law_1.01,0.1669376015663147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,32,power_law_1.01,0.21850879192352296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,32,power_law_1.01,0.29942400455474855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,32,power_law_1.01,0.4066688060760498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,32,power_law_1.01,0.4838399887084961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,32,power_law_1.01,0.8090304374694824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,64,balanced,0.06038400034109751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,64,balanced,0.060640002290407814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,64,balanced,0.06070933242638906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,64,balanced,0.06021333237489065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,64,balanced,0.0698880006869634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,64,balanced,0.0846666693687439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,64,balanced,0.09919466574986775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,64,balanced,0.09869333108266194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,64,balanced,0.09710933764775594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,64,balanced,0.10257066289583842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,64,balanced,0.10131200154622395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,64,balanced,0.09985599915186565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,64,balanced,0.10069866975148518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,64,balanced,0.09920533498128255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,32,power_law_1.01,1.6534656524658202
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,64,balanced,0.1153706709543864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,64,balanced,0.10871466994285583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,64,balanced,0.11437333623568217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,64,balanced,0.119759996732076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,64,balanced,0.12802666425704956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,64,balanced,0.14621333281199136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,64,balanced,0.15704533457756042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,64,balanced,0.18875734011332193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,64,balanced,0.21820799509684244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,64,balanced,0.2771199941635132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,64,balanced,0.3487306833267212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,64,balanced,0.4666239817937215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,1,power_law_1.01,0.08147199749946595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,64,balanced,0.5976746479670206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,64,balanced,0.7482186953226725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,64,balanced,1.1322346528371174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,64,balanced,2.1806507110595703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,1,power_law_1.01,0.10548479557037353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,1,power_law_1.01,0.10784640312194824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,1,power_law_1.01,0.16010240316390992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,1,power_law_1.01,0.20936319828033448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,1,power_law_1.01,0.3415744066238403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,1,power_law_1.01,0.36888320446014405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,1,power_law_1.01,0.3871615886688232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,16,power_law_1.2,0.059724801778793336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,16,4,balanced,0.03884266565243403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,1,power_law_1.01,0.39611520767211916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,16,4,balanced,0.037834666669368744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,16,4,balanced,0.03759466608365377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,16,4,balanced,0.03826666623353958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,16,4,balanced,0.038176000118255615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,16,power_law_1.2,0.0659712016582489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,16,4,balanced,0.03984533250331879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,16,4,balanced,0.038160001238187156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,16,4,balanced,0.03779733429352442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,16,4,balanced,0.04015466570854187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,16,4,balanced,0.03974399964014689
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,16,4,balanced,0.040005333721637726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,16,4,balanced,0.03997866561015447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,16,4,balanced,0.03989866624275843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,16,4,balanced,0.042021334171295166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,16,4,balanced,0.045824001232783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,16,4,balanced,0.046122665206591286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,16,4,balanced,0.04785066843032837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,16,4,balanced,0.054197331269582115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,16,4,balanced,0.05597866574923197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,16,4,balanced,0.06247466802597046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,16,4,balanced,0.07286933561166127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,16,4,balanced,0.0925546685854594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,16,4,balanced,0.10971200466156006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,16,4,balanced,0.14628799756368002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,16,power_law_1.2,0.05449600219726562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,16,4,balanced,0.17670400937398276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,1,power_law_1.01,0.40769281387329104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,16,4,balanced,0.24074665705362955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,16,4,balanced,0.3007520039876302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,16,4,balanced,0.3656746546427409
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,16,4,balanced,0.5518453518549601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,16,4,balanced,1.0491039752960205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,16,power_law_1.2,0.05051519870758057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,1,power_law_1.01,0.4177087783813477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,16,power_law_1.2,0.05208960175514221
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,1,power_law_1.01,0.4372992038726807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,16,power_law_1.2,0.05565440058708191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,1,power_law_1.01,0.4396671772003174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,16,power_law_1.2,0.05676800012588501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,16,power_law_1.2,0.06176000237464905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,1,power_law_1.01,0.4592127799987793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,16,power_law_1.2,0.06180480122566223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,1,power_law_1.01,0.479475212097168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,16,power_law_1.2,0.06314880251884461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,1,power_law_1.01,0.49538559913635255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,16,power_law_1.2,0.06344959735870362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,1,power_law_1.01,0.533568000793457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,16,power_law_1.2,0.06913920044898987
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,1,8,power_law_1.01,0.0721343994140625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,1,8,power_law_1.01,0.07885439991950989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,1,8,power_law_1.01,0.07200000286102295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,1,8,power_law_1.01,0.07109119892120361
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,16,power_law_1.2,0.07014399766921997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,1,8,power_law_1.01,0.07152640223503112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,1,power_law_1.01,0.6192128181457519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,1,8,power_law_1.01,0.07194240093231201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,1,8,power_law_1.01,0.07242879867553711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,1,8,power_law_1.01,0.07338240146636962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,1,8,power_law_1.01,0.07393280267715455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,16,power_law_1.2,0.07262719869613647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,1,8,power_law_1.01,0.0771776020526886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,1,8,power_law_1.01,0.0782912015914917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,1,8,power_law_1.01,0.08395519852638245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,1,8,power_law_1.01,0.08376320004463196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,1,power_law_1.01,0.6669824123382568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,1,8,power_law_1.01,0.09479680061340331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,1,8,power_law_1.01,0.10058879852294922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,16,power_law_1.2,0.08031359910964966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,1,8,power_law_1.01,0.10813440084457397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,1,8,power_law_1.01,0.12350720167160034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,1,8,power_law_1.01,0.1595136046409607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,1,8,power_law_1.01,0.17197439670562745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,1,8,power_law_1.01,0.23141119480133057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,16,power_law_1.2,0.0812991976737976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,1,8,power_law_1.01,0.2669055938720703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,1,power_law_1.01,0.8146047592163086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,1,8,power_law_1.01,0.40025601387023924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,1,8,power_law_1.01,0.5016895771026612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,1,8,power_law_1.01,0.729702377319336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,16,power_law_1.2,0.09069439768791199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,1,8,power_law_1.01,1.0395263671875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,1,power_law_1.01,0.9895423889160156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,1,8,power_law_1.01,1.641516876220703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,16,power_law_1.2,0.1030079960823059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,1,8,power_law_1.01,2.2129087448120117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,16,power_law_1.2,0.11325440406799317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,1,8,power_law_1.01,2.7042112350463867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,1,power_law_1.01,1.2612223625183105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,16,power_law_1.2,0.13211519718170167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,1,8,power_law_1.01,3.795238494873047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,1,power_law_1.01,1.4395968437194824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,16,power_law_1.2,0.14903039932250978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,1,8,power_law_1.01,7.228915405273438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,16,power_law_1.2,0.19292800426483153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,1,power_law_1.01,1.9582975387573243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,16,power_law_1.2,0.23893120288848876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,16,power_law_1.2,0.30307199954986574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,1,power_law_1.01,2.3795135498046873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,16,power_law_1.2,0.39655680656433107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,16,power_law_1.2,0.5469056129455566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,1,power_law_1.01,3.3255104064941405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,16,power_law_1.2,0.6738624095916748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,16,power_law_1.2,0.8982912063598633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,1,power_law_1.01,4.41212158203125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,32,power_law_1.01,0.03511039912700653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,32,power_law_1.01,0.03157120048999786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,16,power_law_1.2,1.421568012237549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,32,power_law_1.01,0.030028799176216127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,32,power_law_1.01,0.030444800853729248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,32,power_law_1.01,0.030451199412345885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,32,power_law_1.01,0.03237760066986084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,32,power_law_1.01,0.033030399680137636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,4,1,power_law_1.01,5.329171371459961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,32,power_law_1.01,0.032595199346542356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,32,power_law_1.01,0.03386879861354828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,16,power_law_1.2,2.5799488067626952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,32,power_law_1.01,0.033855998516082765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,32,power_law_1.01,0.034694400429725644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,32,power_law_1.01,0.03604480028152466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,32,power_law_1.01,0.037459200620651244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,32,power_law_1.01,0.038047999143600464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,32,power_law_1.01,0.04307839870452881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,32,power_law_1.01,0.04268159866333008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,32,power_law_1.01,0.04378879964351654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,4,1,power_law_1.01,8.206515502929687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,32,power_law_1.01,0.048332801461219786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,32,power_law_1.01,0.05418239831924439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,32,power_law_1.01,0.06047359704971313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,32,power_law_1.01,0.0760320007801056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,32,power_law_1.01,0.1025920033454895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,32,power_law_1.01,0.12691199779510498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,32,power_law_1.01,0.17884160280227662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,32,power_law_1.01,0.22012159824371338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,32,power_law_1.01,0.3126399993896484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,32,power_law_1.01,0.3837951898574829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,32,power_law_1.01,0.505075216293335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,32,power_law_1.01,0.7645440101623535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,4,1,power_law_1.01,15.613011169433594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,32,power_law_1.01,1.508351993560791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,32,4,balanced,0.05905066430568695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,32,4,balanced,0.06266666452089946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,32,4,balanced,0.06250666578610738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,32,4,balanced,0.07071466743946075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,32,4,balanced,0.07905599971612294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,32,4,balanced,0.10140800476074219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,32,4,balanced,0.12227200468381245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,32,4,balanced,0.12230400244394939
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,32,4,balanced,0.12385066350301106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,32,4,balanced,0.1239946683247884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,32,4,balanced,0.12030933300654094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,32,4,balanced,0.12416000167528789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,32,4,balanced,0.12589333454767862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,32,4,balanced,0.12917332847913107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,32,4,balanced,0.1358506679534912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,32,4,balanced,0.1384266714255015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,32,4,balanced,0.14857066671053568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,32,4,balanced,0.16291200121243796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,32,4,balanced,0.17531200249989828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,32,4,balanced,0.20340800285339355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,32,4,balanced,0.22869867086410522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,32,4,balanced,0.2849760055541992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,32,4,balanced,0.34065600236256915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,32,4,balanced,0.46059199174245197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,32,4,balanced,0.6191733280817667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,32,4,balanced,0.8612000147501627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,32,4,balanced,1.1373813152313232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,32,4,balanced,1.4139787356058757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,32,4,balanced,2.1833225886027017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,32,4,balanced,4.220394770304362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,8,8,power_law_1.01,0.05419520139694214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,8,8,power_law_1.01,0.057126402854919434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,8,8,power_law_1.01,0.051551997661590576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,8,8,power_law_1.01,0.05431680083274841
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,8,8,power_law_1.01,0.057132798433303836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,8,8,power_law_1.01,0.0622655987739563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,8,8,power_law_1.01,0.06343039870262146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,8,8,power_law_1.01,0.0654528021812439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,8,8,power_law_1.01,0.06709759831428527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,8,8,power_law_1.01,0.06550400257110596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,8,8,power_law_1.01,0.0666815996170044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,8,8,power_law_1.01,0.06704639792442321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,8,8,power_law_1.01,0.07034239768981934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,8,8,power_law_1.01,0.07470719814300537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,8,8,power_law_1.01,0.08010240197181702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,8,8,power_law_1.01,0.08481280207633972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,8,8,power_law_1.01,0.09347839951515198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,8,8,power_law_1.01,0.1093951940536499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,8,8,power_law_1.01,0.12337919473648071
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,8,8,power_law_1.01,0.14858239889144897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,8,8,power_law_1.01,0.1623744010925293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,8,8,power_law_1.01,0.1992576003074646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,8,8,power_law_1.01,0.24726400375366211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,8,8,power_law_1.01,0.3402688026428223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,8,8,power_law_1.01,0.4211264133453369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,8,8,power_law_1.01,0.6543551921844483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,8,8,power_law_1.01,0.8193535804748535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,8,8,power_law_1.01,0.9627391815185546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,8,8,power_law_1.01,1.525004768371582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,8,8,power_law_1.01,2.9056320190429688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,16,power_law_1.2,0.04188160002231598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,16,power_law_1.2,0.03487359881401062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,16,power_law_1.2,0.03335039913654327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,16,power_law_1.2,0.033011201024055484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,16,power_law_1.2,0.034195199608802795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,4,power_law_1.2,0.07494400143623352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,16,power_law_1.2,0.03583360016345978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,16,power_law_1.2,0.03578880131244659
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,16,power_law_1.2,0.035699200630187986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,16,power_law_1.2,0.03630079925060272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,16,power_law_1.2,0.036652800440788266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,4,power_law_1.2,0.09584640264511109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,16,power_law_1.2,0.03687680065631867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,16,power_law_1.2,0.03834879994392395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,16,power_law_1.2,0.038649600744247434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,16,power_law_1.2,0.040863999724388124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,4,power_law_1.2,0.08805760145187377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,16,power_law_1.2,0.045286399126052854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,16,power_law_1.2,0.0461760014295578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,16,power_law_1.2,0.050316798686981204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,16,power_law_1.2,0.05490559935569763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,16,power_law_1.2,0.05788159966468811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,4,power_law_1.2,0.08595839738845826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,16,power_law_1.2,0.06996480226516724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,16,power_law_1.2,0.07749760150909424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,16,power_law_1.2,0.09449599981307984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,4,power_law_1.2,0.12357120513916016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,16,power_law_1.2,0.1260159969329834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,16,power_law_1.2,0.15598080158233643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,16,power_law_1.2,0.1984768033027649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,4,power_law_1.2,0.15471359491348266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,16,power_law_1.2,0.2959167957305908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,16,power_law_1.2,0.40932478904724123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,4,power_law_1.2,0.17812479734420777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,16,power_law_1.2,0.4232128143310547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,4,power_law_1.2,0.2321023941040039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,16,power_law_1.2,0.7334271907806397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,4,power_law_1.2,0.24117119312286378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,16,power_law_1.2,2.2056768417358397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,4,power_law_1.2,0.2338047981262207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,4,power_law_1.2,0.2599168062210083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,4,power_law_1.2,0.26238079071044923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,4,power_law_1.2,0.27654399871826174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,4,power_law_1.2,0.2852288007736206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,4,power_law_1.2,0.31288321018218995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,4,power_law_1.2,0.31536641120910647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,64,power_law_1.01,0.1180351972579956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,4,power_law_1.2,0.3315903902053833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,64,power_law_1.01,0.07463039755821228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,4,power_law_1.2,0.3509119987487793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,64,power_law_1.01,0.06312959790229797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,4,power_law_1.2,0.39193599224090575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,64,power_law_1.01,0.05018240213394165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,64,power_law_1.01,0.0531391978263855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,4,power_law_1.2,0.453542423248291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,64,power_law_1.01,0.05686399936676025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,4,power_law_1.2,0.47882881164550783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,64,power_law_1.01,0.057766401767730714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,64,power_law_1.01,0.060499197244644164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,4,power_law_1.2,0.5951744079589844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,64,power_law_1.01,0.06537600159645081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,4,power_law_1.2,0.6825535774230957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,64,power_law_1.01,0.06499840021133423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,64,power_law_1.01,0.06471679806709289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,4,power_law_1.2,0.8901568412780761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,64,power_law_1.01,0.06601600050926208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,64,power_law_1.01,0.06755840182304382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,4,power_law_1.2,0.9783743858337403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,64,power_law_1.01,0.06981760263442993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,64,power_law_1.01,0.07711359858512878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,4,power_law_1.2,1.4079680442810059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,16,power_law_1.2,0.05924479961395264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,64,power_law_1.01,0.07775359749794006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,16,power_law_1.2,0.06354560256004334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,64,power_law_1.01,0.08310400247573853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,4,power_law_1.2,1.576863956451416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,16,power_law_1.2,0.05452160239219665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,64,power_law_1.01,0.08844799995422363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,16,power_law_1.2,0.052121597528457644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,64,power_law_1.01,0.09896960258483886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,4,power_law_1.2,2.156345558166504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,16,power_law_1.2,0.054790401458740236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,64,power_law_1.01,0.11708159446716308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,16,power_law_1.2,0.05924479961395264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,64,power_law_1.01,0.12627840042114258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,16,power_law_1.2,0.06183040142059326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,4,power_law_1.2,3.1153280258178713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,64,power_law_1.01,0.16433279514312743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,16,power_law_1.2,0.06533759832382202
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,64,power_law_1.01,0.1823232054710388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,16,power_law_1.2,0.06595199704170226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,64,power_law_1.01,0.2450239896774292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,16,power_law_1.2,0.06575999855995178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,64,power_law_1.01,0.331987190246582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,4,power_law_1.2,5.879404830932617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,16,power_law_1.2,0.07008640170097351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,64,power_law_1.01,0.4395135879516602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,16,power_law_1.2,0.07103360295295716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,16,power_law_1.2,0.07260800004005433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,64,power_law_1.01,0.5655935764312744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,16,power_law_1.2,0.07154560089111328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,64,power_law_1.01,0.6719808101654052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,16,power_law_1.2,0.08101760149002075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,16,power_law_1.2,0.08293759822845459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,64,power_law_1.01,0.9738688468933105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,16,power_law_1.2,0.08951680064201355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,16,power_law_1.2,0.10427520275115967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,4,4,balanced,0.038848000268141426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,4,4,balanced,0.03770666569471359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,4,4,balanced,0.03766400118668874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,4,4,balanced,0.041797334949175514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,4,4,balanced,0.04766400158405304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,4,4,balanced,0.04789333542188009
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,4,4,balanced,0.047983999053637184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,4,4,balanced,0.04776533444722494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,16,power_law_1.2,0.12032639980316162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,4,4,balanced,0.049813335140546165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,4,4,balanced,0.047983999053637184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,4,4,balanced,0.048197334011395775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,4,4,balanced,0.050026665131251015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,4,4,balanced,0.05217066903909048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,4,4,balanced,0.0518506666024526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,4,4,balanced,0.05620799958705902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,4,4,balanced,0.057818666100502014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,4,4,balanced,0.05872533222039541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,4,4,balanced,0.06413866579532623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,4,4,balanced,0.06852266689141591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,4,4,balanced,0.08082666496435802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,4,4,balanced,0.0885706643263499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,4,4,balanced,0.12140267093976338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,4,4,balanced,0.13964800039927164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,64,power_law_1.01,2.0022911071777343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,4,4,balanced,0.19131733973821005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,4,4,balanced,0.24810133377710977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,4,4,balanced,0.3513919909795125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,4,4,balanced,0.44651198387145996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,16,power_law_1.2,0.14881279468536376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,4,4,balanced,0.5508586565653483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,4,4,balanced,0.855621337890625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,4,4,balanced,1.6917120615641277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,1,balanced,0.05813866853713989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,1,balanced,0.0621013343334198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,1,balanced,0.07030933101971944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,1,balanced,0.09866666793823242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,1,balanced,0.14256532986958823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,16,power_law_1.2,0.17032320499420167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,1,balanced,0.21563732624053955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,1,balanced,0.28621333837509155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,1,balanced,0.28944534063339233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,1,balanced,0.291594664255778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,1,balanced,0.2927733262379964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,1,balanced,0.29755733410517377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,1,balanced,0.30479466915130615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,1,balanced,0.30881067117055255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,1,balanced,0.3147520025571187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,1,balanced,0.3242986599604289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,1,balanced,0.3323520024617513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,16,power_law_1.2,0.21868159770965576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,1,balanced,0.3566346565882365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,1,balanced,0.40135467052459717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,1,balanced,0.4361279805501302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,1,balanced,0.513152003288269
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,1,balanced,0.5738133192062378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,1,balanced,0.7353973388671875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,1,balanced,0.9260959625244141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,1,balanced,1.2196959654490154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,16,power_law_1.2,0.25598080158233644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,1,balanced,1.642367998758952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,1,balanced,2.341503938039144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,1,balanced,3.0238612492879233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,16,power_law_1.2,0.3337984085083008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,1,balanced,3.794432004292806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,16,power_law_1.2,0.4514944076538086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,1,balanced,5.772469202677409
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,16,power_law_1.2,0.6451903820037842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,1,balanced,11.144480387369791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,16,power_law_1.2,0.8421824455261231
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,8,16,balanced,0.03551466763019562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,8,16,balanced,0.03549333413441976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,8,16,balanced,0.03749333322048187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,8,16,balanced,0.037685332198937736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,8,16,balanced,0.03803733239571253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,8,16,balanced,0.04015466570854187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,8,16,balanced,0.03979733337958654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,8,16,balanced,0.041984001795450844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,8,16,balanced,0.041573333243529
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,8,16,balanced,0.0397119993964831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,8,16,balanced,0.04193066557248434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,8,16,balanced,0.041637333730856575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,8,16,balanced,0.04376000165939331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,8,16,balanced,0.04381866753101349
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,8,16,balanced,0.04571199913819631
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,8,16,balanced,0.047775998711586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,8,16,balanced,0.04771733283996582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,8,16,balanced,0.0516480008761088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,8,16,balanced,0.05397866666316986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,16,power_law_1.2,1.0394495964050292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,8,16,balanced,0.0582239975531896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,8,16,balanced,0.06247466802597046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,8,16,balanced,0.07639466722806294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,8,16,balanced,0.08747200171152751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,8,16,balanced,0.11766933401425679
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,8,16,balanced,0.14455466469128928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,8,16,balanced,0.19458132982254028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,8,16,balanced,0.23560533920923868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,8,16,balanced,0.28693334261576336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,32,power_law_1.01,0.03521920144557953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,8,16,balanced,0.4210133155186971
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,8,16,balanced,0.7889119784037272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,32,power_law_1.01,0.034835198521614076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,32,power_law_1.01,0.030355200171470642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,32,power_law_1.01,0.030329599976539612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,16,power_law_1.2,1.5797951698303223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,32,power_law_1.01,0.030943998694419862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,32,power_law_1.01,0.030847999453544616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,32,power_law_1.01,0.031097599864006044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,32,power_law_1.01,0.03153280019760132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,32,power_law_1.01,0.031865599751472476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,32,power_law_1.01,0.03252480030059814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,16,power_law_1.2,3.2657024383544924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,32,power_law_1.01,0.03255679905414581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,32,power_law_1.01,0.034041601419448855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,32,power_law_1.01,0.03368319869041443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,32,power_law_1.01,0.0344895988702774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,32,power_law_1.01,0.03844479918479919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,32,power_law_1.01,0.03925760090351105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,16,power_law_1.2,0.07961599826812744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,32,power_law_1.01,0.041529598832130435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,32,power_law_1.01,0.04545280039310455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,16,power_law_1.2,0.058195197582244874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,32,power_law_1.01,0.04699519872665405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,32,power_law_1.01,0.05235199928283692
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,16,power_law_1.2,0.05326719880104065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,32,power_law_1.01,0.05597440004348755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,32,power_law_1.01,0.06744319796562195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,8,power_law_1.01,0.06689280271530151
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,16,power_law_1.2,0.05733759999275208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,32,power_law_1.01,0.07925119996070862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,8,power_law_1.01,0.08458240032196045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,32,power_law_1.01,0.11951999664306641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,16,power_law_1.2,0.06844800114631652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,8,power_law_1.01,0.05857920050621033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,32,power_law_1.01,0.14680960178375244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,8,power_law_1.01,0.0661184012889862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,16,power_law_1.2,0.07737600207328796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,32,power_law_1.01,0.2088831901550293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,8,power_law_1.01,0.07162240147590637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,8,power_law_1.01,0.08837760090827942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,32,power_law_1.01,0.23841280937194825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,16,power_law_1.2,0.08053119778633118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,8,power_law_1.01,0.0913856029510498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,2,32,power_law_1.01,0.2728447914123535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,16,power_law_1.2,0.08254719972610473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,8,power_law_1.01,0.09290879964828491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,2,2,power_law_1.01,0.05010560154914856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,8,power_law_1.01,0.09143679738044738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,2,2,power_law_1.01,0.06509439945220948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,16,power_law_1.2,0.08085119724273682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,2,2,power_law_1.01,0.05914239883422852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,2,32,power_law_1.01,0.4642752170562744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,2,2,power_law_1.01,0.06763520240783691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,8,power_law_1.01,0.09270399808883667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,2,2,power_law_1.01,0.06856319904327393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,2,2,power_law_1.01,0.06915199756622314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,2,2,power_law_1.01,0.06982399821281433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,8,power_law_1.01,0.09382399916648865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,16,power_law_1.2,0.0837823987007141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,2,2,power_law_1.01,0.06945279836654664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,2,2,power_law_1.01,0.07297919988632202
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,2,2,power_law_1.01,0.07259520292282104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,8,power_law_1.01,0.09820799827575684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,2,2,power_law_1.01,0.07647359967231751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,2,2,power_law_1.01,0.07715200185775757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,16,power_law_1.2,0.08650239706039428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,2,2,power_law_1.01,0.08102399706840516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,2,32,power_law_1.01,0.8658816337585449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,8,power_law_1.01,0.09781759977340698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,2,2,power_law_1.01,0.08447359800338745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,2,2,power_law_1.01,0.09457280039787293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,2,2,power_law_1.01,0.09950079917907714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,8,power_law_1.01,0.10355199575424194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,2,2,power_law_1.01,0.1030784010887146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,16,power_law_1.2,0.08901119828224183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,2,2,power_law_1.01,0.12222720384597778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,2,2,power_law_1.01,0.13744640350341797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,8,power_law_1.01,0.10848000049591064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,2,2,power_law_1.01,0.1654207944869995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,2,2,power_law_1.01,0.20887680053710939
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,2,2,power_law_1.01,0.2723583936691284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,16,power_law_1.2,0.09045760035514831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,8,power_law_1.01,0.11429120302200317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,2,2,power_law_1.01,0.34138240814208987
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,2,2,power_law_1.01,0.4828224182128906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,8,power_law_1.01,0.12066559791564942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,2,2,power_law_1.01,0.6427775859832764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,16,power_law_1.2,0.09386240243911743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,2,2,power_law_1.01,0.9031807899475097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,8,power_law_1.01,0.13150080442428588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,2,2,power_law_1.01,1.1515968322753907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,16,power_law_1.2,0.10156160593032837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,8,power_law_1.01,0.1474303960800171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,2,2,power_law_1.01,1.5648063659667968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,8,power_law_1.01,0.17637120485305785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,16,power_law_1.2,0.10796799659729003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,2,2,power_law_1.01,2.161235237121582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,8,power_law_1.01,0.1956480026245117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,16,power_law_1.2,0.11459200382232666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,8,power_law_1.01,0.2400063991546631
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,16,power_law_1.2,0.13829760551452636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,2,2,power_law_1.01,4.5792896270751955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,8,power_law_1.01,0.2890304088592529
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,16,power_law_1.2,0.14072320461273194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,8,power_law_1.01,0.3985919952392578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,16,power_law_1.2,0.16758400201797485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,8,power_law_1.01,0.49317121505737305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,16,power_law_1.2,0.20449280738830566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,8,power_law_1.01,0.6723775863647461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,16,power_law_1.2,0.24644479751586915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,8,power_law_1.01,0.8602496147155761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,64,balanced,0.05796800057093302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,64,balanced,0.05005866785844167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,64,balanced,0.050213331977526345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,64,balanced,0.050053333242734276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,16,power_law_1.2,0.3243583917617798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,64,balanced,0.051914667089780174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,64,balanced,0.05227733155091604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,64,balanced,0.05621333420276642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,64,balanced,0.055973331133524575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,64,balanced,0.05559466779232025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,64,balanced,0.05637866755326589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,64,balanced,0.05615466833114624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,64,balanced,0.05844266712665558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,64,balanced,0.05863999823729197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,64,balanced,0.06031466523806254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,64,balanced,0.064751997590065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,64,balanced,0.0668639987707138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,64,balanced,0.07073600093523662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,64,balanced,0.07660800218582153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,64,balanced,0.08399466673533122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,8,power_law_1.01,1.0379008293151855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,64,balanced,0.09707199533780415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,64,balanced,0.11352533102035522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,16,power_law_1.2,0.4242559909820557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,64,balanced,0.1439466675122579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,64,balanced,0.16849599281946817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,64,balanced,0.22206934293111166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,64,balanced,0.2746986746788025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,64,balanced,0.37654932339986164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,64,balanced,0.4836213191350301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,64,balanced,0.5930826663970947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,64,balanced,0.8888053099314371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,16,power_law_1.2,0.5043519973754883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,64,balanced,1.7435626983642578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,8,power_law_1.01,1.6343360900878907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,16,power_law_1.2,0.8091520309448242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,16,power_law_1.2,0.9765119552612305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,8,power_law_1.01,3.291020965576172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,16,power_law_1.2,1.5674304008483886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,2,4,power_law_1.01,0.026080000400543212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,2,4,power_law_1.01,0.028543999791145323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,2,4,power_law_1.01,0.03023360073566437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,2,4,power_law_1.01,0.030752000212669373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,16,power_law_1.2,2.2892799377441406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,2,4,power_law_1.01,0.035545599460601804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,2,4,power_law_1.01,0.046988800168037415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,2,4,power_law_1.01,0.04748159945011139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,2,4,power_law_1.01,0.048844799399375916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,2,4,power_law_1.01,0.050892800092697144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,2,4,power_law_1.01,0.05120000243186951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,2,4,power_law_1.01,0.05189759731292724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,16,power_law_1.2,5.043564987182617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,2,4,power_law_1.01,0.05504000186920166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,2,4,power_law_1.01,0.058822399377822875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,2,4,power_law_1.01,0.05841919779777527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,2,4,power_law_1.01,0.0652288019657135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,2,4,power_law_1.01,0.06666880249977111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,32,2,power_law_1.2,0.06936320066452026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,2,4,power_law_1.01,0.06601600050926208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,2,4,power_law_1.01,0.07251840233802795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,4,power_law_1.01,0.08769919872283935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,32,2,power_law_1.2,0.07509120106697083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,4,power_law_1.01,0.09984639883041382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,4,power_law_1.01,0.10232959985733033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,32,2,power_law_1.2,0.07685760259628296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,4,power_law_1.01,0.1325503945350647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,8,power_law_1.2,0.0466623991727829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,4,power_law_1.01,0.1562559962272644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,32,2,power_law_1.2,0.08112639784812928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,4,power_law_1.01,0.20832641124725343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,8,power_law_1.2,0.04913919866085052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,4,power_law_1.01,0.26703360080718996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,32,2,power_law_1.2,0.09596160054206848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,8,power_law_1.2,0.04903680086135864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,4,power_law_1.01,0.3522559881210327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,32,2,power_law_1.2,0.11558400392532349
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,8,power_law_1.2,0.04696959853172302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,4,power_law_1.01,0.4481152057647705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,32,2,power_law_1.2,0.12813440561294556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,8,power_law_1.2,0.05377280116081238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,4,power_law_1.01,0.5609920024871826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,2,power_law_1.01,0.0419840008020401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,32,2,power_law_1.2,0.15073280334472655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,8,power_law_1.2,0.057708799839019775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,4,power_law_1.01,0.8523967742919922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,2,power_law_1.01,0.04503679871559143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,32,2,power_law_1.2,0.15602560043334962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,2,power_law_1.01,0.050271999835968015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,8,power_law_1.2,0.0593280017375946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,8,power_law_1.2,0.06781439781188965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,32,2,power_law_1.2,0.15960320234298705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,2,power_law_1.01,0.055852800607681274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,4,power_law_1.01,1.556492805480957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,8,power_law_1.2,0.0686464011669159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,32,2,power_law_1.2,0.16754560470581054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,2,power_law_1.01,0.06316800117492676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,2,power_law_1.01,0.08214399814605713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,8,power_law_1.2,0.06906239986419678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,32,2,power_law_1.2,0.1685439944267273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,2,power_law_1.01,0.08453760147094727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,8,power_law_1.2,0.07746559977531434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,32,2,power_law_1.2,0.17463680505752563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,2,power_law_1.01,0.0864960014820099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,8,power_law_1.2,0.07893120050430298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,32,2,power_law_1.2,0.18106240034103394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,2,power_law_1.01,0.08823680281639099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,2,power_law_1.01,0.08989440202713013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,8,power_law_1.2,0.08016639947891235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,32,2,power_law_1.2,0.19230719804763793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,2,power_law_1.01,0.09321600198745728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,8,power_law_1.2,0.0868287980556488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,32,2,power_law_1.2,0.20017919540405274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,2,power_law_1.01,0.09492480158805847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,8,power_law_1.2,0.09773439764976502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,2,power_law_1.01,0.09724159836769104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,32,2,power_law_1.2,0.21704320907592772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,2,power_law_1.01,0.10044159889221191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,8,power_law_1.2,0.10041600465774536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,32,2,power_law_1.2,0.24085121154785155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,2,power_law_1.01,0.10649600028991699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,8,power_law_1.2,0.10979199409484863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,2,power_law_1.01,0.11237759590148926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,4,1,balanced,0.03190933416287104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,4,1,balanced,0.03356799980004629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,4,1,balanced,0.03826133410135905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,4,1,balanced,0.053861334919929504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,32,2,power_law_1.2,0.26714239120483396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,4,1,balanced,0.08654933174451192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,4,1,balanced,0.08643733461697896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,4,1,balanced,0.08656533559163411
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,4,1,balanced,0.08913600444793701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,4,1,balanced,0.0885706643263499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,4,1,balanced,0.09060266613960266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,2,power_law_1.01,0.12106239795684814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,4,1,balanced,0.09107733766237895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,4,1,balanced,0.09251733620961507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,4,1,balanced,0.09360532959302266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,4,1,balanced,0.09921600421269734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,4,1,balanced,0.10402133067448933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,4,1,balanced,0.10683199763298035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,8,power_law_1.2,0.1250815987586975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,4,1,balanced,0.11111467083295186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,4,1,balanced,0.12743999560674033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,1,balanced,0.13037866353988647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,1,balanced,0.1668000022570292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,1,balanced,0.1803200046221415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,1,balanced,0.25994666417439777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,1,balanced,0.30980799595514935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,2,power_law_1.01,0.13696000576019288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,1,balanced,0.43036266167958576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,32,2,power_law_1.2,0.3211071968078613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,1,balanced,0.5418293476104736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,1,balanced,0.7874933083852133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,1,balanced,1.0196373462677002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,8,power_law_1.2,0.13808640241622924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,1,balanced,1.263050635655721
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,2,power_law_1.01,0.15430400371551514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,1,balanced,1.9929866790771484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,32,2,power_law_1.2,0.3778815984725952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,2,power_law_1.01,0.1953727960586548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,1,balanced,3.842735926310221
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,8,power_law_1.2,0.16995199918746948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,2,power_law_1.01,0.21875839233398436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,32,2,power_law_1.2,0.48299517631530764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,2,power_law_1.01,0.29589118957519533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,8,power_law_1.2,0.20385921001434326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,2,power_law_1.01,0.36487679481506347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,32,2,power_law_1.2,0.6158592224121093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,8,power_law_1.2,0.2580287933349609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,2,power_law_1.01,0.5545407772064209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,32,2,power_law_1.2,0.8451647758483887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,8,power_law_1.2,0.3483839988708496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,2,power_law_1.01,0.6774720191955567
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,2,power_law_1.01,0.9713983535766602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,8,power_law_1.2,0.4597631931304932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,32,2,power_law_1.2,1.0533632278442382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,2,power_law_1.01,1.221664047241211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,8,power_law_1.2,0.5846015930175781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,32,2,power_law_1.2,1.528371238708496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,2,power_law_1.01,1.5750271797180175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,8,power_law_1.2,0.8763327598571777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,32,2,power_law_1.2,1.9529535293579101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,2,power_law_1.01,2.4255552291870117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,8,power_law_1.2,1.1131135940551757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,32,2,power_law_1.2,2.4529983520507814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,8,power_law_1.2,1.405900764465332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,2,power_law_1.01,4.776512145996094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,32,2,power_law_1.2,3.862451171875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,8,power_law_1.2,2.1869760513305665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,8,power_law_1.2,4.319903945922851
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,32,2,power_law_1.2,7.805023956298828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,1,balanced,0.053871999184290566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,1,balanced,0.05339199801286062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,1,balanced,0.05745066702365875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,1,balanced,0.06595733265082042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,1,balanced,0.09857599933942159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,1,balanced,0.09908800323804219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,1,balanced,0.10261866450309753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,1,balanced,0.10239467024803162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,1,balanced,0.10400000214576721
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,1,balanced,0.10389866431554158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,1,balanced,0.10800000031789143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,1,balanced,0.1113973359266917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,1,balanced,0.11389866471290588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,1,balanced,0.11853333314259847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,1,balanced,0.12337600191434224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,1,balanced,0.12986666957537332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,1,balanced,0.13871999581654867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,1,balanced,0.1618880033493042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,1,balanced,0.18122132619222006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,1,balanced,0.22667733828226724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,1,balanced,0.2678613265355428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,1,balanced,0.3898080190022786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,1,balanced,0.4798239866892497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,1,balanced,0.6994986534118652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,1,balanced,0.9005813598632812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,1,balanced,1.3180267016092937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,1,balanced,1.7305866877237956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,1,balanced,2.147792021433512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,1,balanced,3.3856213887532554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,1,balanced,6.5921065012613935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,8,balanced,0.04606399933497111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,4,1,power_law_1.2,0.07343360185623168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,8,balanced,0.04997866849104563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,8,balanced,0.07064533233642578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,8,balanced,0.09876267115275066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,4,1,power_law_1.2,0.13523839712142943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,8,balanced,0.14692800243695578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,8,balanced,0.24517333507537842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,8,balanced,0.2449280023574829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,4,1,power_law_1.2,0.12611839771270753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,8,balanced,0.24628265698750815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,8,balanced,0.24789865811665854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,4,1,power_law_1.2,0.20305919647216797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,8,balanced,0.24716265996297201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,8,balanced,0.25116799275080365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,4,1,power_law_1.2,0.20108799934387206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,8,balanced,0.25170133511225384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,8,balanced,0.25380800167719525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,8,balanced,0.2573866645495097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,4,1,power_law_1.2,0.20006399154663085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,8,balanced,0.2670346697171529
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,8,balanced,0.26815466086069745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,4,1,power_law_1.2,0.208076810836792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,8,balanced,0.2733866572380066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,8,balanced,0.28861333926518756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,4,1,power_law_1.2,0.20746879577636718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,8,balanced,0.29816534121831256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,8,balanced,0.3223466674486796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,4,1,power_law_1.2,0.21264638900756835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,8,balanced,0.3490240176518758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,8,balanced,0.3946400086085002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,4,1,power_law_1.2,0.2170880079269409
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,8,balanced,0.4373546838760376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,4,1,power_law_1.2,0.21600639820098877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,8,balanced,0.570575992266337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,8,balanced,0.6343199809392294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,4,1,power_law_1.2,0.2227776050567627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,4,1,power_law_1.2,0.23098878860473632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,8,balanced,0.9102986653645834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,4,1,power_law_1.2,0.2366015911102295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,8,balanced,1.0665280024210613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,4,1,power_law_1.2,0.25194239616394043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,8,balanced,1.3734933535257976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,4,1,power_law_1.2,0.2701119899749756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,4,1,power_law_1.2,0.27210240364074706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,8,balanced,1.9707999229431152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,4,1,power_law_1.2,0.3128767967224121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,4,1,power_law_1.2,0.3353280067443848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,4,1,power_law_1.2,0.42195839881896974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,8,balanced,3.8276265462239585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,4,1,power_law_1.2,0.4876095771789551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,4,1,power_law_1.2,0.6204351902008056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,4,1,power_law_1.2,0.7861055850982666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,4,1,power_law_1.2,1.0701824188232423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,4,1,power_law_1.2,1.400876808166504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,4,1,power_law_1.2,2.062540817260742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,4,1,power_law_1.2,2.6955711364746096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,4,1,power_law_1.2,3.345868682861328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,4,1,power_law_1.2,5.029587173461914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,4,1,power_law_1.2,9.84111328125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,32,power_law_1.2,0.12573440074920655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,32,power_law_1.2,0.17140480279922485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,32,power_law_1.2,0.14143999814987182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,32,power_law_1.2,0.08842880129814149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,32,power_law_1.2,0.0991104006767273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,32,power_law_1.2,0.10518399477005005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,4,8,balanced,0.039808000127474465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,4,8,balanced,0.04164266586303711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,4,8,balanced,0.03966933240493139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,4,8,balanced,0.03966933240493139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,4,8,balanced,0.039503999054431915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,4,8,balanced,0.03962666789690653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,4,8,balanced,0.04154666761557261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,4,8,balanced,0.039733332892258964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,4,8,balanced,0.04165333261092504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,4,8,balanced,0.039877332746982574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,4,8,balanced,0.041802664597829185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,4,8,balanced,0.04162133236726125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,4,8,balanced,0.041877334316571556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,32,power_law_1.2,0.11141760349273681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,4,8,balanced,0.04375466704368591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,4,8,balanced,0.048101335763931274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,4,8,balanced,0.04957333207130432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,4,8,balanced,0.0496373325586319
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,4,8,balanced,0.05598400036493937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,4,8,balanced,0.06198399762312571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,4,8,balanced,0.07333866755167644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,4,8,balanced,0.07849066456158955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,4,8,balanced,0.09701866904894511
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,4,8,balanced,0.11136533816655476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,4,8,balanced,0.13926933209101358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,4,8,balanced,0.16750399271647134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,4,8,balanced,0.22483199834823608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,4,8,balanced,0.27956799666086835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,4,8,balanced,0.33746667702992755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,4,8,balanced,0.5067360003789266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,32,power_law_1.2,0.13014400005340576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,4,8,balanced,0.9564159711201986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,32,power_law_1.2,0.13334399461746216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,32,power_law_1.2,0.14531199932098388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,32,power_law_1.2,0.13969279527664186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,32,power_law_1.2,0.15050239562988282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,32,power_law_1.2,0.15304319858551024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,32,power_law_1.2,0.15516799688339233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,32,power_law_1.2,0.16602879762649536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,32,power_law_1.2,0.16220799684524537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,32,power_law_1.2,0.16281599998474122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,32,power_law_1.2,0.18318079710006713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,1,power_law_1.01,0.08794879913330078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,32,power_law_1.2,0.178656005859375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,1,power_law_1.01,0.12548480033874512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,32,power_law_1.2,0.2127295970916748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,1,power_law_1.01,0.14622080326080322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,32,power_law_1.2,0.24177920818328857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,1,power_law_1.01,0.2517695903778076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,32,power_law_1.2,0.2962368011474609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,1,power_law_1.01,0.34903678894042967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,32,power_law_1.2,0.34029440879821776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,32,power_law_1.2,0.4814591884613037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,1,power_law_1.01,0.609932804107666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,32,power_law_1.2,0.5579648017883301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,1,power_law_1.01,0.6540671825408936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,32,power_law_1.2,0.8236288070678711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,1,power_law_1.01,0.6917183876037598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,1,power_law_1.01,0.7195007801055908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,32,power_law_1.2,1.0623231887817384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,1,power_law_1.01,0.73471999168396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,32,power_law_1.2,1.2498944282531739
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,1,power_law_1.01,0.7544703960418702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,32,power_law_1.2,1.8844415664672851
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,1,power_law_1.01,0.7711103916168213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,1,power_law_1.01,0.7874879837036133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,1,power_law_1.01,0.813542366027832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,32,power_law_1.2,4.352896118164063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,1,power_law_1.01,0.8561984062194824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,1,power_law_1.01,0.8592191696166992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,1,power_law_1.01,0.9050047874450684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,1,power_law_1.01,1.016140842437744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,1,power_law_1.01,1.07456636428833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,1,power_law_1.01,1.2661439895629882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,1,power_law_1.01,1.3475520133972168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,8,power_law_1.01,0.04583680033683777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,8,power_law_1.01,0.049081599712371825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,8,power_law_1.01,0.04746879935264588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,1,power_law_1.01,1.6603904724121095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,8,power_law_1.01,0.04962559938430786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,8,power_law_1.01,0.052051198482513425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,8,power_law_1.01,0.0587007999420166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,8,power_law_1.01,0.06254079937934875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,1,power_law_1.01,1.8319488525390626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,8,power_law_1.01,0.06280320286750793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,8,power_law_1.01,0.06542080044746398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,8,power_law_1.01,0.06700159907341004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,8,power_law_1.01,0.07148799896240235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,1,power_law_1.01,2.4059776306152343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,8,power_law_1.01,0.07197440266609192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,8,power_law_1.01,0.07432960271835327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,8,power_law_1.01,0.07664639949798584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,8,power_law_1.01,0.08513280153274536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,1,power_law_1.01,3.034809684753418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,8,power_law_1.01,0.0923520028591156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,8,power_law_1.01,0.09582719802856446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,8,power_law_1.01,0.10995199680328369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,8,power_law_1.01,0.12328319549560547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,8,power_law_1.01,0.14999040365219116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,1,power_law_1.01,4.320614242553711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,8,power_law_1.01,0.17329280376434325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,8,power_law_1.01,0.22846078872680664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,8,power_law_1.01,0.2824575901031494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,8,power_law_1.01,0.38073599338531494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,8,power_law_1.01,0.464467191696167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,1,power_law_1.01,5.2510017395019535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,8,power_law_1.01,0.6751935958862305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,8,power_law_1.01,0.9916607856750488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,8,power_law_1.01,1.1664511680603027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,2,1,power_law_1.01,6.45638427734375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,8,power_law_1.01,1.6570880889892579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,8,power_law_1.01,3.501804733276367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,2,1,power_law_1.01,9.662655639648438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,2,1,power_law_1.01,18.522866821289064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,4,power_law_1.01,0.14931199550628663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,4,power_law_1.01,0.2399616003036499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,4,power_law_1.01,0.4585408210754395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,4,power_law_1.01,0.2024768114089966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,4,power_law_1.01,0.35817599296569824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,4,power_law_1.01,0.5128448009490967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,4,power_law_1.01,0.5819071769714356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,4,power_law_1.01,0.7807424068450928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,4,power_law_1.01,0.8171968460083008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,16,power_law_1.2,0.07453439831733703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,4,power_law_1.01,0.8505279541015625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,16,power_law_1.2,0.1297600030899048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,4,power_law_1.01,0.8978752136230469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,16,power_law_1.2,0.0786624014377594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,4,power_law_1.01,0.9217791557312012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,16,power_law_1.2,0.06069120168685913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,4,power_law_1.01,0.9395135879516602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,16,power_law_1.2,0.06997759938240052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,4,power_law_1.01,0.9705984115600585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,16,power_law_1.2,0.08115199804306031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,4,power_law_1.01,1.0219903945922852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,16,power_law_1.2,0.08811519742012024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,4,power_law_1.01,1.0091775894165038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,16,power_law_1.2,0.10601600408554077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,16,power_law_1.2,0.11036159992218017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,4,power_law_1.01,1.061030387878418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,16,power_law_1.2,0.10598399639129638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,4,power_law_1.01,1.1531519889831543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,16,power_law_1.2,0.10970239639282227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,4,power_law_1.01,1.2405695915222168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,16,power_law_1.2,0.11533440351486206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,4,power_law_1.01,1.3488448143005372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,16,power_law_1.2,0.11522560119628907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,4,power_law_1.01,1.3754560470581054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,16,power_law_1.2,0.11750400066375732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,16,power_law_1.2,0.1260543942451477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,4,power_law_1.01,1.5716480255126952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,16,power_law_1.2,0.13411200046539307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,4,power_law_1.01,1.7451200485229492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,16,power_law_1.2,0.14433280229568482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,16,power_law_1.2,0.16844799518585205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,4,power_law_1.01,2.177779197692871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,16,power_law_1.2,0.1681023955345154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,4,power_law_1.01,2.215769577026367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,16,power_law_1.2,0.20134398937225342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,16,power_law_1.2,0.23644800186157228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,4,power_law_1.01,2.7873600006103514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,16,power_law_1.2,0.33755519390106203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,4,power_law_1.01,3.4150016784667967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,16,power_law_1.2,0.35868799686431885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,16,power_law_1.2,0.4834496021270752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,4,power_law_1.01,3.8853118896484373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,16,power_law_1.2,0.6250048160552979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,16,power_law_1.2,0.8894720077514648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,4,power_law_1.01,5.626675033569336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,16,power_law_1.2,1.113868808746338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,16,power_law_1.2,1.3490816116333009
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,4,power_law_1.01,10.588428497314453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,16,power_law_1.2,2.2285247802734376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,16,power_law_1.2,4.316499328613281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,1,128,balanced,0.0346666673819224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,1,128,balanced,0.03443199892838796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,1,128,balanced,0.03643200049797694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,1,128,balanced,0.03586133321126302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,1,128,balanced,0.03570666660865148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,1,128,balanced,0.03640533238649368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,1,128,balanced,0.035818666219711304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,1,128,balanced,0.03749333322048187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,1,128,balanced,0.038021333515644073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,1,128,balanced,0.035973332822322845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,1,128,balanced,0.03766933331886927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,1,128,balanced,0.03804266701141993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,1,128,balanced,0.03846933444341024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,1,128,balanced,0.037808001041412354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,1,128,balanced,0.04045333216587702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,1,128,balanced,0.04019733270009359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,1,128,balanced,0.042352000872294106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,1,128,balanced,0.04181866844495138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,128,balanced,0.04197333256403605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,128,balanced,0.04394133388996124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,128,balanced,0.045941332976023354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,128,balanced,0.04795733094215393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,128,balanced,0.0498879998922348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,128,balanced,0.05760533114274343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,128,balanced,0.06189866860707601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,128,balanced,0.07497066756089528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,128,balanced,0.08780266841252644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,128,balanced,0.10559999942779541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,128,balanced,0.14868266383806863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,128,balanced,0.25539199511210126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,1,8,power_law_1.2,0.06470400094985962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,1,8,power_law_1.2,0.07793920040130616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,1,8,power_law_1.2,0.06305279731750488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,1,8,power_law_1.2,0.06357120275497437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,1,8,power_law_1.2,0.06328960061073304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,1,8,power_law_1.2,0.06579840183258057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,1,8,power_law_1.2,0.06615039706230164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,1,8,power_law_1.2,0.06947839856147767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,1,8,power_law_1.2,0.07397119998931885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,1,8,power_law_1.2,0.07702400088310242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,1,8,power_law_1.2,0.07120640277862549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,1,8,power_law_1.2,0.07939199805259704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,1,8,power_law_1.2,0.08240000009536744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,1,8,power_law_1.2,0.08798080086708068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,1,8,power_law_1.2,0.09846400022506714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,1,8,power_law_1.2,0.11870720386505126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,1,8,power_law_1.2,0.13058559894561766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,1,8,power_law_1.2,0.17055360078811646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,1,8,power_law_1.2,0.17601920366287233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,1,8,power_law_1.2,0.24019839763641357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,1,8,power_law_1.2,0.2944191932678223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,1,8,power_law_1.2,0.37411201000213623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,1,8,power_law_1.2,0.5280064105987549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,1,8,power_law_1.2,0.7610047817230224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,1,8,power_law_1.2,0.967801570892334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,1,8,power_law_1.2,1.432697582244873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,1,8,power_law_1.2,1.910163116455078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,1,8,power_law_1.2,2.392915153503418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,1,8,power_law_1.2,3.898406219482422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,1,balanced,0.03961600114901861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,1,balanced,0.038160001238187156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,1,balanced,0.04257066547870636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,1,balanced,0.06132799883683523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,1,balanced,0.08699733018875122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,1,balanced,0.08702400326728821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,1,balanced,0.0890880028406779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,1,balanced,0.09032000104586284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,1,balanced,0.09299733241399129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,1,balanced,0.09397332866986592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,1,balanced,0.09498666723569234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,1,balanced,0.09872532884279887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,1,balanced,0.10268266995747884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,1,balanced,0.10857599973678589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,1,balanced,0.11720533172289531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,1,balanced,0.11998400092124939
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,1,balanced,0.13037332892417908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,1,balanced,0.1463466684023539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,1,balanced,0.16491732994715372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,1,balanced,0.20323199033737183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,1,balanced,0.23293866713841757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,1,balanced,0.33076266447703045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,1,balanced,0.40249598026275635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,1,8,power_law_1.2,7.41717758178711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,1,balanced,0.5865600109100342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,1,balanced,0.7508959770202637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,1,balanced,1.0964639981587727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,1,balanced,1.433616002400716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,1,balanced,1.7831039428710938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,1,balanced,2.8253278732299805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,1,balanced,5.511674880981445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,16,power_law_1.01,0.07400320172309875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,16,power_law_1.01,0.10093439817428589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,16,power_law_1.01,0.1210368037223816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,16,power_law_1.01,0.06125440001487732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,16,power_law_1.01,0.07702400088310242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,16,power_law_1.01,0.09220479726791382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,16,power_law_1.01,0.09501439929008484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,16,power_law_1.01,0.10312319993972778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,16,power_law_1.01,0.10505599975585937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,16,power_law_1.01,0.1076416015625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,16,power_law_1.01,0.11075839996337891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,16,power_law_1.01,0.11797759532928467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,16,power_law_1.01,0.11665279865264892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,16,power_law_1.01,0.11894400119781494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,16,power_law_1.01,0.12836480140686035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,16,power_law_1.01,0.1324928045272827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,16,power_law_1.01,0.1410367965698242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,16,power_law_1.01,0.16458239555358886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,16,power_law_1.01,0.16539520025253296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,16,power_law_1.01,0.19959039688110353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,16,power_law_1.01,0.22891519069671631
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,16,power_law_1.01,0.2926016092300415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,16,power_law_1.01,0.3326848030090332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,16,power_law_1.01,0.4526976108551025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,16,power_law_1.01,0.5435135841369629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,16,power_law_1.01,0.7793087959289551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,16,power_law_1.01,1.0249600410461426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,16,power_law_1.01,1.209823989868164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,16,power_law_1.01,1.871072006225586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,16,power_law_1.01,3.6986110687255858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,4,balanced,0.038949333131313324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,4,balanced,0.037685332198937736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,4,balanced,0.037418665985266365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,4,balanced,0.04158399999141693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,4,balanced,0.048341333866119385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,4,balanced,0.049786667029062905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,4,balanced,0.04831466575463613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,4,balanced,0.04967466493447622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,4,balanced,0.05013866722583771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,4,balanced,0.049733335773150124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,4,balanced,0.05077866713205973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,4,balanced,0.0516533354918162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,4,balanced,0.053488001227378845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,4,balanced,0.05226666728655497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,4,balanced,0.05824000140031179
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,4,balanced,0.05832533538341522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,4,balanced,0.060517330964406334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,4,balanced,0.06599999964237213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,4,balanced,0.07051200171311696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,4,balanced,0.0839359958966573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,4,balanced,0.09566400448481242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,4,balanced,0.1288106640179952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,4,balanced,0.15015467007954916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,4,balanced,0.203274667263031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,4,balanced,0.26150399446487427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,4,balanced,0.3688640197118123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,4,balanced,0.4700213273366292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,4,balanced,0.5819520155588785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,4,balanced,0.8977813720703125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,4,balanced,1.7492693265279133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,1,8,power_law_1.01,0.08238080143928528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,1,8,power_law_1.01,0.0845632016658783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,1,8,power_law_1.01,0.08497920036315917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,1,8,power_law_1.01,0.08193280100822449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,1,8,power_law_1.01,0.08201599717140198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,1,8,power_law_1.01,0.08540160059928895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,1,8,power_law_1.01,0.09283199906349182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,1,8,power_law_1.01,0.11708159446716308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,1,8,power_law_1.01,0.12791039943695068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,1,8,power_law_1.01,0.14591360092163086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,1,8,power_law_1.01,0.1292672038078308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,1,8,power_law_1.01,0.16366080045700074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,1,8,power_law_1.01,0.1778880000114441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,1,8,power_law_1.01,0.12936960458755492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,1,8,power_law_1.01,0.14585599899291993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,1,8,power_law_1.01,0.15110399723052978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,1,8,power_law_1.01,0.17847039699554443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,1,8,power_law_1.01,0.22541439533233643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,1,8,power_law_1.01,0.24959359169006348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,1,8,power_law_1.01,0.3517632007598877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,1,8,power_law_1.01,0.3966847896575928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,1,8,power_law_1.01,0.556224012374878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,1,8,power_law_1.01,0.7552127838134766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,1,8,power_law_1.01,1.2057536125183106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,1,8,power_law_1.01,1.2772607803344727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,1,8,power_law_1.01,2.027168083190918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,1,8,power_law_1.01,2.7397823333740234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,1,8,power_law_1.01,3.734566497802734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,1,8,power_law_1.01,5.8675392150878904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,1,8,power_law_1.01,10.582726287841798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,1,power_law_1.01,0.04835839867591858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,1,power_law_1.01,0.0615231990814209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,1,power_law_1.01,0.08081279993057251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,1,power_law_1.01,0.12995200157165526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,1,power_law_1.01,0.18481919765472413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,1,power_law_1.01,0.2874943971633911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,1,power_law_1.01,0.3142591953277588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,1,power_law_1.01,0.3253376007080078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,1,power_law_1.01,0.34869120121002195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,1,power_law_1.01,0.3401792049407959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,1,power_law_1.01,0.3429951906204224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,1,power_law_1.01,0.35688319206237795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,1,power_law_1.01,0.36888320446014405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,1,power_law_1.01,0.37378559112548826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,1,power_law_1.01,0.3921792030334473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,1,power_law_1.01,0.40274558067321775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,1,power_law_1.01,0.4206528186798096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,1,power_law_1.01,0.4753215789794922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,1,power_law_1.01,0.4666816234588623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,1,power_law_1.01,0.5234879970550537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,1,power_law_1.01,0.5969024181365967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,1,power_law_1.01,0.7040832042694092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,1,power_law_1.01,0.7331967830657959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,1,power_law_1.01,0.883296012878418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,1,power_law_1.01,1.0078911781311035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,1,power_law_1.01,1.331884765625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,1,power_law_1.01,1.6528703689575195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,128,power_law_1.2,0.06092159748077393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,1,1,power_law_1.01,1.9681087493896485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,128,power_law_1.2,0.05724160075187683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,1,1,power_law_1.01,2.9527936935424806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,128,power_law_1.2,0.048127999901771544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,128,power_law_1.2,0.0514240026473999
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,1,1,power_law_1.01,5.550156784057617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,128,power_law_1.2,0.05008640289306641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,128,power_law_1.2,0.052774399518966675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,128,power_law_1.2,0.05364480018615723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,128,power_law_1.2,0.053504002094268796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,128,power_law_1.2,0.052076798677444455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,128,power_law_1.2,0.0541055977344513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,128,power_law_1.2,0.05802879929542541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,128,power_law_1.2,0.06308479905128479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,4,power_law_1.2,0.04020479917526245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,4,power_law_1.2,0.035174399614334106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,4,power_law_1.2,0.03585279881954193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,128,power_law_1.2,0.06616320013999939
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,4,power_law_1.2,0.036985599994659425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,4,power_law_1.2,0.038771200180053714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,4,power_law_1.2,0.044633600115776065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,128,power_law_1.2,0.06906880140304565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,4,power_law_1.2,0.04614399969577789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,4,power_law_1.2,0.04679040014743805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,4,power_law_1.2,0.047328001260757445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,128,power_law_1.2,0.07871999740600585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,4,power_law_1.2,0.047916799783706665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,4,power_law_1.2,0.04965760111808777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,4,power_law_1.2,0.0507968008518219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,128,power_law_1.2,0.07899519801139832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,4,power_law_1.2,0.052025598287582395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,4,power_law_1.2,0.05479679703712463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,4,power_law_1.2,0.06046720147132874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,128,power_law_1.2,0.08771839737892151
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,4,power_law_1.2,0.062015998363494876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,4,power_law_1.2,0.06302719712257385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,4,power_law_1.2,0.07384960055351257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,128,power_law_1.2,0.10580480098724365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,4,power_law_1.2,0.0824895977973938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,4,power_law_1.2,0.10451840162277222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,4,power_law_1.2,0.11781760454177856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,128,power_law_1.2,0.1108672022819519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,4,power_law_1.2,0.1624511957168579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,4,power_law_1.2,0.1895359992980957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,128,power_law_1.2,0.13653759956359862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,4,power_law_1.2,0.25804159641265867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,4,power_law_1.2,0.3911616086959839
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,128,power_law_1.2,0.15434240102767943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,4,power_law_1.2,0.5329023838043213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,4,power_law_1.2,0.6823488235473633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,128,power_law_1.2,0.1954751968383789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,4,power_law_1.2,0.947481632232666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,128,power_law_1.2,0.2024832010269165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,4,power_law_1.2,1.4616127967834474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,128,power_law_1.2,0.28225278854370117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,4,power_law_1.2,3.134860801696777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,128,power_law_1.2,0.3374079942703247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,128,power_law_1.2,0.5207871913909912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,128,power_law_1.2,0.6109824180603027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,2,128,power_law_1.2,0.6966207981109619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,8,power_law_1.2,0.06752640008926392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,8,power_law_1.2,0.05609599947929382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,8,power_law_1.2,0.05716480016708374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,8,power_law_1.2,0.06677759885787964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,2,128,power_law_1.2,1.2903743743896485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,8,power_law_1.2,0.07055360078811646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,8,power_law_1.2,0.08788480162620545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,8,power_law_1.2,0.088755202293396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,8,power_law_1.2,0.0925055980682373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,8,power_law_1.2,0.09333119988441467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,8,power_law_1.2,0.0943615972995758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,8,power_law_1.2,0.09587839841842652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,8,power_law_1.2,0.09695360064506531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,8,power_law_1.2,0.09942399859428405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,8,power_law_1.2,0.105132794380188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,8,power_law_1.2,0.11261440515518188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,2,128,power_law_1.2,3.0340991973876954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,8,power_law_1.2,0.115392005443573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,8,power_law_1.2,0.12277760505676269
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,8,power_law_1.2,0.14105600118637085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,8,power_law_1.2,0.15245440006256103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,8,power_law_1.2,0.19877760410308837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,8,power_law_1.2,0.20430080890655516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,8,power_law_1.2,0.24895360469818115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,8,power_law_1.2,0.33838720321655275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,8,power_law_1.2,0.4376063823699951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,8,power_law_1.2,0.5277376174926758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,8,power_law_1.2,0.7952191829681396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,8,power_law_1.2,1.0287039756774903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,8,power_law_1.2,1.3294272422790527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,8,power_law_1.2,2.0496320724487305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,8,power_law_1.2,5.393452835083008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,32,power_law_1.01,0.08179200291633607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,32,power_law_1.01,0.08606079816818238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,32,power_law_1.01,0.05516160130500793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,32,power_law_1.01,0.06392319798469544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,32,power_law_1.01,0.0735040009021759
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,32,power_law_1.01,0.0756608009338379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,32,power_law_1.01,0.0838591992855072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,32,power_law_1.01,0.08181120157241821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,32,power_law_1.01,0.08263040184974671
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,32,power_law_1.01,0.08360959887504578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,32,power_law_1.01,0.08483200073242188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,32,power_law_1.01,0.08666880130767822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,32,power_law_1.01,0.08561279773712158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,32,power_law_1.01,0.09136639833450318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,32,power_law_1.01,0.10002559423446655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,32,power_law_1.01,0.1003648042678833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,32,power_law_1.01,0.10858880281448365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,32,power_law_1.01,0.11808639764785767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,32,power_law_1.01,0.13902080059051514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,32,power_law_1.01,0.16494079828262329
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,32,power_law_1.01,0.18442239761352539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,32,power_law_1.01,0.24815359115600585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,32,power_law_1.01,0.27404160499572755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,32,power_law_1.01,0.38399999141693114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,32,power_law_1.01,0.47886080741882325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,32,power_law_1.01,0.6300415992736816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,32,power_law_1.01,0.8201791763305664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,32,power_law_1.01,1.1210047721862793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,32,power_law_1.01,1.6236352920532227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,32,power_law_1.01,3.219833755493164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,32,power_law_1.2,0.03614720106124878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,32,power_law_1.2,0.03333120048046112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,32,power_law_1.2,0.030771198868751525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,32,power_law_1.2,0.030579200387001036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,32,power_law_1.2,0.030432000756263733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,32,power_law_1.2,0.030707201361656188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,32,power_law_1.2,0.031020799279212953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,32,power_law_1.2,0.031193599104881287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,32,power_law_1.2,0.0323199987411499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,32,power_law_1.2,0.03215999901294708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,32,power_law_1.2,0.03208959996700287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,32,power_law_1.2,0.033913600444793704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,32,power_law_1.2,0.033932799100875856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,32,power_law_1.2,0.0353408008813858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,32,power_law_1.2,0.03944959938526153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,32,power_law_1.2,0.03993600010871887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,32,power_law_1.2,0.04181120097637177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,32,power_law_1.2,0.046419200301170346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,32,power_law_1.2,0.05148800015449524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,32,power_law_1.2,0.05841919779777527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,32,power_law_1.2,0.060499197244644164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,32,power_law_1.2,0.07704960107803345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,32,power_law_1.2,0.08994560241699219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,32,power_law_1.2,0.13178240060806273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,32,power_law_1.2,0.16744320392608641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,32,power_law_1.2,0.23804159164428712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,32,power_law_1.2,0.28485760688781736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,2,32,power_law_1.2,0.3693887948989868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,2,32,power_law_1.2,0.5543744087219238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,2,32,power_law_1.2,1.2041279792785644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,1,1,power_law_1.01,0.0729856014251709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,1,1,power_law_1.01,0.11182719469070435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,1,1,power_law_1.01,0.10580480098724365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,1,1,power_law_1.01,0.15164799690246583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,1,1,power_law_1.01,0.15852160453796388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,1,1,power_law_1.01,0.16124800443649293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,1,1,power_law_1.01,0.16243840456008912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,1,1,power_law_1.01,0.16387200355529785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,1,1,power_law_1.01,0.16518399715423585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,1,1,power_law_1.01,0.16636799573898314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,1,1,power_law_1.01,0.17256319522857666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,1,1,power_law_1.01,0.17644799947738649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,1,1,power_law_1.01,0.1828160047531128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,1,1,power_law_1.01,0.1873792052268982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,1,1,power_law_1.01,0.20701439380645753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,1,1,power_law_1.01,0.21225600242614745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,1,1,power_law_1.01,0.22970240116119384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,1,1,power_law_1.01,0.26511359214782715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,1,1,power_law_1.01,0.30139520168304446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,1,1,power_law_1.01,0.38638079166412354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,1,1,power_law_1.01,0.46858878135681153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,1,1,power_law_1.01,0.6204800128936767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,1,1,power_law_1.01,0.7691008090972901
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,1,1,power_law_1.01,1.1111680030822755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,1,1,power_law_1.01,1.4398207664489746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,1,1,power_law_1.01,2.1373695373535155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,1,1,power_law_1.01,2.6531967163085937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,1,1,power_law_1.01,3.298643112182617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,1,1,power_law_1.01,4.993292617797851
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,1,1,power_law_1.01,9.754502105712891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,1,power_law_1.2,0.07201279997825623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,1,power_law_1.2,0.08087679743766785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,1,power_law_1.2,0.09497600197792053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,1,power_law_1.2,0.12611839771270753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,1,power_law_1.2,0.16201599836349487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,1,power_law_1.2,0.2025279998779297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,1,power_law_1.2,0.28993279933929444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,1,power_law_1.2,0.30801279544830323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,1,power_law_1.2,0.31665279865264895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,1,power_law_1.2,0.33514881134033203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,1,power_law_1.2,0.34355840682983396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,1,power_law_1.2,0.35068159103393554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,1,power_law_1.2,0.358572793006897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,1,power_law_1.2,0.3714560031890869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,1,power_law_1.2,0.38086400032043455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,1,power_law_1.2,0.3966592073440552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,1,power_law_1.2,0.42458882331848147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,1,power_law_1.2,0.4860544204711914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,1,power_law_1.2,0.5241407871246337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,1,power_law_1.2,0.6273151874542237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,1,power_law_1.2,0.6925119876861572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,1,power_law_1.2,0.8725503921508789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,1,power_law_1.2,1.0525888442993163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,1,power_law_1.2,1.4220352172851562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,1,power_law_1.2,1.775334358215332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,1,power_law_1.2,2.4925312042236327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,1,power_law_1.2,3.2146240234375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,1,power_law_1.2,3.9293182373046873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,1,power_law_1.2,6.032870483398438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,1,power_law_1.2,11.716786956787109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,2,balanced,0.057215998570124306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,2,balanced,0.07612800101439159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,2,balanced,0.11851200461387634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,2,balanced,0.2026080091794332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,2,balanced,0.3705973227818807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,2,balanced,0.3746346632639567
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,2,balanced,0.3739946683247884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,2,balanced,0.3763893445332845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,2,balanced,0.37728532155354816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,2,balanced,0.38229334354400635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,2,balanced,0.3805973529815674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,2,balanced,0.3803040186564128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,2,balanced,0.3838346799214681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,2,balanced,0.38751999537150067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,2,balanced,0.39153067270914715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,2,balanced,0.396448016166687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,2,balanced,0.40411198139190674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,2,balanced,0.4204053481419881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,2,balanced,0.4438240130742391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,2,balanced,0.46667734781901044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,2,balanced,0.49061334133148193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,2,balanced,0.594927986462911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,2,balanced,0.6137119928995768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,2,balanced,0.8208373387654623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,2,balanced,0.9074292977650961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,2,balanced,1.2378719647725422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,2,balanced,1.5948266983032227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,2,balanced,1.9529919624328613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,2,balanced,3.031440099080404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,2,balanced,5.9442933400472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,8,4,power_law_1.2,0.040505599975585935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,8,4,power_law_1.2,0.038201600313186646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,8,4,power_law_1.2,0.03752320110797882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,8,4,power_law_1.2,0.037215998768806456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,8,4,power_law_1.2,0.0366784006357193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,8,4,power_law_1.2,0.037747201323509214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,8,4,power_law_1.2,0.03802880048751831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,8,4,power_law_1.2,0.03828479945659637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,8,4,power_law_1.2,0.03947519958019256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,8,4,power_law_1.2,0.04206719994544983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,8,4,power_law_1.2,0.041555199027061465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,8,4,power_law_1.2,0.0440447986125946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,8,4,power_law_1.2,0.04540160000324249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,8,4,power_law_1.2,0.051712000370025636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,8,4,power_law_1.2,0.058412802219390866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,8,4,power_law_1.2,0.06518399715423584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,8,4,power_law_1.2,0.06236799955368042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,8,4,power_law_1.2,0.07477759718894958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,8,4,power_law_1.2,0.08083199858665466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,8,4,power_law_1.2,0.11617920398712159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,8,4,power_law_1.2,0.13144960403442382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,8,4,power_law_1.2,0.16543359756469728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,8,4,power_law_1.2,0.21305599212646484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,8,4,power_law_1.2,0.272377610206604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,8,4,power_law_1.2,0.339136004447937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,8,4,power_law_1.2,0.5710783958435058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,8,4,power_law_1.2,0.7509056091308594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,8,4,power_law_1.2,0.8262528419494629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,8,4,power_law_1.2,1.2767680168151856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,8,4,power_law_1.2,2.102956771850586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,32,4,power_law_1.2,0.03532800078392029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,32,4,power_law_1.2,0.030880001187324525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,32,4,power_law_1.2,0.031027200818061828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,32,4,power_law_1.2,0.03086079955101013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,32,4,power_law_1.2,0.031231999397277832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,32,4,power_law_1.2,0.03274880051612854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,32,4,power_law_1.2,0.0331712007522583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,32,4,power_law_1.2,0.033932799100875856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,8,balanced,0.047024001677831016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,32,4,power_law_1.2,0.03528960049152374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,8,balanced,0.047983999053637184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,8,balanced,0.052789335449536644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,32,4,power_law_1.2,0.034944000840187076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,8,balanced,0.07717866698900859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,8,balanced,0.10176533460617065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,32,4,power_law_1.2,0.03553920090198517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,8,balanced,0.11356799801190694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,8,balanced,0.11417067050933838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,32,4,power_law_1.2,0.037062400579452516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,8,balanced,0.11880000432332356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,8,balanced,0.11664000153541565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,32,4,power_law_1.2,0.0381632000207901
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,8,balanced,0.11665067076683044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,8,balanced,0.11846400300661723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,32,4,power_law_1.2,0.04035840034484863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,8,balanced,0.11925333738327026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,8,balanced,0.11900267004966736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,32,4,power_law_1.2,0.04446719884872437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,8,balanced,0.1199733316898346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,8,balanced,0.12639466921488443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,32,4,power_law_1.2,0.0469760000705719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,8,balanced,0.12591999769210815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,8,balanced,0.12933866182963052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,32,4,power_law_1.2,0.05134080052375793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,8,balanced,0.1370186706384023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,8,balanced,0.1418293317159017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,32,4,power_law_1.2,0.05775359869003296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,8,balanced,0.16529599825541177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,8,balanced,0.16931732495625815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,32,4,power_law_1.2,0.06865280270576476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,8,balanced,0.22033067544301352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,8,balanced,0.23831466833750406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,32,4,power_law_1.2,0.08592000007629394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,8,balanced,0.3553599913914998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,8,balanced,0.35356799761454266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,32,4,power_law_1.2,0.11606400012969971
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,8,balanced,0.4812053442001343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,32,4,power_law_1.2,0.16649600267410278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,8,balanced,0.6172053416570028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,32,4,power_law_1.2,0.2188096046447754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,8,balanced,0.7069119612375895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,32,4,power_law_1.2,0.2751104116439819
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,8,balanced,1.081781307856242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,32,4,power_law_1.2,0.319270396232605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,8,balanced,2.0300587018330893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,32,4,power_law_1.2,0.5446400165557861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,32,4,power_law_1.2,0.6399936199188232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,32,4,power_law_1.2,0.8700991630554199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,32,4,power_law_1.2,1.238815975189209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,32,4,power_law_1.2,2.8831808090209963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,4,power_law_1.2,0.055206400156021115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,4,power_law_1.2,0.04044159948825836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,4,power_law_1.2,0.046911999583244324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,4,power_law_1.2,0.06166399717330932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,4,power_law_1.2,0.08039039969444275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,4,power_law_1.2,0.10657919645309448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,4,power_law_1.2,0.1134335994720459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,4,power_law_1.2,0.11075199842453003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,4,power_law_1.2,0.11784960031509399
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,4,power_law_1.2,0.1148800015449524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,4,power_law_1.2,0.11839359998703003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,4,power_law_1.2,0.1213312029838562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,4,2,balanced,0.04008533308903376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,4,2,balanced,0.039690665900707245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,4,2,balanced,0.05036266644795736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,4,2,balanced,0.04833066463470459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,4,2,balanced,0.048170665899912514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,4,power_law_1.2,0.12585599422454835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,4,2,balanced,0.05022933085759481
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,4,2,balanced,0.04962133367856344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,4,2,balanced,0.04975466430187225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,4,2,balanced,0.0495306650797526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,4,2,balanced,0.049253334601720176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,4,2,balanced,0.05179200073083242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,4,power_law_1.2,0.1303231954574585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,4,2,balanced,0.05283733208974203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,4,2,balanced,0.05306133131186167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,4,2,balanced,0.061194668213526406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,4,2,balanced,0.06878399848937988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,4,2,balanced,0.073253333568573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,4,2,balanced,0.0697920024394989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,4,power_law_1.2,0.14449280500411987
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,4,2,balanced,0.07991466422875722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,4,2,balanced,0.08582400282224019
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,4,2,balanced,0.10073066751162212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,4,2,balanced,0.1178986628850301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,4,2,balanced,0.1482080022493998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,4,power_law_1.2,0.15312639474868775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,4,2,balanced,0.17942400773366293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,4,2,balanced,0.24080532789230347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,4,2,balanced,0.2871946692466736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,4,2,balanced,0.4076480070749919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,4,power_law_1.2,0.15032960176467897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,4,2,balanced,0.5189226468404134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,1,4,balanced,0.02884799987077713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,1,4,balanced,0.029637334247430164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,4,2,balanced,0.6341226498285929
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,1,4,balanced,0.03166399896144867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,1,4,balanced,0.05166399975617727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,1,4,balanced,0.07688533266385396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,1,4,balanced,0.07684266567230225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,1,4,balanced,0.07828266421953838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,4,2,balanced,0.9645546277364095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,1,4,balanced,0.07754133145014445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,4,power_law_1.2,0.17418240308761596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,1,4,balanced,0.07772266864776611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,1,4,balanced,0.07841599980990092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,1,4,balanced,0.07689600189526875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,1,4,balanced,0.07970133423805237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,1,4,balanced,0.0801333338022232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,1,4,balanced,0.08000533282756805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,1,4,balanced,0.08598933617273967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,4,2,balanced,1.8751519521077473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,1,4,balanced,0.08468799789746602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,1,4,balanced,0.08946133653322856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,1,4,balanced,0.0922933320204417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,4,power_law_1.2,0.18621439933776857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,4,balanced,0.09429867068926494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,4,balanced,0.10938666264216106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,4,balanced,0.11502933502197266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,4,balanced,0.14034133156140646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,4,balanced,0.15470932920773825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,4,balanced,0.21177067359288534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,4,balanced,0.23575466871261597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,4,balanced,0.31563733021418255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,4,power_law_1.2,0.2465343952178955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,4,balanced,0.3989226818084717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,4,balanced,0.47731200853983563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,4,balanced,0.7273333072662354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,4,power_law_1.2,0.25523200035095217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,4,balanced,1.37827730178833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,4,power_law_1.2,0.310809588432312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,4,power_law_1.2,0.385152006149292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,4,power_law_1.2,0.5516223907470703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,4,power_law_1.2,0.6188864231109619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,4,power_law_1.2,1.0298239707946777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,4,power_law_1.2,1.4132543563842774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,4,power_law_1.2,1.5950400352478027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,4,power_law_1.2,2.349760055541992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,1,power_law_1.01,0.037939199805259706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,8,balanced,0.029525332152843475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,8,balanced,0.029781334102153778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,8,balanced,0.03126399964094162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,8,balanced,0.029440000653266907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,8,balanced,0.03491200009981791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,8,balanced,0.03594133257865906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,8,balanced,0.035375999907652535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,8,balanced,0.03551999976237615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,8,balanced,0.03566399961709976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,8,balanced,0.035375999907652535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,8,balanced,0.03737599899371465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,8,balanced,0.03711999952793121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,8,balanced,0.03644266724586487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,8,balanced,0.03938666731119156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,8,balanced,0.04159466673930486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,8,balanced,0.04151466737190882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,4,power_law_1.2,4.344543838500977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,8,balanced,0.04389866689840952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,8,balanced,0.04584000011285146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,8,balanced,0.05420266588528951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,8,balanced,0.05797866483529409
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,8,balanced,0.06444799900054932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,8,balanced,0.08411199847857158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,8,balanced,0.10603732864061992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,1,power_law_1.01,0.04602240025997162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,8,balanced,0.13140799601872763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,8,balanced,0.1541759967803955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,8,balanced,0.21967999140421549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,8,balanced,0.2624906698862712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,8,balanced,0.3309173385302226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,8,balanced,0.4827253421147664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,8,balanced,0.8966186841328939
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,1,power_law_1.01,0.06876800060272217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,1,power_law_1.01,0.0938368022441864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,1,power_law_1.01,0.11571840047836304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,1,power_law_1.01,0.18544000387191772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,1,power_law_1.01,0.18852479457855226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,1,power_law_1.01,0.19975039958953858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,1,power_law_1.01,0.20193920135498047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,1,power_law_1.01,0.20635519027709961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,1,power_law_1.01,0.21310720443725586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,1,power_law_1.01,0.22147200107574463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,1,power_law_1.01,0.22329599857330323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,1,power_law_1.01,0.23237121105194092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,1,power_law_1.01,0.24303359985351564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,1,power_law_1.01,0.2545919895172119
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,1,power_law_1.01,0.26703360080718996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,4,power_law_1.01,0.051475197076797485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,1,power_law_1.01,0.3002880096435547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,4,power_law_1.01,0.0663424015045166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,1,power_law_1.01,0.33628799915313723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,4,power_law_1.01,0.044998401403427125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,4,power_law_1.01,0.061260801553726194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,1,power_law_1.01,0.4208064079284668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,1,power_law_1.01,0.08597760200500489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,4,power_law_1.01,0.07360640168190002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,1,power_law_1.01,0.4786687850952148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,4,power_law_1.01,0.10548479557037353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,4,power_law_1.01,0.10705280303955078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,1,power_law_1.01,0.09260159730911255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,1,power_law_1.01,0.6115200042724609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,4,power_law_1.01,0.10872960090637207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,4,power_law_1.01,0.10998400449752807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,1,power_law_1.01,0.7436223983764648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,1,power_law_1.01,0.12087039947509766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,4,power_law_1.01,0.11465599536895751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,4,power_law_1.01,0.11682560443878173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,1,power_law_1.01,1.0219072341918944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,1,power_law_1.01,0.14238719940185546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,4,power_law_1.01,0.11749759912490845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,4,power_law_1.01,0.1180799961090088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,1,power_law_1.01,1.2588095664978027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,4,power_law_1.01,0.12429440021514893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,1,power_law_1.01,0.20349440574645997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,4,power_law_1.01,0.12362240552902222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,1,power_law_1.01,1.7798271179199219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,4,power_law_1.01,0.12804479598999025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,1,power_law_1.01,0.2723968029022217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,4,power_law_1.01,0.14280320405960084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,4,power_law_1.01,0.16264959573745727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,1,power_law_1.01,0.3101567983627319
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,1,power_law_1.01,2.296249580383301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,4,power_law_1.01,0.15429760217666627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,4,power_law_1.01,0.18613760471343993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,1,power_law_1.01,0.38706560134887696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,4,power_law_1.01,0.19624320268630982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,1,power_law_1.01,2.8209280014038085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,4,power_law_1.01,0.2458048105239868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,1,power_law_1.01,0.3938688039779663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,4,power_law_1.01,0.2448064088821411
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,4,power_law_1.01,0.30358400344848635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,1,power_law_1.01,0.40316162109375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,4,power_law_1.01,0.35574400424957275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,1,power_law_1.01,4.327360153198242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,4,power_law_1.01,0.45525760650634767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,1,power_law_1.01,0.43213438987731934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,4,power_law_1.01,0.6199808120727539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,1,power_law_1.01,0.4409023761749268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,1,4,power_law_1.01,0.7577792167663574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,1,power_law_1.01,0.448089599609375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,1,4,power_law_1.01,1.0147392272949218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,1,power_law_1.01,8.379583740234375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,1,power_law_1.01,0.47470722198486326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,1,power_law_1.01,0.4926144123077393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,1,4,power_law_1.01,2.0641855239868163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,1,power_law_1.01,0.5048255920410156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,1,power_law_1.01,0.5264832019805908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,1,power_law_1.01,0.5785855770111084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,1,power_law_1.01,0.6306240081787109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,1,power_law_1.01,0.7320320129394531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,1,power_law_1.01,0.816044807434082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,1,power_law_1.01,1.004691219329834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,1,power_law_1.01,1.1960895538330079
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,8,balanced,0.03182400017976761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,8,balanced,0.03177600105603536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,8,balanced,0.03383466601371765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,8,balanced,0.03517866631348928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,8,balanced,0.057989334066708885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,8,balanced,0.05676266551017761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,8,balanced,0.05791999896367391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,8,balanced,0.05712000032265981
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,8,balanced,0.058005332946777344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,8,balanced,0.058517331878344216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,8,balanced,0.05749866863091787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,8,balanced,0.06039466460545858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,8,balanced,0.05860800047715505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,8,balanced,0.06065066655476888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,8,balanced,0.06534933547178905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,8,balanced,0.0661599983771642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,8,balanced,0.06699199974536896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,1,power_law_1.01,0.09074559807777405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,8,balanced,0.0707893321911494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,8,balanced,0.07311466832955678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,8,balanced,0.08660266796747844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,8,balanced,0.0905013382434845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,1,power_law_1.01,1.5707648277282715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,8,balanced,0.11020267009735107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,8,balanced,0.12523733576138815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,8,balanced,0.16776533921559653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,8,balanced,0.18739734093348184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,8,balanced,0.24740799268086752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,8,balanced,0.30592532952626544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,8,balanced,0.3726613521575928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,8,balanced,0.5529066721598307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,8,balanced,1.0176479816436768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,1,power_law_1.01,1.9564031600952148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,1,power_law_1.01,0.11537920236587525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,1,power_law_1.01,2.708755111694336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,1,power_law_1.01,0.14814079999923707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,1,power_law_1.01,3.4974590301513673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,1,power_law_1.01,0.17045120000839234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,1,power_law_1.01,0.2627455949783325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,1,power_law_1.01,4.299161529541015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,1,power_law_1.01,0.3787008047103882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,1,power_law_1.01,6.515711975097656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,1,power_law_1.01,0.4354239940643311
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,2,power_law_1.2,0.14572160243988036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,1,power_law_1.01,0.5466368198394775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,2,power_law_1.2,0.22926080226898193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,2,power_law_1.2,0.18462079763412476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,1,power_law_1.01,0.5807487964630127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,2,power_law_1.2,0.3309184074401855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,1,power_law_1.01,12.625382232666016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,2,power_law_1.2,0.48172798156738283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,1,power_law_1.01,0.5834303855895996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,2,power_law_1.2,0.7003776073455811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,2,power_law_1.2,1.0753600120544433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,1,power_law_1.01,0.6242559909820556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,2,power_law_1.2,1.108902359008789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,1,power_law_1.01,0.6464896202087402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,2,power_law_1.2,1.1422783851623535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,2,power_law_1.2,1.2421055793762208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,1,power_law_1.01,0.6606847763061523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,2,power_law_1.2,1.2566399574279785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,2,power_law_1.2,1.3485055923461915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,1,power_law_1.01,0.6912191867828369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,2,power_law_1.2,1.3400575637817382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,1,power_law_1.01,0.7242752075195312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,2,power_law_1.2,1.42805757522583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,2,power_law_1.2,1.4658111572265624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,1,power_law_1.01,0.7343935966491699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,8,power_law_1.2,0.045132800936698914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,2,power_law_1.2,1.470847988128662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,8,power_law_1.2,0.054118400812149046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,1,power_law_1.01,0.7904575824737549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,8,power_law_1.2,0.04423680007457733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,2,power_law_1.2,1.655308723449707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,8,power_law_1.2,0.045747199654579164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,2,power_law_1.2,1.7738176345825196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,1,power_law_1.01,0.8776063919067383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,8,power_law_1.2,0.04878720045089722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,2,power_law_1.2,1.7916479110717773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,8,power_law_1.2,0.05414400100708008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,1,power_law_1.01,0.9466431617736817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,8,power_law_1.2,0.058847999572753905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,2,power_law_1.2,1.9846912384033204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,8,power_law_1.2,0.060031998157501223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,2,power_law_1.2,1.9427583694458008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,1,power_law_1.01,1.0827648162841796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,8,power_law_1.2,0.06178560256958008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,8,power_law_1.2,0.06250879764556885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,2,power_law_1.2,2.1883968353271483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,8,power_law_1.2,0.06591359972953796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,1,power_law_1.01,1.1433792114257812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,8,power_law_1.2,0.0680512011051178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,2,power_law_1.2,2.442515182495117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,8,power_law_1.2,0.07125120162963867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,1,power_law_1.01,1.3877568244934082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,8,power_law_1.2,0.0766207993030548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,2,power_law_1.2,3.1825727462768554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,8,power_law_1.2,0.08611199855804444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,1,power_law_1.01,1.6350656509399415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,8,power_law_1.2,0.09099519848823548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,2,power_law_1.2,3.2700927734375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,8,power_law_1.2,0.09598720073699951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,8,power_law_1.2,0.11541119813919068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,1,power_law_1.01,2.0982336044311523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,2,power_law_1.2,4.367948913574219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,8,power_law_1.2,0.12672640085220338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,8,power_law_1.2,0.15413119792938232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,16,power_law_1.01,0.03688960075378418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,1,power_law_1.01,2.539616012573242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,2,power_law_1.2,5.104403305053711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,16,power_law_1.01,0.030425599217414855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,8,power_law_1.2,0.19393279552459716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,16,power_law_1.01,0.02771199941635132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,8,power_law_1.2,0.23809919357299805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,16,power_law_1.01,0.02812800109386444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,16,power_law_1.01,0.02975359857082367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,16,power_law_1.01,0.0308351993560791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,8,power_law_1.2,0.302239990234375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,1,power_law_1.01,3.3964607238769533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,16,power_law_1.01,0.03086720108985901
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,2,power_law_1.2,6.553228759765625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,16,power_law_1.01,0.03113600015640259
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,8,power_law_1.2,0.4307712078094482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,16,power_law_1.01,0.031763198971748355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,16,power_law_1.01,0.031539198756217954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,8,power_law_1.2,0.5432896137237548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,16,power_law_1.01,0.0339711993932724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,16,power_law_1.01,0.03464959859848023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,1,power_law_1.01,4.167808151245117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,16,power_law_1.01,0.034835198521614076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,8,power_law_1.2,0.774348783493042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,16,power_law_1.01,0.03660799860954285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,2,power_law_1.2,8.517171478271484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,16,power_law_1.01,0.04032639861106872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,16,power_law_1.01,0.04092159867286682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,8,power_law_1.2,1.0598591804504394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,16,power_law_1.01,0.04211199879646301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,16,power_law_1.01,0.04612480103969574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,1,power_law_1.01,5.014720153808594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,16,power_law_1.01,0.04999679923057556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,8,power_law_1.2,1.3615488052368163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,16,power_law_1.01,0.05885440111160278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,16,power_law_1.01,0.06587520241737366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,16,power_law_1.01,0.08143360018730164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,8,power_law_1.2,1.9652992248535157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,16,power_law_1.01,0.09836159944534302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,16,power_law_1.01,0.14517120122909546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,2,power_law_1.2,17.234291076660156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,16,power_law_1.01,0.19101439714431762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,1,power_law_1.01,7.571942138671875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,16,power_law_1.01,0.2351232051849365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,16,power_law_1.01,0.33457920551300047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,8,power_law_1.2,4.404646301269532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,16,power_law_1.01,0.3929152011871338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,16,power_law_1.01,0.6230976104736328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,1,power_law_1.01,14.257369995117188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,16,power_law_1.01,1.164358425140381
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,16,power_law_1.2,0.09765120148658753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,16,power_law_1.2,0.15811840295791627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,16,power_law_1.2,0.06224640011787415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,128,power_law_1.01,0.07580159902572632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,16,power_law_1.2,0.07463039755821228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,128,power_law_1.01,0.07708160281181335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,16,power_law_1.2,0.0894976019859314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,16,balanced,0.03617066641648611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,16,balanced,0.036362667878468834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,128,power_law_1.01,0.0729088008403778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,16,balanced,0.03585600107908249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,16,balanced,0.03385066737731298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,16,balanced,0.03770666569471359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,16,balanced,0.038704000413417816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,16,balanced,0.03835733234882355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,16,balanced,0.0377866675456365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,16,power_law_1.2,0.0955136001110077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,16,balanced,0.03766400118668874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,16,balanced,0.038773333032925926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,16,balanced,0.03794133414824804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,16,balanced,0.04029333343108495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,16,balanced,0.03823466598987579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,16,balanced,0.03999999910593033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,16,balanced,0.04479999840259552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,128,power_law_1.01,0.0650111973285675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,32,balanced,0.0347680002450943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,16,balanced,0.043978666265805565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,32,balanced,0.033904001116752625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,16,balanced,0.04509866734345754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,32,balanced,0.031925333042939506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,16,balanced,0.04757333298524221
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,32,balanced,0.03403199960788091
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,32,balanced,0.035775999228159584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,16,balanced,0.05026133358478546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,32,balanced,0.033717334270477295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,32,balanced,0.03614933292071024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,16,balanced,0.052111998200416565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,16,power_law_1.2,0.12645119428634644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,32,balanced,0.0359253336985906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,32,balanced,0.036117332677046456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,16,balanced,0.05960000058015188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,32,balanced,0.0359253336985906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,16,balanced,0.06402133405208588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,32,balanced,0.037632000943024956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,32,balanced,0.03619199991226196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,16,balanced,0.07162133355935414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,32,balanced,0.03779733429352442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,32,balanced,0.037674665451049805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,16,balanced,0.09185066819190979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,32,balanced,0.04177066683769226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,128,power_law_1.01,0.06936320066452026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,32,balanced,0.04041599979003271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,16,balanced,0.1164959967136383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,32,balanced,0.041637333730856575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,32,balanced,0.04403733213742574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,32,balanced,0.04365866879622141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,16,balanced,0.14841066797574362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,32,balanced,0.04610666632652283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,32,balanced,0.04830400149027506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,16,balanced,0.18107734123865762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,32,balanced,0.053674668073654175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,32,balanced,0.05625600119431814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,16,balanced,0.2148426572481791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,16,power_law_1.2,0.1313024044036865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,32,balanced,0.07263466715812683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,32,balanced,0.08281066517035167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,16,balanced,0.30957865715026855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,32,balanced,0.11027200023333232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,32,balanced,0.1341546674569448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,16,balanced,0.5703893502553304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,32,balanced,0.15821333726247153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,128,power_law_1.01,0.06912000179290771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,32,balanced,0.2134880026181539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,32,balanced,0.3813440004984538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,16,power_law_1.2,0.13127039670944213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,128,power_law_1.01,0.07026559710502625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,16,power_law_1.2,0.12676479816436767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,128,power_law_1.01,0.07153919935226441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,16,power_law_1.2,0.13971840143203734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,128,power_law_1.01,0.07203199863433837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,16,power_law_1.2,0.13556480407714844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,128,power_law_1.01,0.07349119782447815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,16,power_law_1.2,0.15591039657592773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,128,power_law_1.01,0.07225599884986877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,16,power_law_1.2,0.15695359706878662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,128,power_law_1.01,0.073990398645401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,16,power_law_1.2,0.16547199487686157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,128,power_law_1.01,0.073471999168396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,16,power_law_1.2,0.18250240087509156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,128,power_law_1.01,0.07269759774208069
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,16,power_law_1.2,0.1650879979133606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,128,power_law_1.01,0.08033279776573181
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,16,power_law_1.2,0.17971199750900269
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,128,power_law_1.01,0.08392320275306701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,16,power_law_1.2,0.1965376019477844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,128,power_law_1.01,0.08617600202560424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,16,power_law_1.2,0.23686399459838867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,128,power_law_1.01,0.09941120147705078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,16,power_law_1.2,0.243123197555542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,128,power_law_1.01,0.10903680324554443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,32,power_law_1.2,0.0643392026424408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,128,power_law_1.01,0.13645440340042114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,16,power_law_1.2,0.28995840549468993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,32,power_law_1.2,0.05742080211639404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,128,power_law_1.01,0.16159360408782958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,16,power_law_1.2,0.35055999755859374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,32,power_law_1.2,0.04912000000476837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,128,power_law_1.01,0.21673600673675536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,16,power_law_1.2,0.4540863990783691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,32,power_law_1.2,0.04981119930744171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,128,power_law_1.01,0.2537791967391968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,16,power_law_1.2,0.5506239891052246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,32,power_law_1.2,0.05084159970283508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,128,power_law_1.01,0.34520959854125977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,32,power_law_1.2,0.051545602083206174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,16,power_law_1.2,0.8458304405212402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,128,power_law_1.01,0.43578238487243653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,32,power_law_1.2,0.05355520248413086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,32,power_law_1.2,0.05484799742698669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,16,power_law_1.2,1.0012224197387696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,128,power_law_1.01,0.6741055965423584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,32,power_law_1.2,0.05493119955062866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,32,power_law_1.2,0.05539839863777161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,16,power_law_1.2,1.2106304168701172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,128,power_law_1.01,0.8102720260620118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,32,power_law_1.2,0.05767040252685547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,128,power_law_1.01,1.01910400390625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,32,power_law_1.2,0.05779839754104614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,16,power_law_1.2,2.098086357116699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,32,power_law_1.2,0.05901439785957337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,32,power_law_1.2,0.05950719714164734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,128,power_law_1.01,1.7615743637084962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,32,power_law_1.2,0.06784639954566955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,32,power_law_1.2,0.07060480117797852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,16,power_law_1.2,3.3727745056152343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,32,power_law_1.2,0.07455360293388366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,32,power_law_1.2,0.08580480217933655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,128,power_law_1.01,3.114246368408203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,32,power_law_1.2,0.09514880180358887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,32,power_law_1.2,0.11754239797592163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,32,power_law_1.2,0.1338047981262207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,32,power_law_1.2,0.169868803024292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,32,power_law_1.2,0.20597760677337645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,32,power_law_1.2,0.27560958862304685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,32,power_law_1.2,0.3659840106964111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,32,power_law_1.2,0.4951807975769043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,32,power_law_1.2,0.6800511837005615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,4,balanced,0.03364799916744232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,4,balanced,0.03196800003449122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,4,balanced,0.035690667728583016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,4,balanced,0.039834665755430855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,4,balanced,0.06029866635799408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,4,balanced,0.060991997520128884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,4,balanced,0.0625439981619517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,4,balanced,0.06306133170922597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,4,balanced,0.0621919979651769
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,4,balanced,0.06436266501744588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,4,balanced,0.06460266808668773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,4,balanced,0.0683840016523997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,4,balanced,0.06912533442179362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,4,balanced,0.07165866593519847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,32,power_law_1.2,0.785427188873291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,4,balanced,0.07869866490364075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,4,balanced,0.07876266539096832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,4,balanced,0.08085866769154866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,4,balanced,0.08953600128491719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,4,balanced,0.0995199978351593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,4,balanced,0.11838933825492859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,4,balanced,0.13411733508110046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,4,16,balanced,0.03748266647259394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,4,16,balanced,0.03548266738653183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,4,balanced,0.17274133364359537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,4,16,balanced,0.03482666611671448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,4,balanced,0.20360533396402994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,4,16,balanced,0.035402665535608925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,4,16,balanced,0.03756266583998998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,4,balanced,0.2844533324241638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,4,16,balanced,0.037477334340413414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,4,16,balanced,0.03773866593837738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,4,balanced,0.3670186599095662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,4,16,balanced,0.037578667203585304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,4,16,balanced,0.03759466608365377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,4,16,balanced,0.03760000069936117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,4,balanced,0.5223626693089803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,4,16,balanced,0.03759466608365377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,4,16,balanced,0.03949866692225138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,4,16,balanced,0.037589333951473236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,4,16,balanced,0.03957866628964742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,4,balanced,0.6752160390218099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,4,16,balanced,0.043653334180514015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,4,16,balanced,0.04201066493988037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,4,16,balanced,0.04366933306058248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,4,balanced,0.8285973072052002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,4,16,balanced,0.04566933214664459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,4,16,balanced,0.04774933556715647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,4,16,balanced,0.04931733508904775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,4,16,balanced,0.056517332792282104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,4,16,balanced,0.062165334820747375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,4,balanced,1.283087968826294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,4,16,balanced,0.0666240006685257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,4,16,balanced,0.0825439989566803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,4,16,balanced,0.10847999652226765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,4,16,balanced,0.1400213340918223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,4,16,balanced,0.17061867316563925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,32,power_law_1.2,1.2574399948120116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,4,16,balanced,0.20720533529917398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,4,balanced,2.459130605061849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,4,16,balanced,0.2944906751314799
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,4,16,balanced,0.5403733253479004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,32,power_law_1.2,2.5671808242797853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,128,power_law_1.2,0.034790399670600894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,128,power_law_1.2,0.034560000896453856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,128,power_law_1.2,0.03543039858341217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,2,power_law_1.2,0.11800320148468017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,128,power_law_1.2,0.03498240113258362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,128,power_law_1.2,0.03503359854221344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,2,balanced,0.03602133442958196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,2,balanced,0.04164800047874451
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,2,balanced,0.04398400088151296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,2,balanced,0.055861334005991616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,2,balanced,0.08282666901747386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,2,balanced,0.08263466755549113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,2,balanced,0.08427733182907104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,2,balanced,0.08067733546098073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,128,power_law_1.2,0.035955199599266054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,2,balanced,0.07924800117810567
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,2,balanced,0.08036266764005025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,2,balanced,0.08103999992211659
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,2,balanced,0.08342933654785156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,2,power_law_1.2,0.11845760345458985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,2,balanced,0.08406399687131245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,2,balanced,0.08303999900817871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,2,balanced,0.09080533186594646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,2,balanced,0.09226133426030476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,2,balanced,0.09825066725413005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,2,balanced,0.10936533411343892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,128,power_law_1.2,0.03687680065631867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,2,balanced,0.11591999729474385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,2,balanced,0.13307199875513712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,2,balanced,0.14774933457374573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,2,balanced,0.21585599581400552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,2,balanced,0.25804799795150757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,2,balanced,0.3582613468170166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,128,power_law_1.2,0.0369024008512497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,2,balanced,0.4602400064468384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,2,balanced,0.6580213308334351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,2,balanced,0.8450986544291178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,128,power_law_1.2,0.036697599291801455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,2,balanced,1.0415039857228596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,2,power_law_1.2,0.15191680192947388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,2,balanced,1.6318079630533855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,128,power_law_1.2,0.03688960075378418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,2,balanced,3.2376267115275064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,128,power_law_1.2,0.036499199271202085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,128,power_law_1.2,0.03768320083618164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,2,power_law_1.2,0.2323456048965454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,128,power_law_1.2,0.03797760009765625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,128,power_law_1.2,0.03827199935913086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,128,power_law_1.2,0.041152000427246094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,2,power_law_1.2,0.30776960849761964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,128,power_law_1.2,0.041228801012039185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,128,power_law_1.2,0.04312320053577423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,128,power_law_1.2,0.04537599980831146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,2,power_law_1.2,0.5992896080017089
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,128,power_law_1.2,0.046489599347114566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,128,power_law_1.2,0.05406079888343811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,128,power_law_1.2,0.059411197900772095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,2,power_law_1.2,0.6574719905853271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,128,power_law_1.2,0.0689408004283905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,128,power_law_1.2,0.08569599986076355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,2,power_law_1.2,0.701145601272583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,128,power_law_1.2,0.11438720226287842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,128,power_law_1.2,0.13349759578704834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,2,power_law_1.2,0.6922688007354736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,128,power_law_1.2,0.18788479566574096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,128,power_law_1.2,0.2447295904159546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,2,power_law_1.2,0.7284031867980957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,128,power_law_1.2,0.3077248096466064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,2,power_law_1.2,0.7472832202911377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,128,power_law_1.2,0.4651328086853027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,2,power_law_1.2,0.7723328113555908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,2,power_law_1.2,0.7932864189147949
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,128,power_law_1.2,0.8972864151000977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,2,power_law_1.2,0.8133055686950683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,2,power_law_1.2,0.8646271705627442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,2,power_law_1.2,0.8850751876831054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,2,power_law_1.2,0.8926464080810547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,2,power_law_1.2,0.9804672241210938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,2,power_law_1.2,1.0688896179199219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,16,8,power_law_1.01,0.03187839984893799
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,16,8,power_law_1.01,0.033241599798202515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,16,8,power_law_1.01,0.029344001412391664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,16,8,power_law_1.01,0.029145601391792297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,16,8,power_law_1.01,0.029023998975753786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,16,8,power_law_1.01,0.030054399371147157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,16,8,power_law_1.01,0.030329599976539612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,16,8,power_law_1.01,0.03110400140285492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,2,power_law_1.2,1.2067456245422363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,16,8,power_law_1.01,0.03150080144405365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,16,8,power_law_1.01,0.03185279965400696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,16,8,power_law_1.01,0.032051199674606325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,16,8,power_law_1.01,0.03457280099391937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,16,8,power_law_1.01,0.03415679931640625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,16,8,power_law_1.01,0.03504000008106232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,16,8,power_law_1.01,0.042124798893928526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,16,8,power_law_1.01,0.04482559859752655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,16,8,power_law_1.01,0.04642559885978699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,2,power_law_1.2,1.348031997680664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,16,8,power_law_1.01,0.051692801713943484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,16,8,power_law_1.01,0.05616639852523804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,16,8,power_law_1.01,0.06912000179290771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,16,8,power_law_1.01,0.07522559762001038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,16,8,power_law_1.01,0.10177279710769653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,16,8,power_law_1.01,0.1153216004371643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,16,8,power_law_1.01,0.15608960390090942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,16,8,power_law_1.01,0.1793727993965149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,16,8,power_law_1.01,0.2612799882888794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,2,power_law_1.2,1.6506687164306642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,16,8,power_law_1.01,0.3528448104858398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,16,8,power_law_1.01,0.3830912113189697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,16,8,power_law_1.01,0.6329152107238769
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,16,8,power_law_1.01,1.210425567626953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,2,power_law_1.2,1.716806411743164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,2,power_law_1.2,2.0798208236694338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,2,power_law_1.2,2.5389568328857424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,2,power_law_1.2,3.6984832763671873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,4,power_law_1.01,0.0721343994140625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,2,power_law_1.2,4.205228805541992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,4,power_law_1.01,0.09321600198745728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,4,power_law_1.01,0.06843519806861878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,4,power_law_1.01,0.08794239759445191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,4,power_law_1.01,0.11975040435791015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,2,2,power_law_1.2,4.898329544067383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,4,power_law_1.01,0.13841279745101928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,4,power_law_1.01,0.18695679903030396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,4,power_law_1.01,0.19342080354690552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,4,power_law_1.01,0.19542399644851685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,4,power_law_1.01,0.20229759216308593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,4,power_law_1.01,0.2124608039855957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,2,2,power_law_1.2,7.472364807128907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,4,power_law_1.01,0.21328001022338866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,4,power_law_1.01,0.22181119918823242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,4,power_law_1.01,0.23420159816741942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,4,power_law_1.01,0.2501568078994751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,4,power_law_1.01,0.2462143898010254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,4,power_law_1.01,0.24697599411010743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,4,power_law_1.01,0.29164800643920896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,4,power_law_1.01,0.3006720066070557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,4,power_law_1.01,0.34962561130523684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,4,power_law_1.01,0.37862401008605956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,2,2,power_law_1.2,14.530636596679688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,4,power_law_1.01,0.468339204788208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,4,power_law_1.01,0.509388780593872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,4,power_law_1.01,0.6654463768005371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,4,power_law_1.01,0.7895167827606201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,4,power_law_1.01,1.063321590423584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,4,power_law_1.01,1.3396032333374024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,4,power_law_1.01,1.6847103118896485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,4,power_law_1.01,2.309907150268555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,32,1,balanced,0.03969600051641464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,32,1,balanced,0.03755199909210205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,32,1,balanced,0.04159999887148539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,32,1,balanced,0.042538667718569435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,32,1,balanced,0.04154666761557261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,32,1,balanced,0.0436160018046697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,32,1,balanced,0.04307733476161957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,32,1,balanced,0.04383466641108195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,32,1,balanced,0.04365866879622141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,32,1,balanced,0.04382933179537455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,32,1,balanced,0.04568533102671305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,32,1,balanced,0.04599999884764353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,32,1,balanced,0.048341333866119385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,32,1,balanced,0.04929066697756449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,32,1,balanced,0.055957332253456116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,32,1,balanced,0.058373332023620605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,32,1,balanced,0.06010666489601135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,32,1,balanced,0.07500799993673961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,32,1,balanced,0.08662933111190796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,32,1,balanced,0.10673600435256958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,32,1,balanced,0.12338133653004964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,32,1,balanced,0.16455466548601785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,32,1,balanced,0.1926986575126648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,32,1,balanced,0.2716053326924642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,32,1,balanced,0.35143999258677167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,32,1,balanced,0.49996264775594074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,32,1,balanced,0.6489546696345011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,32,1,balanced,0.7986079851786295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,32,1,balanced,1.2302026748657227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,32,1,balanced,2.3980053265889487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,4,power_law_1.01,4.816134262084961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,1,4,balanced,0.05810666580994924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,1,4,balanced,0.06029866635799408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,1,4,balanced,0.07573866844177246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,1,4,balanced,0.07425599793593089
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,1,4,balanced,0.0755573312441508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,1,4,balanced,0.07459733386834462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,1,4,balanced,0.07706666489442189
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,32,balanced,0.04450133442878723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,1,4,balanced,0.07505066692829132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,1,4,balanced,0.07445333401362102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,32,balanced,0.044064000248909
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,1,4,balanced,0.07569066683451335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,32,balanced,0.04362666606903076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,1,4,balanced,0.07630933324495952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,32,balanced,0.045663997530937195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,1,4,balanced,0.0784800002972285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,32,balanced,0.04599999884764353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,1,4,balanced,0.0790826678276062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,32,balanced,0.05072000126043955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,1,4,balanced,0.08264533181985219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,32,balanced,0.052005335688591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,1,4,balanced,0.08672533432642619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,32,balanced,0.05225066840648651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,1,4,balanced,0.08757866422335307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,32,balanced,0.053802669048309326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,1,4,balanced,0.09221866726875305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,32,balanced,0.05568000177542368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,32,balanced,0.05402133365472158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,1,4,balanced,0.10553066929181416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,32,balanced,0.05779733260472616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,1,4,balanced,0.11416533589363098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,32,balanced,0.05967999994754791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,1,4,balanced,0.1337279975414276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,32,balanced,0.06020799775918325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,1,4,balanced,0.15476266543070474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,32,balanced,0.06580266853173573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,32,balanced,0.06905599931875865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,1,4,balanced,0.19843200842539468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,32,balanced,0.07648533085982005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,1,4,balanced,0.23710399866104126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,32,balanced,0.08738133311271667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,32,balanced,0.09515733520189922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,1,4,balanced,0.32103466987609863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,32,balanced,0.11504000425338745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,1,4,balanced,0.4115360180536906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,32,balanced,0.12988266348838806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,32,balanced,0.16180266936620077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,1,4,balanced,0.5795626640319824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,32,balanced,0.19546133279800415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,32,balanced,0.2611413399378459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,1,4,balanced,0.7501546541849772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,32,balanced,0.3259999950726827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,1,4,balanced,0.9498346646626791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,32,balanced,0.4676640033721924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,32,balanced,0.5804320176442465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,1,4,balanced,1.4721652666727703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,32,balanced,0.7237386703491211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,32,balanced,1.0868319670359294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,1,4,balanced,2.7420212427775064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,32,balanced,2.1285707155863443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,2,power_law_1.01,0.17586560249328614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,2,power_law_1.01,0.26270720958709715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,2,power_law_1.01,0.22081279754638672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,2,power_law_1.01,0.3608896017074585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,2,power_law_1.01,0.43902077674865725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,2,power_law_1.01,0.8364607810974121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,2,power_law_1.01,0.9121855735778809
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,2,power_law_1.01,0.9125503540039063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,2,power_law_1.01,0.9571200370788574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,2,power_law_1.01,0.9666111946105957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,2,power_law_1.01,1.0123583793640136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,2,power_law_1.01,1.0010047912597657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,2,power_law_1.01,1.0474623680114745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,2,power_law_1.01,1.0668288230895997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,2,power_law_1.01,1.1033344268798828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,2,power_law_1.01,1.1141823768615722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,2,power_law_1.01,1.1781375885009766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,2,power_law_1.01,1.3009471893310547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,2,power_law_1.01,1.3675519943237304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,2,power_law_1.01,1.6264575958251952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,2,power_law_1.01,1.679417610168457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,2,power_law_1.01,2.0442495346069336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,2,power_law_1.01,2.1507007598876955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,2,power_law_1.01,2.5291519165039062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,2,power_law_1.01,2.7054399490356444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,2,power_law_1.01,3.7887039184570312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,2,power_law_1.01,5.228736114501953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,2,power_law_1.01,6.195411300659179
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,2,power_law_1.01,8.84158706665039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,2,power_law_1.01,16.067481994628906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,32,power_law_1.01,0.1591423988342285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,32,power_law_1.01,0.14501760005950928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,32,power_law_1.01,0.11282559633255004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,32,power_law_1.01,0.08577280044555664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,32,power_law_1.01,0.11119999885559081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,32,power_law_1.01,0.11679999828338623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,32,power_law_1.01,0.12932480573654176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,32,power_law_1.01,0.13293440341949464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,32,power_law_1.01,0.13281919956207275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,32,power_law_1.01,0.1254464030265808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,32,power_law_1.01,0.13738240003585817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,32,power_law_1.01,0.14383360147476196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,32,power_law_1.01,0.1514431953430176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,32,power_law_1.01,0.168614399433136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,32,power_law_1.01,0.18776960372924806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,32,power_law_1.01,0.18255360126495362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,32,power_law_1.01,0.16255359649658202
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,32,power_law_1.01,0.17950719594955444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,32,power_law_1.01,0.1807039976119995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,32,power_law_1.01,0.21237759590148925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,32,power_law_1.01,0.22242560386657714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,32,power_law_1.01,0.2631999969482422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,32,power_law_1.01,0.33201279640197756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,32,power_law_1.01,0.3996671915054321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,32,power_law_1.01,0.5069632053375244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,32,power_law_1.01,0.6630208015441894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,32,power_law_1.01,1.011731243133545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,32,power_law_1.01,1.0240639686584472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,32,power_law_1.01,1.7130495071411134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,32,power_law_1.01,3.001094436645508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,64,power_law_1.01,0.03524479866027832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,64,power_law_1.01,0.03296639919281006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,64,power_law_1.01,0.03219839930534363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,64,power_law_1.01,0.03331199884414673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,64,power_law_1.01,0.0332863986492157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,64,power_law_1.01,0.03358719944953918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,64,power_law_1.01,0.03351039886474609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,64,power_law_1.01,0.033990401029586795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,64,power_law_1.01,0.0343423992395401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,64,power_law_1.01,0.03521279990673065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,64,power_law_1.01,0.03523840010166168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,64,power_law_1.01,0.035417601466178894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,64,power_law_1.01,0.03601920008659363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,64,power_law_1.01,0.03696640133857727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,64,power_law_1.01,0.03979519903659821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,8,balanced,0.03796799977620443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,8,balanced,0.038058665891488395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,8,balanced,0.03932799895604452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,8,balanced,0.04561600089073181
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,8,balanced,0.06733866532643636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,64,power_law_1.01,0.04044159948825836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,8,balanced,0.06698666512966156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,8,balanced,0.06702400247255962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,8,balanced,0.06888000170389812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,8,balanced,0.06852266689141591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,8,balanced,0.0680159976085027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,8,balanced,0.06915199756622314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,8,balanced,0.06929066777229309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,8,balanced,0.07187733550866444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,64,power_law_1.01,0.0404992014169693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,8,balanced,0.07264000177383423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,8,balanced,0.07872533301512401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,8,balanced,0.07946133116881053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,8,balanced,0.08408533533414204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,8,balanced,0.08544533451398213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,8,balanced,0.10123733679453532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,8,balanced,0.10937066872914632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,64,power_law_1.01,0.042412799596786496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,8,balanced,0.1125440001487732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,8,balanced,0.1413653294245402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,8,balanced,0.16103999813397726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,8,balanced,0.2087200085322062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,8,balanced,0.2482186754544576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,64,power_law_1.01,0.04397439956665039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,8,balanced,0.33957334359486896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,8,balanced,0.4286506573359172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,8,balanced,0.5204693476359049
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,64,power_law_1.01,0.04986239969730377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,8,balanced,0.7949546972910563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,8,balanced,1.4942453702290852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,64,power_law_1.01,0.052697598934173584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,64,power_law_1.01,0.06182399988174438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,64,power_law_1.01,0.07214080095291138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,64,power_law_1.01,0.09552639722824097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,64,power_law_1.01,0.11012480258941651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,64,power_law_1.01,0.1667904019355774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,64,power_law_1.01,0.20083200931549072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,64,power_law_1.01,0.24524800777435302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,64,power_law_1.01,0.34133119583129884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,64,power_law_1.01,0.7234560012817383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,2,power_law_1.2,0.0718720018863678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,2,power_law_1.2,0.09743360280990601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,2,power_law_1.2,0.08079360127449035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,2,power_law_1.2,0.12766720056533815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,2,power_law_1.2,0.16577279567718506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,2,power_law_1.2,0.21779201030731202
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,2,1,balanced,0.07954666515191396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,2,1,balanced,0.1551146705945333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,2,1,balanced,0.2067520022392273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,2,1,balanced,0.2050293286641439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,2,1,balanced,0.2074293295542399
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,2,1,balanced,0.20867733160654703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,2,1,balanced,0.20965866247812906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,2,1,balanced,0.21363733212153116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,2,1,balanced,0.2161173423131307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,2,1,balanced,0.21524800856908163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,2,1,balanced,0.22002132733662924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,2,power_law_1.2,0.31958398818969724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,2,1,balanced,0.22207999229431152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,2,1,balanced,0.22563733657201132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,2,1,balanced,0.2347093423207601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,2,1,balanced,0.25511467456817627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,2,1,balanced,0.2550613284111023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,2,1,balanced,0.25994133949279785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,2,1,balanced,0.3229866623878479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,2,1,balanced,0.3380426565806071
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,2,1,balanced,0.41282665729522705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,2,1,balanced,0.4675626754760742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,2,1,balanced,0.633242646853129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,2,1,balanced,0.794111967086792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,2,power_law_1.2,0.32515840530395507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,2,1,balanced,1.1452693144480388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,2,1,balanced,1.5096640586853027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,2,1,balanced,2.215376059214274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,2,power_law_1.2,0.33783679008483886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,2,1,balanced,2.9188321431477866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,2,1,balanced,3.6314452489217124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,2,power_law_1.2,0.3512768030166626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,2,1,balanced,5.731253306070964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,2,power_law_1.2,0.37068159580230714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,2,power_law_1.2,0.3815167903900146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,2,1,balanced,11.343055725097656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,2,power_law_1.2,0.3901439905166626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,2,power_law_1.2,0.42748160362243653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,2,power_law_1.2,0.4474048137664795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,2,power_law_1.2,0.44495358467102053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,2,power_law_1.2,0.47886080741882325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,2,power_law_1.2,0.5303616046905517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,2,power_law_1.2,0.5614399909973145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,2,power_law_1.2,0.6457280158996582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,2,power_law_1.2,0.6996352195739746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,64,balanced,0.0521066685517629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,64,balanced,0.05397333204746246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,64,balanced,0.053930665055910744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,64,balanced,0.05398400127887726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,64,balanced,0.05589333176612854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,64,balanced,0.061103999614715576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,64,balanced,0.0601440022389094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,64,balanced,0.062037333846092224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,64,balanced,0.06307200094064076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,64,balanced,0.06425066788991292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,64,balanced,0.06429333488146464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,2,power_law_1.2,0.8576895713806152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,64,balanced,0.06423466900984447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,64,balanced,0.0661653329928716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,64,balanced,0.06649066507816315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,64,balanced,0.07007466753323872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,64,balanced,0.07423466444015503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,64,balanced,0.07584533095359802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,64,balanced,0.08258666594823201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,64,balanced,0.09074133634567261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,64,balanced,0.11129066348075867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,64,balanced,0.13225066661834717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,64,balanced,0.17157334089279175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,64,balanced,0.2029973268508911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,64,balanced,0.27218133211135864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,64,balanced,0.33686399459838867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,64,balanced,0.47226134936014813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,64,balanced,0.6042240063349406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,2,power_law_1.2,0.9619327545166015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,64,balanced,0.7426933447519938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,64,balanced,1.1387360095977783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,64,balanced,2.2436854044596353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,2,power_law_1.2,1.3225472450256348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,2,power_law_1.2,1.5016768455505372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,2,power_law_1.2,2.1110208511352537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,2,power_law_1.2,2.6747840881347655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,2,power_law_1.2,3.3082817077636717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,2,power_law_1.2,5.114265441894531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,4,power_law_1.01,0.0559935986995697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,4,power_law_1.01,0.06212480068206787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,4,power_law_1.01,0.05618559718132019
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,4,power_law_1.01,0.0660863995552063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,4,power_law_1.01,0.07232000231742859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,2,power_law_1.2,9.252652740478515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,4,power_law_1.01,0.08492799997329711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,4,power_law_1.01,0.1023103952407837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,4,power_law_1.01,0.1056447982788086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,4,power_law_1.01,0.10578559637069702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,4,power_law_1.01,0.10743039846420288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,4,power_law_1.01,0.11182719469070435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,4,power_law_1.01,0.11681280136108399
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,4,power_law_1.01,0.11731840372085571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,4,power_law_1.01,0.12079999446868897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,4,power_law_1.01,0.12980480194091798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,4,power_law_1.01,0.13805439472198486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,4,power_law_1.01,0.13752959966659545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,4,power_law_1.01,0.1515071988105774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,4,power_law_1.01,0.17210880517959595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,64,balanced,0.056176001826922096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,64,balanced,0.0506933331489563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,64,balanced,0.04984533290068308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,64,balanced,0.049973333875338234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,64,balanced,0.05006400247414907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,64,balanced,0.05212266743183136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,64,balanced,0.05376000205675761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,64,balanced,0.054287999868392944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,64,balanced,0.05409599840641022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,64,balanced,0.05365866422653198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,64,balanced,0.055957332253456116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,64,balanced,0.05599466462930044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,64,balanced,0.056202664971351624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,4,power_law_1.01,0.19720319509506226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,64,balanced,0.05761066575845083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,64,balanced,0.062352001667022705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,64,balanced,0.06238933404286703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,64,balanced,0.0665226678053538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,64,balanced,0.07073600093523662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,64,balanced,0.07461866736412048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,64,balanced,0.0874826709429423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,64,balanced,0.10134933392206828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,64,balanced,0.12295466661453247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,64,balanced,0.14641599853833517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,64,balanced,0.18372267484664917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,64,balanced,0.22781866788864136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,64,balanced,0.3080693284670512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,4,power_law_1.01,0.22823679447174072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,64,balanced,0.3757546742757161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,1,power_law_1.2,0.03733760118484497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,64,balanced,0.47115198771158856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,64,balanced,0.6864799658457438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,64,balanced,1.3465226491292317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,4,power_law_1.01,0.28967039585113524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,1,power_law_1.2,0.03654400110244751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,4,power_law_1.01,0.3537856101989746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,1,power_law_1.2,0.04177280068397522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,4,power_law_1.01,0.4719808101654053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,1,power_law_1.2,0.04647040069103241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,1,power_law_1.2,0.055795198678970336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,4,power_law_1.01,0.6398591995239258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,1,power_law_1.2,0.07579519748687744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,1,power_law_1.2,0.08240000009536744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,4,power_law_1.01,0.8528127670288086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,1,power_law_1.2,0.08517119884490967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,1,power_law_1.2,0.08753920197486878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,4,power_law_1.01,1.1304767608642579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,1,power_law_1.2,0.08984320163726807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,1,power_law_1.2,0.09219840168952942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,4,power_law_1.01,1.4346303939819336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,1,power_law_1.2,0.09735680222511292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,1,power_law_1.2,0.10154240131378174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,1,power_law_1.2,0.10695680379867553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,4,power_law_1.01,2.0593536376953123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,1,power_law_1.2,0.11797759532928467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,1,power_law_1.2,0.12180479764938354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,4,balanced,0.05855466425418854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,4,balanced,0.0611413319905599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,4,balanced,0.06109866499900818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,4,balanced,0.07628799974918365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,4,balanced,0.10233599940935771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,4,balanced,0.14320000012715658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,4,balanced,0.18182400862375894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,4,balanced,0.18401066462198892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,4,balanced,0.184063990910848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,4,balanced,0.18477332592010498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,1,power_law_1.2,0.1320512056350708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,4,balanced,0.18443200985590616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,4,balanced,0.18761066595713297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,4,balanced,0.1883093317349752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,4,balanced,0.19331733385721842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,4,balanced,0.19830934206644693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,4,balanced,0.1994826594988505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,4,balanced,0.20878400405248007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,4,balanced,0.2254400054613749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,4,balanced,0.23735467592875162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,4,balanced,0.26517866055170697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,1,power_law_1.2,0.15224319696426392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,4,balanced,0.2897439996401469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,4,balanced,0.34722665945688885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,4,balanced,0.40334399541219074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,4,balanced,0.5248266855875651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,4,balanced,0.6888159910837809
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,4,power_law_1.01,4.131161499023437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,1,power_law_1.2,0.16908799409866332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,4,balanced,0.918224016825358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,4,balanced,1.2152000268300374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,4,balanced,1.5099679629007976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,1,power_law_1.2,0.21056640148162842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,4,balanced,2.287333329518636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,1,power_law_1.2,0.24635519981384277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,4,balanced,4.378464063008626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,1,power_law_1.2,0.3341759920120239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,1,power_law_1.2,0.4226240158081055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,1,power_law_1.2,0.5933440208435059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,1,power_law_1.2,0.7593599796295166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,1,power_law_1.2,1.0887807846069335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,1,power_law_1.2,1.4226816177368165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,1,power_law_1.2,1.7522111892700196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,1,power_law_1.2,2.7444480895996093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,8,2,power_law_1.2,0.036524799466133115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,8,2,power_law_1.2,0.04733439981937408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,8,2,power_law_1.2,0.04429439902305603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,8,2,power_law_1.2,0.05304319858551025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,8,2,power_law_1.2,0.053862398862838744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,8,2,power_law_1.2,0.05749760270118713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,8,2,power_law_1.2,0.058380800485610965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,8,2,power_law_1.2,0.05818880200386047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,8,2,power_law_1.2,0.059487998485565186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,8,2,power_law_1.2,0.059385597705841064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,1,power_law_1.2,5.377651214599609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,8,2,power_law_1.2,0.06229760050773621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,8,2,power_law_1.2,0.06379520297050476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,8,2,power_law_1.2,0.06417279839515685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,8,2,power_law_1.2,0.07057920098304749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,8,2,power_law_1.2,0.07978240251541138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,8,2,power_law_1.2,0.08268160223960877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,8,2,power_law_1.2,0.09475200176239014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,8,2,power_law_1.2,0.1071552038192749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,8,2,power_law_1.2,0.12312320470809937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,1,power_law_1.2,0.07162240147590637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,8,2,power_law_1.2,0.15324800014495848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,8,2,power_law_1.2,0.19415040016174318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,8,2,power_law_1.2,0.2630336046218872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,8,2,power_law_1.2,0.3005568027496338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,1,power_law_1.2,0.11016319990158081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,8,2,power_law_1.2,0.42016000747680665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,8,2,power_law_1.2,0.5239039897918701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,8,2,power_law_1.2,0.8333696365356446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,1,power_law_1.2,0.11098239421844483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,8,2,power_law_1.2,1.0652223587036134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,8,2,power_law_1.2,1.2073792457580566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,1,power_law_1.2,0.17492480278015138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,8,2,power_law_1.2,2.0368576049804688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,1,power_law_1.2,0.24922239780426025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,16,power_law_1.01,0.03713920116424561
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,16,power_law_1.01,0.04207360148429871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,1,power_law_1.2,0.3626496076583862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,16,power_law_1.01,0.02746880054473877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,8,2,power_law_1.2,3.9950401306152346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,16,power_law_1.01,0.031353598833084105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,1,power_law_1.2,0.5561535835266114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,16,power_law_1.01,0.03113600015640259
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,16,power_law_1.01,0.03313280045986176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,1,power_law_1.2,0.6136320114135743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,16,power_law_1.01,0.03344640135765076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,16,power_law_1.01,0.03468799889087677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,1,power_law_1.2,0.6256127834320069
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,16,power_law_1.01,0.033958399295806886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,16,power_law_1.01,0.03468799889087677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,1,power_law_1.2,0.65731840133667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,16,power_law_1.01,0.03506560027599335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,16,power_law_1.01,0.03578880131244659
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,1,balanced,0.12947199741999307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,1,balanced,0.16057067116101584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,1,power_law_1.2,0.6862847805023193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,1,balanced,0.20266133546829224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,16,power_law_1.01,0.03727999925613403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,1,balanced,0.300437331199646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,1,balanced,0.4915200074513753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,1,balanced,0.6767199834187826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,16,power_law_1.01,0.03903999924659729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,1,balanced,0.6825653711954752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,1,balanced,0.687269369761149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,1,power_law_1.2,0.6849792003631592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,1,balanced,0.6933866341908773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,16,power_law_1.01,0.04419200122356415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,1,balanced,0.6980746587117513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,1,balanced,0.707594633102417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,1,balanced,0.7118026415506998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,16,power_law_1.01,0.047065600752830505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,1,balanced,0.7231573263804117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,1,balanced,0.7368053595225016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,1,power_law_1.2,0.7428864002227783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,1,balanced,0.7587893009185791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,16,power_law_1.01,0.04933120012283325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,1,balanced,0.7789386908213297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,1,balanced,0.8218932946523031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,16,power_law_1.01,0.054048001766204834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,1,balanced,0.9234613577524821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,1,balanced,1.0161333084106445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,1,power_law_1.2,0.7820608139038085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,16,power_law_1.01,0.05863040089607239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,1,balanced,1.2541706562042236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,1,balanced,1.3455573717753093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,16,power_law_1.01,0.06585599780082703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,1,balanced,2.0238986015319824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,1,power_law_1.2,0.7983168125152588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,16,power_law_1.01,0.06796159744262695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,1,balanced,2.2169386545817056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,16,power_law_1.01,0.08199040293693542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,1,balanced,3.338063875834147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,1,power_law_1.2,0.8252799987792969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,16,power_law_1.01,0.09852160215377807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,1,balanced,3.8989973068237305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,1,power_law_1.2,0.9135295867919921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,16,power_law_1.01,0.13103359937667847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,1,balanced,5.940074920654297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,16,power_law_1.01,0.15321600437164307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,1,power_law_1.2,0.9833344459533692
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,16,power_law_1.01,0.19544320106506347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,1,balanced,7.174832026163737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,1,power_law_1.2,1.0241920471191406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,16,power_law_1.01,0.21829760074615479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,2,16,power_law_1.01,0.27336320877075193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,2,1,balanced,9.12167994181315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,1,power_law_1.2,1.1553215980529785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,2,16,power_law_1.01,0.3727168083190918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,1,power_law_1.2,1.2314175605773925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,2,1,balanced,13.844351450602213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,1,power_law_1.2,1.4519231796264649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,2,16,power_law_1.01,0.6562047958374023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,1,power_law_1.2,1.7068992614746095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,1,power_law_1.2,2.150246429443359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,1,power_law_1.2,2.5627264022827148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,2,1,balanced,26.78575388590495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,1,power_law_1.2,3.462105560302734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,1,power_law_1.2,4.343910217285156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,1,power_law_1.2,5.246956634521484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,2,4,balanced,0.05801600217819214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,2,4,balanced,0.05787200232346853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,2,4,balanced,0.07859733204046886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,2,4,balanced,0.07978666822115581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,2,4,balanced,0.07876800000667572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,2,4,balanced,0.07881600161393483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,2,4,balanced,0.08027199904123943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,2,4,balanced,0.08127466837565105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,2,4,balanced,0.082997332016627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,2,4,balanced,0.08332266906897227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,2,4,balanced,0.08627200126647949
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,2,4,balanced,0.0846506655216217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,2,4,balanced,0.08692266543706258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,2,4,balanced,0.09062932928403218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,2,4,balanced,0.09866133332252502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,2,4,balanced,0.09814400474230449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,2,4,balanced,0.10266666611035664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,2,4,balanced,0.12777066230773926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,2,4,balanced,0.13407466808954874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,2,4,balanced,0.16740800937016806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,2,4,balanced,0.18726933002471924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,2,4,balanced,0.24260799090067545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,2,4,balanced,0.28380266825358075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,2,4,balanced,0.40265599886576336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,2,4,balanced,0.5077120065689087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,2,4,balanced,0.7341972986857096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,2,4,balanced,0.9474240144093832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,2,4,balanced,1.1653066476186116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,1,power_law_1.2,7.676735687255859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,2,4,balanced,1.8352905909220378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,2,4,balanced,3.58020814259847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,1,2,power_law_1.2,0.19023360013961793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,1,2,power_law_1.2,0.27206399440765383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,1,2,power_law_1.2,0.28086400032043457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,1,2,power_law_1.2,0.3768255949020386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,1,2,power_law_1.2,0.3897599935531616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,1,2,power_law_1.2,0.3973567962646484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,1,2,power_law_1.2,0.40020480155944826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,1,2,power_law_1.2,0.42247681617736815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,1,2,power_law_1.2,0.43291521072387695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,1,2,power_law_1.2,0.42230401039123533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,1,power_law_1.2,14.462425231933594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,1,2,power_law_1.2,0.4246272087097168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,1,2,power_law_1.2,0.44614400863647463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,1,2,power_law_1.2,0.4371840000152588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,1,2,power_law_1.2,0.45978879928588867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,1,2,power_law_1.2,0.509600019454956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,1,2,power_law_1.2,0.5157440185546875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,1,2,power_law_1.2,0.5748032093048095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,1,2,power_law_1.2,0.6985023975372314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,1,2,power_law_1.2,0.7559423923492432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,1,2,power_law_1.2,0.8786879539489746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,1,2,power_law_1.2,1.0974016189575195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,1,2,power_law_1.2,1.7590335845947265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,1,2,power_law_1.2,2.195155143737793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,1,2,power_law_1.2,3.0336767196655274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,4,balanced,0.054666668176651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,4,balanced,0.07665066421031952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,4,balanced,0.08636266986529033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,4,balanced,0.12628266215324402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,4,balanced,0.18317866325378418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,4,balanced,0.21486934026082358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,4,balanced,0.21526400248209634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,4,balanced,0.22004799048105875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,1,2,power_law_1.2,4.134361648559571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,4,balanced,0.21767999728520712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,4,balanced,0.22011200586954752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,4,balanced,0.2233333388964335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,4,balanced,0.22317334016164145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,4,balanced,0.22487467527389526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,4,balanced,0.22868800163269043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,4,balanced,0.23628799120585123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,4,balanced,0.2374239961306254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,4,balanced,0.2452053427696228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,4,balanced,0.2562026580174764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,4,balanced,0.26790400346120197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,4,balanced,0.32877333958943683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,4,balanced,0.32343467076619464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,4,balanced,0.43162135283152264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,4,balanced,0.44379734992980957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,4,balanced,0.6511946519215902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,4,balanced,0.7372533480326334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,1,2,power_law_1.2,6.319609451293945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,4,balanced,1.0897119839986165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,4,balanced,1.3161706924438477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,4,balanced,1.5289546648661296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,4,balanced,2.457354704538981
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,1,2,power_law_1.2,7.523583984375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,4,balanced,4.561333338419597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,8,8,power_law_1.2,0.03702400028705597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,8,8,power_law_1.2,0.03746559917926788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,8,8,power_law_1.2,0.03529599905014038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,8,8,power_law_1.2,0.035155200958251955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,8,8,power_law_1.2,0.035897600650787356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,8,8,power_law_1.2,0.036025598645210266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,8,8,power_law_1.2,0.03642880022525787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,8,8,power_law_1.2,0.03686400055885315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,8,8,power_law_1.2,0.037324801087379456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,8,8,power_law_1.2,0.03824639916419983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,1,2,power_law_1.2,10.396339416503906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,8,8,power_law_1.2,0.03797119855880737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,8,8,power_law_1.2,0.04087679982185364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,8,8,power_law_1.2,0.04007680118083954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,8,8,power_law_1.2,0.04433279931545257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,8,8,power_law_1.2,0.051475197076797485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,8,8,power_law_1.2,0.05479679703712463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,8,8,power_law_1.2,0.057132798433303836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,8,8,power_law_1.2,0.06832640171051026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,8,8,power_law_1.2,0.06985599994659424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,8,8,power_law_1.2,0.0866047978401184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,8,8,power_law_1.2,0.10083839893341065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,8,8,power_law_1.2,0.13831679821014403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,8,8,power_law_1.2,0.1641535997390747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,8,8,power_law_1.2,0.22279679775238037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,8,8,power_law_1.2,0.27472639083862305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,1,2,power_law_1.2,14.884608459472656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,8,8,power_law_1.2,0.3897984027862549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,8,8,power_law_1.2,0.5087039947509766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,8,8,power_law_1.2,0.6241983890533447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,8,8,power_law_1.2,0.9424960136413574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,8,8,power_law_1.2,1.8241727828979493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,16,power_law_1.2,0.05568640232086182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,1,2,power_law_1.2,30.172479248046876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,16,power_law_1.2,0.05978239774703979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,16,power_law_1.2,0.05018240213394165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,16,power_law_1.2,0.05162879824638367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,16,power_law_1.2,0.05345919728279114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,16,power_law_1.2,0.05466880202293396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,16,power_law_1.2,0.05811200141906738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,16,power_law_1.2,0.058246397972106935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,16,power_law_1.2,0.05980160236358643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,16,power_law_1.2,0.060153597593307497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,16,power_law_1.2,0.062054401636123656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,16,power_law_1.2,0.06366080045700073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,16,power_law_1.2,0.06522240042686463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,16,power_law_1.2,0.06539520025253295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,16,power_law_1.2,0.07148159742355346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,8,power_law_1.2,0.0758463978767395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,16,power_law_1.2,0.07303680181503296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,16,power_law_1.2,0.07893120050430298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,8,power_law_1.2,0.06620799899101257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,16,power_law_1.2,0.0910975992679596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,8,power_law_1.2,0.06614400148391723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,16,power_law_1.2,0.10272639989852905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,8,power_law_1.2,0.0797439992427826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,16,power_law_1.2,0.12255359888076782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,8,power_law_1.2,0.08156160116195679
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,16,power_law_1.2,0.14006400108337402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,8,power_law_1.2,0.10375679731369018
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,16,power_law_1.2,0.18396159410476684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,8,power_law_1.2,0.10464639663696289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,16,power_law_1.2,0.22366719245910643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,8,power_law_1.2,0.10969599485397338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,16,power_law_1.2,0.29619200229644777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,8,power_law_1.2,0.1128767967224121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,16,power_law_1.2,0.4147327899932861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,8,power_law_1.2,0.11510399580001832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,16,power_law_1.2,0.5462207794189453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,8,power_law_1.2,0.1134719967842102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,8,power_law_1.2,0.11832319498062134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,16,power_law_1.2,0.7406144142150879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,8,power_law_1.2,0.12036479711532592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,16,power_law_1.2,0.8449536323547363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,8,power_law_1.2,0.1273151993751526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,8,power_law_1.2,0.13864320516586304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,8,power_law_1.2,0.14234880208969117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,16,power_law_1.2,1.618067169189453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,8,power_law_1.2,0.1519551992416382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,8,power_law_1.2,0.18052480220794678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,8,power_law_1.2,0.18341120481491088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,16,power_law_1.2,3.3670272827148438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,8,power_law_1.2,0.23486080169677734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,8,power_law_1.2,0.24929919242858886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,8,power_law_1.2,0.31857919692993164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,8,power_law_1.2,0.39635839462280276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,8,power_law_1.2,0.5516799926757813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,8,power_law_1.2,0.716153621673584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,8,power_law_1.2,1.0405183792114259
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,8,power_law_1.2,1.11278076171875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,8,power_law_1.2,1.6468416213989259
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,8,power_law_1.2,2.308012771606445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,8,power_law_1.2,4.951436614990234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,8,power_law_1.2,0.056831997632980344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,8,power_law_1.2,0.06836479902267456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,8,power_law_1.2,0.0575872004032135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,8,power_law_1.2,0.0599232017993927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,8,power_law_1.2,0.06630399823188782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,8,power_law_1.2,0.07008640170097351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,8,power_law_1.2,0.08458240032196045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,8,power_law_1.2,0.08686720132827759
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,8,power_law_1.2,0.08977280259132385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,8,power_law_1.2,0.09283840060234069
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,8,power_law_1.2,0.09313279986381531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,8,power_law_1.2,0.09493759870529175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,8,power_law_1.2,0.09660800099372864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,8,power_law_1.2,0.0985152006149292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,8,power_law_1.2,0.10732799768447876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,8,power_law_1.2,0.11370879411697388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,8,power_law_1.2,0.12143360376358033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,8,power_law_1.2,0.1412992000579834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,8,power_law_1.2,0.15930240154266356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,8,power_law_1.2,0.18176000118255614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,8,power_law_1.2,0.20804479122161865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,8,power_law_1.2,0.26185600757598876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,8,power_law_1.2,0.28759040832519533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,8,power_law_1.2,0.4116032123565674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,8,power_law_1.2,0.5092544078826904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,8,power_law_1.2,0.739251184463501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,8,power_law_1.2,1.0105216026306152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,8,power_law_1.2,1.1507776260375977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,8,power_law_1.2,2.1237567901611327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,8,power_law_1.2,4.1128894805908205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,16,4,balanced,0.037989333271980286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,16,4,balanced,0.03192000091075897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,16,4,balanced,0.0324799989660581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,16,4,balanced,0.03329599897066752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,16,4,balanced,0.032111999889214836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,16,4,balanced,0.03411199897527695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,16,4,balanced,0.03384000062942505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,16,4,balanced,0.034143999218940735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,16,4,balanced,0.03393599887688955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,16,4,balanced,0.03349866718053818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,16,4,balanced,0.03457599878311157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,16,4,balanced,0.034458667039871216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,16,4,balanced,0.03572266548871994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,16,4,balanced,0.03570666660865148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,16,4,balanced,0.03999999910593033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,16,4,balanced,0.04394666850566864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,16,4,balanced,0.04181866844495138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,16,4,balanced,0.047322665651639305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,16,4,balanced,0.047925333182017006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,16,4,balanced,0.05958933134873708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,16,4,balanced,0.06066666543483734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,16,4,balanced,0.08116266628106435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,16,4,balanced,0.09156800309816997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,16,4,balanced,0.12382933497428894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,16,4,balanced,0.13965333501497904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,16,4,balanced,0.1832639972368876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,16,4,balanced,0.21890133619308472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,16,4,balanced,0.2690346638361613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,16,4,balanced,0.36955201625823975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,16,4,balanced,0.6815360387166342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,32,4,balanced,0.05463466544946035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,32,4,balanced,0.05913066864013672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,32,4,balanced,0.05898666878541311
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,32,4,balanced,0.06697600086530049
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,32,4,balanced,0.07709333300590515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,32,4,balanced,0.0977280040582021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,32,4,balanced,0.09756267070770264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,32,4,balanced,0.10148266951243083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,32,4,balanced,0.10061333576838176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,32,4,balanced,0.1013813316822052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,32,4,balanced,0.10384000341097514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,32,4,balanced,0.10531733433405559
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,32,4,balanced,0.10583999752998352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,32,4,balanced,0.10871466994285583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,32,4,balanced,0.11516799529393514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,32,4,balanced,0.11786666512489319
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,32,4,balanced,0.11990933616956075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,32,4,balanced,0.13354133566220602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,32,4,balanced,0.14773866534233093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,32,4,balanced,0.17213332653045654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,32,4,balanced,0.19889599084854126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,32,4,balanced,0.24458666642506918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,32,4,balanced,0.2911626696586609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,32,4,balanced,0.4210133155186971
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,32,4,balanced,0.5290613174438477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,32,4,balanced,0.7679200172424316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,32,4,balanced,0.997920036315918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,32,4,balanced,1.2411786715189617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,32,4,balanced,1.902469317118327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,32,4,balanced,3.71012274424235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,32,balanced,0.046480000019073486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,32,balanced,0.04725866516431173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,32,balanced,0.04630400240421295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,32,balanced,0.04896000027656555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,32,balanced,0.05172266562779745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,32,balanced,0.06454933186372121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,32,balanced,0.06527466575304668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,32,balanced,0.06569066643714905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,32,balanced,0.06744533280531566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,32,balanced,0.06896533568700154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,32,balanced,0.06891733407974243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,32,balanced,0.07006399830182393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,32,balanced,0.07272000114123027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,32,balanced,0.07406400144100189
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,32,balanced,0.07994133234024048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,32,balanced,0.08169599870840709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,32,balanced,0.08474133412043254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,32,balanced,0.09363200267155965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,32,balanced,0.09521067142486572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,32,balanced,0.11498133341471355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,32,balanced,0.12402133146921794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,32,balanced,0.15771200259526572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,32,balanced,0.1837973395983378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,32,balanced,0.25114667415618896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,32,balanced,0.2897439996401469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,32,balanced,0.42418134212493896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,32,balanced,0.50764266649882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,32,balanced,0.6054986715316772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,32,balanced,0.9251680374145508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,32,balanced,1.7918400764465332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,4,64,power_law_1.2,0.03578880131244659
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,4,64,power_law_1.2,0.03246079981327057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,4,64,power_law_1.2,0.03313280045986176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,4,64,power_law_1.2,0.03344640135765076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,4,64,power_law_1.2,0.03333120048046112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,4,64,power_law_1.2,0.03351680040359497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,4,64,power_law_1.2,0.03368319869041443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,4,64,power_law_1.2,0.03452160060405731
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,4,64,power_law_1.2,0.034297600388526917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,4,64,power_law_1.2,0.03484799861907959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,4,power_law_1.01,0.048563200235366824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,4,64,power_law_1.2,0.035692799091339114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,4,64,power_law_1.2,0.03607679903507233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,4,64,power_law_1.2,0.03622399866580963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,4,power_law_1.01,0.053907197713851926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,4,64,power_law_1.2,0.036582401394844054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,4,64,power_law_1.2,0.04004479944705963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,4,power_law_1.01,0.051795202493667605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,4,64,power_law_1.2,0.03954559862613678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,4,64,power_law_1.2,0.04190079867839813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,4,64,power_law_1.2,0.042387199401855466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,4,power_law_1.01,0.05851519703865051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,4,64,power_law_1.2,0.045465600490570066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,4,power_law_1.01,0.06421120166778564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,4,64,power_law_1.2,0.051622402667999265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,4,64,power_law_1.2,0.05711359977722168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,4,power_law_1.01,0.08281599879264831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,4,64,power_law_1.2,0.07220479846000671
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,4,64,power_law_1.2,0.07834240198135375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,4,power_law_1.01,0.08390399813652039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,4,64,power_law_1.2,0.1134719967842102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,4,power_law_1.01,0.08885120153427124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,4,64,power_law_1.2,0.15089279413223267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,4,power_law_1.01,0.0870527982711792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,4,64,power_law_1.2,0.1995967984199524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,4,power_law_1.01,0.09159680008888245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,4,64,power_law_1.2,0.2705600023269653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,4,power_law_1.01,0.09359999895095825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,4,64,power_law_1.2,0.3076479911804199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,4,power_law_1.01,0.09804800152778625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,4,power_law_1.01,0.10096640586853027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,4,64,power_law_1.2,0.59552001953125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,4,power_law_1.01,0.1009984016418457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,4,power_law_1.01,0.1115839958190918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,4,64,power_law_1.2,1.2653056144714356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,4,power_law_1.01,0.1125823974609375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,4,power_law_1.01,0.12057600021362305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,4,power_law_1.01,0.1402943968772888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,4,power_law_1.01,0.15957759618759154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,4,power_law_1.01,0.18916480541229247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,4,power_law_1.01,0.23264639377593993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,4,power_law_1.01,0.30583040714263915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,4,power_law_1.01,0.3911360025405884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,4,power_law_1.01,0.5070335865020752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,4,power_law_1.01,0.640934419631958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,4,power_law_1.01,0.9206015586853027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,64,balanced,0.06483733157316844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,64,balanced,0.05008000135421753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,64,balanced,0.050101334849993386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,64,balanced,0.0506933331489563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,64,balanced,0.05273066461086273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,64,balanced,0.059018666545550026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,64,balanced,0.058975999553998314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,64,balanced,0.06042666733264923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,64,balanced,0.0603413333495458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,64,balanced,0.06025599936644236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,64,balanced,0.06262933214505513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,64,balanced,0.06539733211199443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,64,balanced,0.06474133332570393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,4,power_law_1.01,1.1729087829589844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,64,balanced,0.06644799808661143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,64,balanced,0.07501866420110066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,64,balanced,0.07634133100509644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,64,balanced,0.08124266564846039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,64,balanced,0.08912533521652222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,64,balanced,0.09119466940561931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,64,balanced,0.1053013304869334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,64,balanced,0.12378133336702983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,64,balanced,0.1423733333746592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,64,balanced,0.1607093314329783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,64,balanced,0.20166399081548056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,64,balanced,0.23600000143051147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,64,balanced,0.32202666997909546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,64,balanced,0.40305066108703613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,64,balanced,0.4894719918568929
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,64,balanced,0.7306613127390543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,4,power_law_1.01,1.4260671615600586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,64,balanced,1.411253293355306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,4,power_law_1.01,2.3542591094970704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,4,power_law_1.01,4.4050239562988285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,2,balanced,0.03373866776625315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,2,balanced,0.032501332461833954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,2,balanced,0.036117332677046456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,2,balanced,0.03775999943415324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,2,balanced,0.04529599845409393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,2,balanced,0.04558399816354116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,2,balanced,0.045925334095954895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,2,balanced,0.04586666822433472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,2,balanced,0.045909335215886436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,2,balanced,0.047770669062932335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,2,balanced,0.048021331429481506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,2,balanced,0.05177066723505656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,2,balanced,0.052111998200416565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,2,balanced,0.05277866621812185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,2,balanced,0.062218666076660156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,2,balanced,0.06446933249632518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,2,balanced,0.06565333406130473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,2,balanced,0.07866666714350383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,2,balanced,0.09199466307957967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,2,balanced,0.11502400040626526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,2,balanced,0.12703466415405273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,2,balanced,0.17163199186325073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,2,balanced,0.2037280003229777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,2,balanced,0.2926933368047078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,2,balanced,0.38419198989868164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,2,balanced,0.5753066539764404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,2,balanced,0.7634399731953939
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,2,balanced,0.950762669245402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,2,balanced,1.5065174102783203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,2,balanced,2.881429354349772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,64,power_law_1.01,0.10606720447540283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,64,power_law_1.01,0.10289280414581299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,64,power_law_1.01,0.07355520129203796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,2,1,power_law_1.01,0.11860480308532714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,2,1,power_law_1.01,0.18355200290679932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,2,1,power_law_1.01,0.21604480743408203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,2,1,power_law_1.01,0.36632959842681884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,2,1,power_law_1.01,0.3771008014678955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,2,1,power_law_1.01,0.37482240200042727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,2,1,power_law_1.01,0.3835839986801147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,64,power_law_1.01,0.08842880129814149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,2,1,power_law_1.01,0.38373119831085206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,2,1,power_law_1.01,0.39394559860229494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,2,1,power_law_1.01,0.3891072034835815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,2,1,power_law_1.01,0.3936959981918335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,2,1,power_law_1.01,0.4062528133392334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,2,1,power_law_1.01,0.4076223850250244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,2,1,power_law_1.01,0.418009614944458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,2,1,power_law_1.01,0.44408321380615234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,2,1,power_law_1.01,0.4785600185394287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,2,1,power_law_1.01,0.5010176181793213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,2,1,power_law_1.01,0.5643712043762207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,64,power_law_1.01,0.08640639781951905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,2,1,power_law_1.01,0.642195177078247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,2,1,power_law_1.01,0.7683199882507324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,2,1,power_law_1.01,0.8873279571533204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,2,1,power_law_1.01,1.206726360321045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,2,1,power_law_1.01,1.4636223793029786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,2,1,power_law_1.01,2.075366401672363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,64,power_law_1.01,0.09737600088119507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,2,1,power_law_1.01,2.6727167129516602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,2,1,power_law_1.01,3.920512008666992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,64,power_law_1.01,0.09694079756736755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,2,1,power_law_1.01,5.113516616821289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,2,1,power_law_1.01,6.388415908813476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,64,power_law_1.01,0.1022528052330017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,64,power_law_1.01,0.1038591980934143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,2,1,power_law_1.01,9.948274993896485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,64,power_law_1.01,0.10736639499664306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,64,power_law_1.01,0.10316159725189208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,2,1,power_law_1.01,19.40015411376953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,64,power_law_1.01,0.10650880336761474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,64,power_law_1.01,0.10793600082397461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,64,power_law_1.01,0.11057920455932617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,64,power_law_1.01,0.11559040546417236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,64,power_law_1.01,0.12076799869537354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,64,power_law_1.01,0.14224640130996705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,2,1,power_law_1.01,0.08006399869918823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,2,1,power_law_1.01,0.11384960412979125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,2,1,power_law_1.01,0.15281280279159545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,2,1,power_law_1.01,0.23457279205322265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,2,1,power_law_1.01,0.22992000579833985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,2,1,power_law_1.01,0.2350719928741455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,2,1,power_law_1.01,0.23539841175079346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,2,1,power_law_1.01,0.24030721187591553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,2,1,power_law_1.01,0.24540159702301026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,2,1,power_law_1.01,0.2503488063812256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,64,power_law_1.01,0.1597632050514221
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,2,1,power_law_1.01,0.26245760917663574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,2,1,power_law_1.01,0.2619839906692505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,2,1,power_law_1.01,0.2646912097930908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,2,1,power_law_1.01,0.26128640174865725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,2,1,power_law_1.01,0.2925695896148682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,2,1,power_law_1.01,0.2914752006530762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,2,1,power_law_1.01,0.31150720119476316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,2,1,power_law_1.01,0.3591423988342285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,2,1,power_law_1.01,0.3968319892883301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,64,power_law_1.01,0.14979200363159179
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,2,1,power_law_1.01,0.47032961845397947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,2,1,power_law_1.01,0.5690112113952637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,2,1,power_law_1.01,0.7345024108886719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,2,1,power_law_1.01,0.9101311683654785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,2,1,power_law_1.01,1.2965567588806153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,2,1,power_law_1.01,1.6283903121948242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,64,power_law_1.01,0.19167360067367553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,2,1,power_law_1.01,2.3807743072509764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,2,1,power_law_1.01,3.0846847534179687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,64,power_law_1.01,0.21155838966369628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,2,1,power_law_1.01,3.8942337036132812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,2,1,power_law_1.01,5.969696044921875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,64,power_law_1.01,0.2916928052902222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,2,1,power_law_1.01,11.654354858398438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,64,power_law_1.01,0.27626240253448486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,64,power_law_1.01,0.3837824106216431
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,64,power_law_1.01,0.42812161445617675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,64,power_law_1.01,0.6376704216003418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,2,balanced,0.056128000219662987
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,2,balanced,0.08256533245245616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,2,balanced,0.09719467163085938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,2,balanced,0.1479680041472117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,2,balanced,0.2453706661860148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,2,balanced,0.4416373173395793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,2,balanced,0.4452213446299235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,64,power_law_1.01,0.7597760200500489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,2,balanced,0.44310398896535236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,2,balanced,0.4471413294474284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,2,balanced,0.4485386610031128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,2,balanced,0.44947731494903564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,2,balanced,0.45307199160257977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,2,balanced,0.45815467834472656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,2,balanced,0.459386666615804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,2,balanced,0.4750613371531169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,2,balanced,0.4771466652552287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,2,balanced,0.4869973262151082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,2,balanced,0.5082613229751587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,2,balanced,0.5278026660283407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,2,balanced,0.5715680122375488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,2,balanced,0.6147626638412476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,2,balanced,0.6906773249308268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,2,balanced,0.7762400309244791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,2,balanced,0.9729920228322347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,2,balanced,1.11517333984375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,2,balanced,1.564458688100179
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,2,balanced,1.9204586346944172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,2,balanced,2.4228426615397134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,1,64,power_law_1.01,0.9430912017822266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,2,balanced,3.5044533411661782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,2,balanced,6.747568130493164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,1,64,power_law_1.01,1.3791935920715332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,1,64,power_law_1.01,2.6315776824951174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,4,1,balanced,0.03808533400297165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,4,1,balanced,0.06014933188756307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,4,1,balanced,0.08584533135096233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,4,1,balanced,0.08478933572769165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,4,1,balanced,0.08442133665084839
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,4,1,balanced,0.08538666367530823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,4,1,balanced,0.08727999528249104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,4,1,balanced,0.08680533369382222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,4,1,balanced,0.0885599950949351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,4,1,balanced,0.08737066388130188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,4,1,balanced,0.08959999680519104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,4,1,balanced,0.09193600217501323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,4,1,balanced,0.09122133255004883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,4,1,balanced,0.09673600395520528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,4,1,balanced,0.10557867089907329
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,4,1,balanced,0.10873599847157796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,4,1,balanced,0.11077866951624553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,4,1,balanced,0.13235732913017273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,4,1,balanced,0.14058666427930197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,4,1,balanced,0.17493333419164023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,4,1,balanced,0.19723733266194662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,4,1,balanced,0.2640373309453328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,4,1,balanced,0.32788266738255817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,4,1,balanced,0.4663413365681966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,4,1,balanced,0.5949600140253702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,4,1,balanced,0.8632319768269857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,4,1,balanced,1.113813320795695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,4,1,balanced,1.3792479832967122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,4,1,balanced,2.147637367248535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,4,1,balanced,4.214069366455078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,32,8,power_law_1.2,0.048640000820159915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,32,8,power_law_1.2,0.050380802154541014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,1,balanced,0.039962666730086006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,1,balanced,0.0399893323580424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,1,balanced,0.05835199852784475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,1,balanced,0.08156799773375194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,1,balanced,0.12597333391507468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,1,balanced,0.1276533305644989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,1,balanced,0.13013866543769836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,1,balanced,0.12924800316492716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,1,balanced,0.1320853332678477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,1,balanced,0.13594667116800943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,1,balanced,0.13640000422795615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,1,balanced,0.1420693298180898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,1,balanced,0.1437013347943624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,1,balanced,0.14702933033307394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,1,balanced,0.15441600481669107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,1,balanced,0.16014400124549866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,1,balanced,0.17147733767827353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,1,balanced,0.1886826753616333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,32,8,power_law_1.2,0.045817598700523376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,1,balanced,0.2108373244603475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,1,balanced,0.26444266239802044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,1,balanced,0.2978293299674988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,1,balanced,0.40427732467651367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,1,balanced,0.47807466983795166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,1,balanced,0.6735520362854004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,1,balanced,0.8491413593292236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,1,balanced,1.23963729540507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,1,balanced,1.6458133061726887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,32,8,power_law_1.2,0.046854400634765626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,1,balanced,2.051898638407389
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,1,balanced,3.16429869333903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,32,8,power_law_1.2,0.05053439736366272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,1,balanced,6.168250401814778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,32,8,power_law_1.2,0.05304960012435913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,32,8,power_law_1.2,0.05623040199279785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,32,8,power_law_1.2,0.058182400465011594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,32,8,power_law_1.2,0.05980799794197082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,32,8,power_law_1.2,0.058937597274780276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,32,8,power_law_1.2,0.06376960277557372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,32,8,power_law_1.2,0.06604160070419311
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,32,8,power_law_1.2,0.06798719763755798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,32,8,power_law_1.2,0.073471999168396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,32,8,power_law_1.2,0.07895039916038513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,32,power_law_1.01,0.06340479850769043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,32,power_law_1.01,0.060524797439575194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,32,8,power_law_1.2,0.08487039804458618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,32,power_law_1.01,0.05200639963150024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,32,power_law_1.01,0.054176002740859985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,32,8,power_law_1.2,0.08646399974822998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,32,power_law_1.01,0.05538560152053833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,32,power_law_1.01,0.05614079833030701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,32,8,power_law_1.2,0.09715200066566468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,32,power_law_1.01,0.057625597715377806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,32,power_law_1.01,0.05931519865989685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,32,8,power_law_1.2,0.11633280515670777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,32,power_law_1.01,0.059033602476119995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,32,power_law_1.01,0.058348798751831056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,32,8,power_law_1.2,0.1539199948310852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,32,power_law_1.01,0.06004480123519897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,32,power_law_1.01,0.06082559823989868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,32,power_law_1.01,0.06117759943008423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,32,8,power_law_1.2,0.1937343955039978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,32,power_law_1.01,0.06197119951248169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,32,power_law_1.01,0.06659200191497802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,32,8,power_law_1.2,0.242956805229187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,32,power_law_1.01,0.06714879870414733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,32,power_law_1.01,0.07107200026512146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,32,8,power_law_1.2,0.3162879943847656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,32,power_law_1.01,0.08278399705886841
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,32,power_law_1.01,0.09100800156593322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,32,8,power_law_1.2,0.40965118408203127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,32,power_law_1.01,0.11345920562744141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,32,power_law_1.01,0.11953279972076417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,32,8,power_law_1.2,0.525055980682373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,32,power_law_1.01,0.1615615963935852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,32,power_law_1.01,0.18689279556274413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,32,8,power_law_1.2,0.7898431777954101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,32,power_law_1.01,0.24815359115600585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,32,power_law_1.01,0.29441280364990235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,32,8,power_law_1.2,1.0308671951293946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,32,power_law_1.01,0.4393919944763184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,32,power_law_1.01,0.5636735916137695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,32,8,power_law_1.2,1.2659135818481446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,32,power_law_1.01,0.6842175960540772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,32,8,power_law_1.2,1.9875200271606446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,32,power_law_1.01,1.144435214996338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,32,power_law_1.01,2.3991039276123045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,32,8,power_law_1.2,3.873183822631836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,16,balanced,0.05080533524354299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,16,balanced,0.05057600140571594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,16,balanced,0.050586665670077004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,16,balanced,0.05226133267084757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,16,balanced,0.056176001826922096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,16,balanced,0.060191998879114784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,16,balanced,0.06249066690603892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,16,balanced,0.06278933087984721
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,16,balanced,0.06287466486295064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,16,balanced,0.0637600024541219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,16,balanced,0.06418133278687795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,16,balanced,0.06623466809590657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,16,balanced,0.06468800206979115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,16,balanced,0.06653333206971486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,16,balanced,0.07328000168005626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,16,balanced,0.07278400162855785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,16,balanced,0.07460266848405202
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,16,balanced,0.08338666955629985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,16,balanced,0.0892693301041921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,16,balanced,0.10728533069292705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,16,balanced,0.12942933042844137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,16,balanced,0.15268266201019287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,16,balanced,0.17715734243392944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,16,balanced,0.23549866676330566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,16,balanced,0.28352532784144086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,16,balanced,0.3927626609802246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,16,balanced,0.5024746656417847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,16,balanced,0.6201493342717489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,16,balanced,0.9380479653676351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,16,balanced,1.827221393585205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,8,power_law_1.2,0.12075519561767578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,8,power_law_1.2,0.11875200271606445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,8,power_law_1.2,0.0813696026802063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,8,power_law_1.2,0.10162559747695923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,8,power_law_1.2,0.13233920335769653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,1,balanced,0.0870293378829956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,1,balanced,0.15552533666292825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,1,balanced,0.1909760038057963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,1,balanced,0.3269546627998352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,1,balanced,0.5948746601740519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,1,balanced,0.8499200344085693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,1,balanced,0.8549173672993978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,1,balanced,0.8613386948903402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,1,balanced,0.8655947049458822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,1,balanced,0.8707040150960287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,1,balanced,0.8760000069936117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,1,balanced,0.8840426603953043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,8,power_law_1.2,0.19685759544372558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,1,balanced,0.8893919785817465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,1,balanced,0.9065013726552328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,1,balanced,0.9227946599324545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,1,balanced,0.9319413503011068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,1,balanced,0.9591093063354492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,1,balanced,1.160266637802124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,1,balanced,1.0705013275146484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,1,balanced,1.382256031036377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,1,balanced,1.3286026318868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,8,power_law_1.2,0.20945920944213867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,1,balanced,1.6978240013122559
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,1,balanced,1.790298620859782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,1,balanced,2.542149384816488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,1,balanced,2.9039414723714194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,8,power_law_1.2,0.21708159446716307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,1,balanced,4.583930651346843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,1,balanced,5.224437395731608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,8,power_law_1.2,0.22199039459228515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,4,1,balanced,6.656869252522786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,4,1,balanced,10.194677352905273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,8,power_law_1.2,0.22576000690460205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,8,power_law_1.2,0.2442624092102051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,4,1,balanced,19.41300328572591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,8,power_law_1.2,0.2408128023147583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,8,power_law_1.2,0.24539520740509033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,8,power_law_1.2,0.25747840404510497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,8,power_law_1.2,0.27216639518737795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,8,power_law_1.2,0.27469439506530763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,8,power_law_1.2,0.30053761005401614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,8,balanced,0.034341332813103996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,8,balanced,0.03832533210515976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,8,balanced,0.05551999807357788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,8,balanced,0.07477866609891255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,8,balanced,0.12609066565831503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,8,balanced,0.1209440032641093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,8,balanced,0.12102933724721272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,8,balanced,0.12360533078511556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,8,balanced,0.12173866232236226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,8,balanced,0.1207413375377655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,8,balanced,0.12199466427167256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,8,balanced,0.12455466389656067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,8,balanced,0.12435199817021687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,8,balanced,0.1244533360004425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,8,balanced,0.1299626628557841
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,8,balanced,0.12966932853062949
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,8,balanced,0.13144532839457193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,8,balanced,0.13338133692741394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,8,power_law_1.2,0.3454400062561035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,8,balanced,0.14270933469136557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,8,balanced,0.14888532956441244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,8,balanced,0.15876799821853638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,8,balanced,0.18961066007614136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,8,balanced,0.19561066230138144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,8,balanced,0.2600906689961751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,8,balanced,0.29046400388081867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,8,balanced,0.37251198291778564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,8,balanced,0.47095998128255206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,8,balanced,0.5599733193715414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,8,balanced,0.8469333648681641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,8,balanced,1.6051519711812336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,8,power_law_1.2,0.3432960033416748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,8,power_law_1.2,0.4299839973449707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,8,power_law_1.2,0.48888959884643557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,8,power_law_1.2,0.5804287910461425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,1,8,power_law_1.01,0.028838399052619933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,8,power_law_1.2,0.6017280101776123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,1,8,power_law_1.01,0.026873600482940675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,1,8,power_law_1.01,0.030880001187324525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,1,8,power_law_1.01,0.03306879997253418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,1,8,power_law_1.01,0.037945601344108584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,1,8,power_law_1.01,0.046348801255226134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,8,power_law_1.2,0.7664000034332276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,64,power_law_1.2,0.07697280049324036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,1,8,power_law_1.01,0.047225600481033324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,1,8,power_law_1.01,0.04747520089149475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,1,8,power_law_1.01,0.049414399266242984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,64,power_law_1.2,0.04717440009117126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,1,8,power_law_1.01,0.05057280063629151
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,1,8,power_law_1.01,0.05292159914970398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,64,power_law_1.2,0.04521600008010864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,1,8,power_law_1.01,0.0546176016330719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,8,power_law_1.2,1.0074239730834962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,64,power_law_1.2,0.04806399941444397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,1,8,power_law_1.01,0.0523967981338501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,1,8,power_law_1.01,0.05634559988975525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,64,power_law_1.2,0.04737919867038727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,1,8,power_law_1.01,0.060236799716949466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,1,8,power_law_1.01,0.06377599835395813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,64,power_law_1.2,0.04707199931144714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,1,8,power_law_1.01,0.06432639956474304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,1,8,power_law_1.01,0.0725055992603302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,64,power_law_1.2,0.04790399968624115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,8,power_law_1.2,1.3362943649291992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,8,power_law_1.01,0.07344639897346497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,64,power_law_1.2,0.05305600166320801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,8,power_law_1.01,0.08310400247573853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,8,power_law_1.01,0.0896448016166687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,64,power_law_1.2,0.05377280116081238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,8,power_law_1.01,0.10717439651489258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,64,power_law_1.2,0.05389440059661865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,8,power_law_1.01,0.13401600122451782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,64,power_law_1.2,0.05311999917030334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,8,power_law_1.01,0.16880639791488647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,8,power_law_1.2,1.6354751586914062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,8,power_law_1.01,0.21006081104278565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,64,power_law_1.2,0.0554751992225647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,8,power_law_1.01,0.27418880462646483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,64,power_law_1.2,0.0578495979309082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,8,power_law_1.01,0.3519551992416382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,64,power_law_1.2,0.05998079776763916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,64,power_law_1.2,0.06803200244903565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,8,power_law_1.01,0.4344768047332764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,64,power_law_1.2,0.0670527994632721
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,1,8,power_law_1.2,1.8657983779907226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,8,power_law_1.01,0.7025728225708008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,64,power_law_1.2,0.06996480226516724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,64,power_law_1.2,0.08408960103988647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,64,power_law_1.2,0.095769602060318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,8,power_law_1.01,1.0529791831970214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,64,power_law_1.2,0.10673919916152955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,64,power_law_1.2,0.12315520048141479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,1,8,power_law_1.2,2.9768447875976562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,64,power_law_1.2,0.16149760484695436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,64,power_law_1.2,0.18287359476089476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,64,power_law_1.2,0.25032958984375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,64,power_law_1.2,0.36012160778045654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,64,power_law_1.2,0.544377613067627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,64,power_law_1.2,0.6009984016418457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,1,8,power_law_1.2,6.266764831542969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,64,power_law_1.2,0.6555903911590576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,64,power_law_1.2,1.0913023948669434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,1,power_law_1.2,0.07322239875793457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,1,power_law_1.2,0.09896960258483886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,64,power_law_1.2,2.4141056060791017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,1,power_law_1.2,0.1236415982246399
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,1,power_law_1.2,0.16715519428253173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,1,power_law_1.2,0.2960576057434082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,1,power_law_1.2,0.4343552112579346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,1,power_law_1.2,0.5196800231933594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,1,8,balanced,0.02603733291228612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,1,8,balanced,0.026159999271233875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,1,8,balanced,0.027914665639400482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,1,8,balanced,0.034341332813103996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,1,8,balanced,0.05193600058555603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,1,8,balanced,0.05275199810663859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,1,8,balanced,0.05285866558551788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,1,8,balanced,0.05266666909058889
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,1,8,balanced,0.05384000142415365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,1,8,balanced,0.05454400181770325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,1,8,balanced,0.05412266651789347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,1,8,balanced,0.05435200035572052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,1,power_law_1.2,0.7120575904846191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,1,8,balanced,0.05561066667238871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,1,8,balanced,0.05819199979305267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,1,8,balanced,0.06278933087984721
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,1,8,balanced,0.06126399834950765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,1,8,balanced,0.06264000137646993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,1,8,balanced,0.06810133159160614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,8,balanced,0.06837333242098491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,8,balanced,0.07505600154399872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,8,balanced,0.07642666498819987
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,8,balanced,0.09173867106437683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,8,balanced,0.10249599814414978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,8,balanced,0.12687466541926065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,8,balanced,0.15632533033688864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,8,balanced,0.2022613286972046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,8,balanced,0.23574932416280112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,1,power_law_1.2,0.7889920234680176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,8,balanced,0.28055999676386517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,8,balanced,0.4092106819152832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,8,balanced,0.7505653699239095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,1,power_law_1.2,0.8184191703796386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,1,power_law_1.2,0.843238353729248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,64,power_law_1.01,0.032287999987602234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,64,power_law_1.01,0.03470720052719116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,1,power_law_1.2,0.8984127998352051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,64,power_law_1.01,0.029900801181793214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,64,power_law_1.01,0.030924800038337707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,1,power_law_1.2,0.9311871528625488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,64,power_law_1.01,0.03020800054073334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,64,power_law_1.01,0.030687999725341798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,64,power_law_1.01,0.030636799335479737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,1,power_law_1.2,0.9851648330688476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,64,power_law_1.01,0.030694401264190672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,64,power_law_1.01,0.030803200602531434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,64,power_law_1.01,0.03111039996147156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,1,power_law_1.2,1.0200063705444335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,64,power_law_1.01,0.03221119940280914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,64,power_law_1.01,0.03214719891548157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,1,power_law_1.2,1.0333056449890137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,64,power_law_1.01,0.0336896002292633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,64,power_law_1.01,0.03522560000419617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,64,power_law_1.01,0.037484800815582274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,1,power_law_1.2,1.1370944023132323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,64,power_law_1.01,0.03863039910793305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,2,balanced,0.05816533168156942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,2,balanced,0.06971199810504913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,2,balanced,0.09407466650009155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,2,balanced,0.1272266705830892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,64,power_law_1.01,0.04033919870853424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,2,balanced,0.19697066148122153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,2,balanced,0.26224533716837567
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,2,balanced,0.2650559941927592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,2,balanced,0.2659839987754822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,1,power_law_1.2,1.2059647560119628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,64,power_law_1.01,0.04423039853572845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,2,balanced,0.2683039903640747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,2,balanced,0.26827200253804523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,2,balanced,0.27000532547632855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,2,balanced,0.2754133343696594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,64,power_law_1.01,0.04769279956817627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,2,balanced,0.28068800767262775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,2,balanced,0.2848693331082662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,2,balanced,0.2924319903055827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,2,balanced,0.30144532521565753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,2,balanced,0.3152586619059245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,64,power_law_1.01,0.056364798545837404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,2,balanced,0.3466453154881795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,2,balanced,0.37459198633829754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,1,power_law_1.2,1.3118271827697754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,2,balanced,0.4944426616032918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,64,power_law_1.01,0.06090239882469177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,2,balanced,0.5155733426411947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,2,balanced,0.7884746392567953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,2,balanced,0.802127997080485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,64,power_law_1.01,0.06961920261383056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,2,balanced,1.197808027267456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,2,balanced,1.3268266518910725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,1,power_law_1.2,1.523750400543213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,64,power_law_1.01,0.08592640161514283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,2,balanced,1.9867253303527832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,2,balanced,2.4680320421854653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,64,power_law_1.01,0.11388800144195557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,8,2,balanced,3.1786346435546875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,1,power_law_1.2,1.5894271850585937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,64,power_law_1.01,0.14476799964904785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,8,2,balanced,4.719189325968425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,64,power_law_1.01,0.1879040002822876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,1,power_law_1.2,2.012019157409668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,64,power_law_1.01,0.21315839290618896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,8,2,balanced,9.049445470174154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,64,power_law_1.01,0.2663104057312012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,1,power_law_1.2,2.0500415802001952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,64,power_law_1.01,0.5189504146575927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,1,power_law_1.2,2.533184051513672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,64,power_law_1.01,0.7663296222686767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,1,power_law_1.2,3.1056127548217773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,1,power_law_1.2,3.962393569946289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,1,power_law_1.2,4.902694320678711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,32,power_law_1.01,0.03442560136318207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,1,power_law_1.2,5.871833419799804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,32,power_law_1.01,0.031040000915527343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,32,power_law_1.01,0.03144960105419159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,32,balanced,0.03160533308982849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,32,balanced,0.02961066613594691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,32,balanced,0.03124266614516576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,32,balanced,0.031770666440327965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,32,balanced,0.03161066770553589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,32,balanced,0.033376000821590424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,32,balanced,0.03359466542800268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,32,balanced,0.03532800078392029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,32,balanced,0.03369066615899404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,32,balanced,0.0347626656293869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,32,power_law_1.01,0.03023360073566437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,32,balanced,0.0352906659245491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,32,balanced,0.03577066709597906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,32,balanced,0.03752533346414566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,32,balanced,0.03734400123357773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,32,balanced,0.04162133236726125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,32,balanced,0.03977066775163015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,32,balanced,0.041834667325019836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,32,balanced,0.04353066782156626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,32,balanced,0.045647998650868736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,32,balanced,0.04798933366934458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,32,balanced,0.054085334142049156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,32,balanced,0.06667733192443848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,32,balanced,0.08974400162696838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,32,balanced,0.12158399820327759
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,32,balanced,0.15362133582433066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,1,power_law_1.2,8.803539276123047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,32,balanced,0.2081973354021708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,32,power_law_1.01,0.031411200761795044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,32,balanced,0.258842666943868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,32,balanced,0.31028799215952557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,32,balanced,0.46255465348561603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,32,balanced,0.867258628209432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,32,power_law_1.01,0.032595199346542356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,32,power_law_1.01,0.03213439881801605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,32,power_law_1.01,0.032307198643684386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,32,power_law_1.01,0.032543998956680295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,32,power_law_1.01,0.0333759993314743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,32,power_law_1.01,0.033881598711013795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,32,power_law_1.01,0.034694400429725644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,1,power_law_1.2,16.01947479248047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,32,power_law_1.01,0.03710080087184906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,32,power_law_1.01,0.03668479919433594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,32,power_law_1.01,0.04035840034484863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,32,power_law_1.01,0.04077439904212952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,32,power_law_1.01,0.04318720102310181
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,32,power_law_1.01,0.04620159864425659
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,32,power_law_1.01,0.05092480182647705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,32,power_law_1.01,0.055257600545883176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,32,power_law_1.01,0.06172159910202026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,16,power_law_1.01,0.09991040229797363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,32,power_law_1.01,0.08277119994163513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,32,power_law_1.01,0.09353600144386291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,16,power_law_1.01,0.15077120065689087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,32,power_law_1.01,0.13047679662704467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,16,power_law_1.01,0.19589120149612427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,32,power_law_1.01,0.17165440320968628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,16,power_law_1.01,0.07306879758834839
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,32,power_law_1.01,0.2285248041152954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,2,power_law_1.2,0.03110400140285492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,16,power_law_1.01,0.10431360006332398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,32,power_law_1.01,0.272108793258667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,2,power_law_1.2,0.03408640027046204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,2,power_law_1.2,0.03568640053272247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,2,power_law_1.2,0.04161919951438904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,16,power_law_1.01,0.1230463981628418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,32,power_law_1.01,0.41377921104431153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,2,power_law_1.2,0.048230400681495665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,2,power_law_1.2,0.0674560010433197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,16,power_law_1.01,0.1310528039932251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,2,power_law_1.2,0.07276160120964051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,32,power_law_1.01,0.5412799835205078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,2,power_law_1.2,0.07346559762954712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,2,power_law_1.2,0.07325440049171447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,16,power_law_1.01,0.15310720205307007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,2,power_law_1.2,0.07745280265808105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,2,power_law_1.2,0.07928320169448852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,16,power_law_1.01,0.1546623945236206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,2,power_law_1.2,0.08135679960250855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,32,power_law_1.01,0.9938559532165527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,2,power_law_1.2,0.08315520286560059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,2,power_law_1.2,0.08394880294799804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,16,power_law_1.01,0.15851520299911498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,2,power_law_1.2,0.09005439877510071
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,2,power_law_1.2,0.09529600143432618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,16,power_law_1.01,0.17119359970092773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,2,power_law_1.2,0.0991104006767273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,2,power_law_1.2,0.11251200437545776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,16,power_law_1.01,0.16992000341415406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,2,power_law_1.2,0.11639679670333862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,2,power_law_1.2,0.136953604221344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,2,power_law_1.2,0.16035200357437135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,16,power_law_1.01,0.17367680072784425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,2,power_law_1.2,0.18881280422210694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,2,power_law_1.2,0.2185663938522339
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,16,power_law_1.01,0.17737599611282348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,2,power_law_1.2,0.2867392063140869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,2,power_law_1.2,0.3704832077026367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,16,power_law_1.01,0.1902400016784668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,2,power_law_1.2,0.49548802375793455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,16,power_law_1.01,0.19032319784164428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,2,power_law_1.2,0.6642879962921142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,16,power_law_1.01,0.20880639553070068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,2,2,power_law_1.2,0.7766016006469727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,16,power_law_1.01,0.22936320304870605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,2,2,power_law_1.2,1.2020288467407227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,64,power_law_1.2,0.05260159969329834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,16,power_law_1.01,0.24780800342559814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,64,power_law_1.2,0.031891199946403506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,16,power_law_1.01,0.29084799289703367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,64,power_law_1.2,0.03333759903907776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,2,2,power_law_1.2,2.228767967224121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,16,power_law_1.01,0.3065984010696411
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,64,power_law_1.2,0.03172479867935181
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,16,power_law_1.01,0.3857919931411743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,64,power_law_1.2,0.031379199028015135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,64,power_law_1.2,0.03271040022373199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,16,power_law_1.01,0.4772160053253174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,64,power_law_1.2,0.0320576012134552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,16,power_law_1.01,0.6096255779266357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,64,power_law_1.2,0.03238399922847748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,64,power_law_1.2,0.032332798838615416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,16,power_law_1.01,0.6783487796783447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,64,power_law_1.2,0.032576000690460204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,64,power_law_1.2,0.03395200073719025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,16,power_law_1.01,0.9174912452697754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,64,power_law_1.2,0.034067198634147644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,64,power_law_1.2,0.034406399726867674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,64,power_law_1.2,0.03558399975299835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,16,power_law_1.01,1.177126407623291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,128,power_law_1.2,0.034534400701522826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,64,power_law_1.2,0.03976320028305054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,128,power_law_1.2,0.03489919900894165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,128,power_law_1.2,0.034585601091384886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,64,power_law_1.2,0.04007680118083954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,128,power_law_1.2,0.03468799889087677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,16,power_law_1.01,1.4114368438720704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,64,power_law_1.2,0.04202240109443665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,128,power_law_1.2,0.03531520068645477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,64,power_law_1.2,0.04557439982891083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,128,power_law_1.2,0.03603839874267578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,128,power_law_1.2,0.036396801471710205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,64,power_law_1.2,0.05014399886131286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,128,power_law_1.2,0.03644160032272339
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,64,power_law_1.2,0.05568000078201294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,128,power_law_1.2,0.03594880104064942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,16,power_law_1.01,2.0768064498901366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,128,power_law_1.2,0.03691520094871521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,64,power_law_1.2,0.06407039761543273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,128,power_law_1.2,0.03674240112304687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,64,power_law_1.2,0.07762560248374939
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,128,power_law_1.2,0.036857599020004274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,128,power_law_1.2,0.03710080087184906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,64,power_law_1.2,0.11184639930725097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,128,power_law_1.2,0.03804160058498383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,128,power_law_1.2,0.041440001130104064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,64,power_law_1.2,0.1529536008834839
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,128,power_law_1.2,0.041459199786186215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,16,power_law_1.01,4.252620697021484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,64,power_law_1.2,0.17123839855194092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,128,power_law_1.2,0.04203520119190216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,128,power_law_1.2,0.0430400013923645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,64,power_law_1.2,0.240447998046875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,128,power_law_1.2,0.04514560103416443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,128,power_law_1.2,0.04888319969177246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,64,power_law_1.2,0.3054336071014404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,128,power_law_1.2,0.05518720149993896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,128,power_law_1.2,0.06705920100212097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,1,64,power_law_1.2,0.39495038986206055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,128,power_law_1.2,0.08398720026016235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,128,power_law_1.2,0.10914560556411743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,128,power_law_1.2,0.1457919955253601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,1,64,power_law_1.2,0.6887296199798584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,128,power_law_1.2,0.19863040447235109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,128,power_law_1.2,0.2794496059417725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,1,64,power_law_1.2,1.3324031829833984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,2,128,power_law_1.2,0.3309760093688965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,2,128,power_law_1.2,0.5345600128173829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,2,128,power_law_1.2,1.1085311889648437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,16,2,balanced,0.034261333445707955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,16,2,balanced,0.03807999938726425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,16,2,balanced,0.03586133321126302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,16,2,balanced,0.037903999288876854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,16,2,balanced,0.03755733370780945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,16,2,balanced,0.03585600107908249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,16,2,balanced,0.03787733366092046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,16,2,balanced,0.03808533400297165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,16,2,balanced,0.037658666570981346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,16,2,balanced,0.03798400113979975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,16,2,balanced,0.037808001041412354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,16,2,balanced,0.039664000272750854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,16,2,balanced,0.039621333281199135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,16,2,balanced,0.040037333965301514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,16,2,balanced,0.04602666695912679
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,16,2,balanced,0.04665066798528036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,16,2,balanced,0.04993600149949392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,16,2,balanced,0.05808533231417338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,16,2,balanced,0.06052266558011373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,16,2,balanced,0.07272533575693767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,16,2,balanced,0.08268266419569652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,16,2,balanced,0.10989866654078166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,16,2,balanced,0.12308800220489502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,16,2,balanced,0.1606826682885488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,16,2,balanced,0.19834667444229126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,16,2,balanced,0.2812533378601074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,16,2,balanced,0.3586719830830892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,16,2,balanced,0.4410026470820109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,16,2,balanced,0.6781173547108968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,16,2,balanced,1.3044533729553223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,1,4,power_law_1.01,0.07433599829673768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,1,4,power_law_1.01,0.07774720191955567
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,1,4,power_law_1.01,0.08625919818878174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,1,4,power_law_1.01,0.08755840063095092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,1,4,power_law_1.01,0.08748800158500672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,1,4,power_law_1.01,0.08922880291938781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,1,4,power_law_1.01,0.09322239756584168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,1,4,power_law_1.01,0.0967743992805481
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,1,4,power_law_1.01,0.10152319669723511
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,1,4,power_law_1.01,0.10713599920272827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,1,4,power_law_1.01,0.10001920461654663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,1,4,power_law_1.01,0.10349440574645996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,1,4,power_law_1.01,0.10663679838180543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,1,4,power_law_1.01,0.11778559684753417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,1,4,power_law_1.01,0.12789759635925294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,1,4,power_law_1.01,0.13401600122451782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,1,4,power_law_1.01,0.16087679862976073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,1,4,power_law_1.01,0.208076810836792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,1,4,power_law_1.01,0.23291521072387694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,1,4,power_law_1.01,0.2526400089263916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,1,4,power_law_1.01,0.38184959888458253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,1,4,power_law_1.01,0.6297472000122071
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,1,4,power_law_1.01,0.5247360229492187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,1,4,power_law_1.01,0.8566080093383789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,1,balanced,0.052202666799227394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,1,balanced,0.091648002465566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,1,balanced,0.10492799679438274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,1,balanced,0.15871999661127725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,1,balanced,0.26411734024683636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,1,4,power_law_1.01,1.4093440055847168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,1,balanced,0.3195040027300517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,1,balanced,0.32252800464630127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,1,balanced,0.3252906600634257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,1,balanced,0.3277759949366252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,1,balanced,0.32876267035802204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,1,balanced,0.33641600608825684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,1,balanced,0.34120531876881915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,1,balanced,0.34601600964864093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,1,4,power_law_1.01,2.0415231704711916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,1,balanced,0.35148266951243085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,1,balanced,0.36370666821797687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,1,balanced,0.3744959831237793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,1,balanced,0.38867731889088947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,1,balanced,0.42420268058776855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,1,balanced,0.454037348429362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,1,balanced,0.5762720108032227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,1,balanced,0.6146986484527588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,1,4,power_law_1.01,2.5661376953125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,1,balanced,0.9181013107299805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,1,balanced,0.9317599932352701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,1,balanced,1.3885599772135417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,1,balanced,1.6280959447224934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,1,4,power_law_1.01,3.182796859741211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,1,balanced,2.441434701283773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,1,balanced,3.035365422566732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,1,4,power_law_1.01,4.752108764648438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,1,balanced,3.5727361043294272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,1,balanced,5.655626932779948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,1,4,power_law_1.01,10.118093109130859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,1,balanced,10.833338419596354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,1,8,balanced,0.06724800169467926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,1,8,balanced,0.07896000146865845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,1,8,balanced,0.06566933294137318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,1,8,balanced,0.0665280024210612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,1,8,balanced,0.06645333270231883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,1,8,balanced,0.06615466872851054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,1,8,balanced,0.06648533542950948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,1,8,balanced,0.06693866848945618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,1,8,balanced,0.06833066542943318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,1,8,balanced,0.06794133285681407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,1,8,balanced,0.06950933237870534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,1,8,balanced,0.07057066758473714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,1,8,balanced,0.06956799825032552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,1,8,balanced,0.07452266911665599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,1,8,balanced,0.07918933530648549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,1,8,balanced,0.08118933439254761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,1,8,balanced,0.08592533071835835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,1,8,balanced,0.09268266956011455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,1,8,balanced,0.09542399644851685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,1,8,balanced,0.1362879971663157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,1,8,balanced,0.1304693321386973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,1,8,balanced,0.17777599891026816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,1,8,balanced,0.20367467403411865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,1,8,balanced,0.2796853383382161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,1,8,balanced,0.34699201583862305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,1,8,balanced,0.49353599548339844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,1,8,balanced,0.6423999865849813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,1,8,balanced,0.7873067061106364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,1,8,balanced,1.1986400286356609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,128,power_law_1.2,0.11208319664001465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,1,8,balanced,2.368096033732096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,128,power_law_1.2,0.10738559961318969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,128,power_law_1.2,0.10550400018692016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,128,power_law_1.2,0.09242879748344421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,128,power_law_1.2,0.10149760246276855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,128,power_law_1.2,0.10900479555130005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,64,power_law_1.2,0.03091199994087219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,128,power_law_1.2,0.10474879741668701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,64,power_law_1.2,0.031040000915527343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,64,power_law_1.2,0.03056640028953552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,64,power_law_1.2,0.03154560029506683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,64,power_law_1.2,0.03086079955101013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,128,power_law_1.2,0.11531519889831543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,64,power_law_1.2,0.03242239952087402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,64,power_law_1.2,0.03253760039806366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,64,power_law_1.2,0.03249920010566711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,128,power_law_1.2,0.10820480585098266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,64,power_law_1.2,0.03287039995193482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,64,power_law_1.2,0.03311359882354736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,64,power_law_1.2,0.034860798716545106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,128,power_law_1.2,0.11856640577316284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,64,power_law_1.2,0.03493120074272156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,64,power_law_1.2,0.03529599905014038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,64,power_law_1.2,0.038515201210975646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,64,power_law_1.2,0.04255360066890716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,128,power_law_1.2,0.11324160099029541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,64,power_law_1.2,0.0440064013004303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,64,power_law_1.2,0.04642559885978699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,64,power_law_1.2,0.05496320128440857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,128,power_law_1.2,0.11431679725646973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,64,power_law_1.2,0.06004480123519897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,64,power_law_1.2,0.06974719762802124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,64,power_law_1.2,0.09191679954528809
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,2,power_law_1.2,0.054028797149658206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,128,power_law_1.2,0.10830080509185791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,64,power_law_1.2,0.11140480041503906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,2,power_law_1.2,0.059545600414276124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,64,power_law_1.2,0.14741120338439942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,2,power_law_1.2,0.06539520025253295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,64,power_law_1.2,0.22916479110717775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,2,power_law_1.2,0.08067200183868409
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,128,power_law_1.2,0.10688639879226684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,2,power_law_1.2,0.09758719801902771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,64,power_law_1.2,0.28857600688934326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,2,power_law_1.2,0.14515199661254882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,64,power_law_1.2,0.4234816074371338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,2,power_law_1.2,0.1585088014602661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,128,power_law_1.2,0.12044800519943237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,2,power_law_1.2,0.1598080039024353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,64,power_law_1.2,0.5813055992126465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,2,power_law_1.2,0.17427200078964233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,128,power_law_1.2,0.11210240125656128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,2,power_law_1.2,0.17377279996871947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,64,power_law_1.2,0.7029888153076171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,2,power_law_1.2,0.18186880350112916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,2,power_law_1.2,0.17986559867858887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,128,power_law_1.2,0.11719679832458496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,64,power_law_1.2,1.2256640434265136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,2,power_law_1.2,0.19111679792404174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,2,power_law_1.2,0.19997440576553344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,2,power_law_1.2,0.2020927906036377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,128,power_law_1.2,0.13495680093765258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,2,power_law_1.2,0.21262080669403077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,64,power_law_1.2,2.378982353210449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,2,power_law_1.2,0.22578558921813965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,2,power_law_1.2,0.2549439907073975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,128,power_law_1.2,0.1531775951385498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,2,power_law_1.2,0.271616005897522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,2,power_law_1.2,0.3255359888076782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,128,power_law_1.2,0.17923840284347534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,2,power_law_1.2,0.3919424057006836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,2,power_law_1.2,0.5043776035308838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,128,power_law_1.2,0.20624001026153566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,2,power_law_1.2,0.6250944137573242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,2,power_law_1.2,0.8030143737792969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,128,power_law_1.2,0.26591360569000244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,2,power_law_1.2,1.0324224472045898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,2,power_law_1.2,1.3495360374450684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,128,power_law_1.2,0.3091840028762817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,2,power_law_1.2,1.7404096603393555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,128,power_law_1.2,0.4359936237335205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,2,power_law_1.2,2.2205184936523437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,128,power_law_1.2,0.5336319923400878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,2,power_law_1.2,3.4912704467773437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,128,power_law_1.2,0.8325504302978516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,2,power_law_1.2,5.861209487915039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,128,power_law_1.2,1.1134528160095214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,128,power_law_1.2,1.2848447799682616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,128,power_law_1.2,2.326470375061035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,128,power_law_1.2,4.655551910400391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,4,balanced,0.05420266588528951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,4,balanced,0.058186665177345276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,4,balanced,0.061994666854540505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,4,balanced,0.08959466218948364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,4,balanced,0.12429333726565044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,4,balanced,0.181877334912618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,4,balanced,0.1818293333053589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,4,balanced,0.18521066506703696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,4,balanced,0.18758400281270346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,4,balanced,0.18650132417678833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,4,balanced,0.19033600886662802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,4,balanced,0.1902773380279541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,4,balanced,0.19497066736221313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,4,balanced,0.19442667563756308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,4,balanced,0.2032053271929423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,4,balanced,0.20679465929667154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,4,balanced,0.21198399861653647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,4,balanced,0.22740799188613892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,4,balanced,0.24499199787775675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,4,balanced,0.27508799235026044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,4,balanced,0.30899200836817425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,4,balanced,0.35717864831288654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,4,balanced,0.40382933616638184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,4,balanced,0.5620106856028239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,4,balanced,0.6492000023523966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,4,balanced,0.9242346286773682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,4,balanced,1.1597546736399333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,4,balanced,1.4458719889322917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,4,balanced,2.2161280314127603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,4,balanced,4.210570653279622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,4,power_law_1.2,0.043532800674438474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,4,power_law_1.2,0.038848000764846805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,4,power_law_1.2,0.040575999021530154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,4,power_law_1.2,0.04188799858093262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,4,power_law_1.2,0.04730879962444305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,32,power_law_1.2,0.03671680092811584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,4,power_law_1.2,0.05598719716072083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,32,power_law_1.2,0.0328575998544693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,4,power_law_1.2,0.05664640069007874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,32,power_law_1.2,0.03271040022373199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,32,power_law_1.2,0.03242239952087402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,4,power_law_1.2,0.05873280167579651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,32,power_law_1.2,0.032364800572395325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,32,power_law_1.2,0.0341376006603241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,4,power_law_1.2,0.058764797449111936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,32,power_law_1.2,0.03400320112705231
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,4,power_law_1.2,0.05865600109100342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,32,power_law_1.2,0.03485440015792847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,32,power_law_1.2,0.03484799861907959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,4,power_law_1.2,0.061478400230407716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,32,power_law_1.2,0.03516159951686859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,4,power_law_1.2,0.06236799955368042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,32,power_law_1.2,0.03558399975299835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,32,power_law_1.2,0.0361407995223999
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,4,power_law_1.2,0.06471679806709289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,32,power_law_1.2,0.03692800104618073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,32,power_law_1.2,0.037651199102401736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,4,power_law_1.2,0.07019519805908203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,32,power_law_1.2,0.040729600191116336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,4,power_law_1.2,0.07896320223808288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,32,power_law_1.2,0.04160000085830688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,32,power_law_1.2,0.04294399917125702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,4,power_law_1.2,0.08095359802246094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,32,power_law_1.2,0.04741120040416717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,4,power_law_1.2,0.08936960101127625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,32,power_law_1.2,0.050937598943710326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,4,power_law_1.2,0.10346239805221558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,32,power_law_1.2,0.0606656014919281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,32,power_law_1.2,0.06452479958534241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,4,power_law_1.2,0.11015039682388306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,32,power_law_1.2,0.0756928026676178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,4,power_law_1.2,0.12832000255584716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,32,power_law_1.2,0.0946943998336792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,32,power_law_1.2,0.1285248041152954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,4,power_law_1.2,0.1523967981338501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,32,power_law_1.2,0.19625600576400756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,4,power_law_1.2,0.19431040287017823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,32,power_law_1.2,0.22272000312805176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,4,power_law_1.2,0.26232960224151614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,32,power_law_1.2,0.34267520904541016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,4,power_law_1.2,0.36126079559326174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,32,power_law_1.2,0.39203200340270994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,4,power_law_1.2,0.49222397804260254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,4,power_law_1.2,0.7495935916900635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,32,power_law_1.2,0.7745279788970947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,4,power_law_1.2,1.0377280235290527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,32,power_law_1.2,1.5279999732971192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,4,power_law_1.2,1.1964480400085449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,4,power_law_1.2,1.9246208190917968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,4,power_law_1.2,4.025811386108399
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,64,power_law_1.01,0.040556800365447995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,64,power_law_1.01,0.03511039912700653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,64,power_law_1.01,0.034995201230049136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,64,power_law_1.01,0.036287999153137206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,64,power_law_1.01,0.03589119911193848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,64,power_law_1.01,0.0353408008813858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,64,power_law_1.01,0.036032000184059144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,64,power_law_1.01,0.03611519932746887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,64,power_law_1.01,0.03615359961986542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,64,power_law_1.01,0.03639039993286133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,32,8,power_law_1.01,0.03335680067539215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,64,power_law_1.01,0.036652800440788266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,32,8,power_law_1.01,0.03285120129585266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,32,8,power_law_1.01,0.03225600123405457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,64,power_law_1.01,0.0379967987537384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,32,8,power_law_1.01,0.03146879971027374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,64,power_law_1.01,0.03864319920539856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,32,8,power_law_1.01,0.03146879971027374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,64,power_law_1.01,0.040703999996185306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,32,8,power_law_1.01,0.03328000009059906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,32,8,power_law_1.01,0.03329919874668121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,64,power_law_1.01,0.04426240026950836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,32,8,power_law_1.01,0.033542400598526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,64,power_law_1.01,0.044614401459693906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,32,8,power_law_1.01,0.03406080007553101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,32,8,power_law_1.01,0.034380799531936644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,64,power_law_1.01,0.04989440143108368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,32,8,power_law_1.01,0.03560959994792938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,32,8,power_law_1.01,0.036671999096870425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,64,power_law_1.01,0.05280640125274658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,32,8,power_law_1.01,0.037062400579452516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,64,power_law_1.01,0.05677440166473389
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,32,8,power_law_1.01,0.03765760064125061
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,32,8,power_law_1.01,0.045311999320983884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,64,power_law_1.01,0.06795520186424256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,32,8,power_law_1.01,0.04547840058803558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,32,8,power_law_1.01,0.0469760000705719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,64,power_law_1.01,0.07285119891166687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,32,8,power_law_1.01,0.05920000076293945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,32,8,power_law_1.01,0.06185600161552429
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,64,power_law_1.01,0.09340800046920776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,32,8,power_law_1.01,0.07799680233001709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,32,8,power_law_1.01,0.1007807970046997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,64,power_law_1.01,0.12029440402984619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,32,8,power_law_1.01,0.14075520038604736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,32,8,power_law_1.01,0.1695039987564087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,64,power_law_1.01,0.1648319959640503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,32,8,power_law_1.01,0.24840960502624512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,32,8,power_law_1.01,0.30628480911254885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,64,power_law_1.01,0.1712000012397766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,32,8,power_law_1.01,0.45061120986938474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,64,power_law_1.01,0.25420799255371096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,32,8,power_law_1.01,0.5740992069244385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,32,8,power_law_1.01,0.6507904052734375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,64,power_law_1.01,0.3138943910598755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,32,8,power_law_1.01,1.0520383834838867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,64,power_law_1.01,0.4128960132598877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,32,8,power_law_1.01,2.1067455291748045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,64,power_law_1.01,0.6557568073272705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,64,power_law_1.01,1.2566975593566894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,1,power_law_1.01,0.0648576021194458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,1,power_law_1.01,0.07034879922866821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,2,power_law_1.2,0.06570879817008972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,1,power_law_1.01,0.07912319898605347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,2,power_law_1.2,0.07445120215415954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,2,power_law_1.2,0.08275840282440186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,1,power_law_1.01,0.10974080562591552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,2,power_law_1.2,0.09646080136299133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,1,power_law_1.01,0.13508479595184325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,2,power_law_1.2,0.12234879732131958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,1,power_law_1.01,0.1656448006629944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,2,power_law_1.2,0.14940160512924194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,1,power_law_1.01,0.21438078880310057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,2,power_law_1.2,0.16733440160751342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,1,power_law_1.01,0.22683520317077638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,2,power_law_1.2,0.2123647928237915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,1,power_law_1.01,0.2366719961166382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,2,power_law_1.2,0.22116479873657227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,1,power_law_1.01,0.24039039611816407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,2,power_law_1.2,0.23359360694885253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,1,power_law_1.01,0.24642560482025147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,2,power_law_1.2,0.2450239896774292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,1,power_law_1.01,0.2604736089706421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,2,power_law_1.2,0.2507967948913574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,1,power_law_1.01,0.2659199953079224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,2,power_law_1.2,0.26299519538879396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,1,power_law_1.01,0.27594239711761476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,2,power_law_1.2,0.2747136116027832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,1,power_law_1.01,0.28703999519348145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,2,power_law_1.2,0.27781119346618655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,1,power_law_1.01,0.29341440200805663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,2,power_law_1.2,0.2938944101333618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,1,power_law_1.01,0.3111104011535645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,2,power_law_1.2,0.3044480085372925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,1,power_law_1.01,0.3638463973999023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,2,power_law_1.2,0.33879680633544923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,1,power_law_1.01,0.39724159240722656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,2,power_law_1.2,0.37080960273742675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,1,power_law_1.01,0.4884607791900635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,2,power_law_1.2,0.4346816062927246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,1,power_law_1.01,0.5625855922698975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,2,power_law_1.2,0.4973696231842041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,1,power_law_1.01,0.7282879829406739
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,2,power_law_1.2,0.6281407833099365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,1,power_law_1.01,0.8810175895690918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,2,power_law_1.2,0.7069503784179687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,1,power_law_1.01,1.2040639877319337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,2,power_law_1.2,0.9364735603332519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,1,power_law_1.01,1.5159168243408203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,2,power_law_1.2,1.14835844039917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,1,power_law_1.01,2.1535167694091797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,2,power_law_1.2,1.6319999694824219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,1,power_law_1.01,2.8021823883056642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,2,power_law_1.2,2.0662912368774413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,1,power_law_1.01,3.4271488189697266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,2,power_law_1.2,2.5881343841552735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,2,power_law_1.2,3.8210113525390623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,1,power_law_1.01,5.421113586425781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,2,power_law_1.2,7.932288360595703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,1,power_law_1.01,10.629388427734375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,128,balanced,0.041797334949175514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,128,balanced,0.04164266586303711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,128,balanced,0.03955733279387156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,128,balanced,0.04145599901676178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,128,balanced,0.04196799794832865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,128,balanced,0.04161600023508072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,128,balanced,0.043050666650136314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,128,balanced,0.0408746674656868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,128,balanced,0.04277333120505015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,128,balanced,0.04295999805132548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,128,balanced,0.04242133100827535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,128,balanced,0.0409706657131513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,128,balanced,0.04318400224049886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,128,balanced,0.043605332573254905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,128,balanced,0.04667733112970988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,128,balanced,0.045114666223526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,128,balanced,0.045514668027559914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,128,balanced,0.047237331668535866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,128,balanced,0.04747733473777771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,128,balanced,0.04914666712284088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,128,balanced,0.04930133124192556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,128,balanced,0.0558240016301473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,128,balanced,0.057029331723848976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,128,balanced,0.06506666541099548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,128,balanced,0.07343466579914093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,128,balanced,0.08961600065231323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,128,balanced,0.10382933417956035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,1,128,balanced,0.12276800473531087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,1,128,balanced,0.17149333159128824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,1,128,balanced,0.2977973421414693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,64,balanced,0.049733335773150124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,64,balanced,0.053488001227378845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,64,balanced,0.0552106648683548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,64,balanced,0.07380266487598419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,64,balanced,0.08218133449554443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,64,balanced,0.08548800150553386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,64,balanced,0.08306133250395457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,64,balanced,0.0830506682395935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,64,balanced,0.08509332935015361
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,64,balanced,0.08585066596666972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,64,balanced,0.08488000432650249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,64,balanced,0.08941866954167683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,64,balanced,0.08790399630864461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,64,balanced,0.08739733695983887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,64,balanced,0.09488000472386678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,64,balanced,0.09507733583450317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,64,balanced,0.09980799754460652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,64,balanced,0.11471466223398845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,64,balanced,0.11813867092132568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,64,balanced,0.14475199580192566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,64,balanced,0.15702933073043823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,64,balanced,0.20067733526229858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,64,balanced,0.2049600084622701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,64,balanced,0.2614346742630005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,64,balanced,0.3011626601219177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,64,balanced,0.4039040009180705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,64,balanced,0.49404799938201904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,1,64,balanced,0.5925386746724447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,1,64,balanced,0.864736000696818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,1,64,balanced,1.8266293207804363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,64,power_law_1.2,0.08451840281486511
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,64,power_law_1.2,0.08046079874038696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,64,power_law_1.2,0.07794560194015503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,64,power_law_1.2,0.06421759724617004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,64,power_law_1.2,0.06823679804801941
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,64,power_law_1.2,0.06853119730949402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,64,power_law_1.2,0.0706496000289917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,64,power_law_1.2,0.07365760207176208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,64,power_law_1.2,0.06963199973106385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,64,power_law_1.2,0.07189120054244995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,64,power_law_1.2,0.0762880027294159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,64,power_law_1.2,0.07592960000038147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,64,power_law_1.2,0.07588480114936828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,64,power_law_1.2,0.07694079875946044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,64,power_law_1.2,0.08108800053596496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,64,power_law_1.2,0.08360959887504578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,64,power_law_1.2,0.0898751974105835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,64,power_law_1.2,0.10186879634857178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,64,power_law_1.2,0.11270400285720825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,64,power_law_1.2,0.13751039505004883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,64,power_law_1.2,0.15928959846496582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,64,power_law_1.2,0.1993407964706421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,64,power_law_1.2,0.254419207572937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,64,power_law_1.2,0.3278847932815552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,64,power_law_1.2,0.4075200080871582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,64,power_law_1.2,0.6171072006225586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,64,power_law_1.2,0.7925375938415528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,64,power_law_1.2,1.0304896354675293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,64,power_law_1.2,1.6902912139892579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,64,power_law_1.2,3.0778751373291016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,2,power_law_1.2,0.052108800411224364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,2,power_law_1.2,0.05367680191993714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,2,power_law_1.2,0.05792639851570129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,2,power_law_1.2,0.06771199703216553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,2,power_law_1.2,0.08526080250740051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,2,power_law_1.2,0.09999359846115112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,2,power_law_1.2,0.13611520528793336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,2,power_law_1.2,0.13934719562530518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,2,power_law_1.2,0.14479360580444336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,2,power_law_1.2,0.14874240159988403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,2,power_law_1.2,0.1590783953666687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,2,power_law_1.2,0.17163519859313964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,2,power_law_1.2,0.18625919818878173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,2,power_law_1.2,0.19111039638519287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,2,power_law_1.2,0.21175680160522461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,2,power_law_1.2,0.22050559520721436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,2,power_law_1.2,0.21890559196472167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,2,power_law_1.2,0.25413761138916013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,2,power_law_1.2,0.2694144010543823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,2,power_law_1.2,0.3222592115402222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,2,power_law_1.2,0.34876160621643065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,2,power_law_1.2,0.4473599910736084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,2,power_law_1.2,0.5625855922698975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,2,power_law_1.2,0.7088960170745849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,2,power_law_1.2,0.931385612487793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,2,power_law_1.2,1.2865344047546388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,2,power_law_1.2,1.7171327590942382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,2,power_law_1.2,2.145587158203125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,2,power_law_1.2,3.138924789428711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,2,power_law_1.2,6.037011337280274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,16,power_law_1.2,0.07278720140457154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,16,power_law_1.2,0.09653120040893555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,16,power_law_1.2,0.09968640208244324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,16,power_law_1.2,0.05985280275344849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,16,power_law_1.2,0.06814079880714416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,16,power_law_1.2,0.08170239925384522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,16,power_law_1.2,0.09088000059127807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,16,power_law_1.2,0.10077439546585083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,64,balanced,0.0562720000743866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,64,balanced,0.058037335673967995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,64,balanced,0.05670933425426483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,64,balanced,0.060122668743133545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,64,balanced,0.06015466650327047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,64,balanced,0.06032533446947733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,64,balanced,0.06211733321348826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,64,balanced,0.06201066573460897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,64,balanced,0.0621919979651769
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,64,balanced,0.061994666854540505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,64,balanced,0.062074666221936546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,64,balanced,0.0643039991458257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,64,balanced,0.06451199948787689
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,64,balanced,0.06412800153096516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,64,balanced,0.07022400200366974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,64,balanced,0.07034666836261749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,64,balanced,0.07467199862003326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,64,balanced,0.08272533118724823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,64,balanced,0.08966400225957234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,64,balanced,0.12172800302505493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,64,balanced,0.12945600350697836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,64,balanced,0.17712533473968506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,64,balanced,0.17594132820765176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,64,balanced,0.226474662621816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,64,balanced,0.24928534030914307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,64,balanced,0.3347626527150472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,16,power_law_1.2,0.10039039850234985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,64,balanced,0.3992533286412557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,2,64,balanced,0.48456533749898273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,2,64,balanced,0.7052960395812988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,2,64,balanced,1.5139892896016438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,16,power_law_1.2,0.10444799661636353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,16,power_law_1.2,0.10525439977645874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,16,power_law_1.2,0.10706559419631959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,16,power_law_1.2,0.12012159824371338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,16,power_law_1.2,0.11548800468444824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,16,power_law_1.2,0.12814079523086547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,16,power_law_1.2,0.13671679496765138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,2,power_law_1.01,0.03797119855880737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,16,power_law_1.2,0.13514879941940308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,2,power_law_1.01,0.04421760141849518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,2,power_law_1.01,0.04699519872665405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,2,power_law_1.01,0.06053760051727295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,16,power_law_1.2,0.14864640235900878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,2,power_law_1.01,0.07401599884033203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,2,power_law_1.01,0.10613759756088256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,2,power_law_1.01,0.10666879415512084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,2,power_law_1.01,0.1126911997795105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,16,power_law_1.2,0.16828800439834596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,2,power_law_1.01,0.1168511986732483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,2,power_law_1.01,0.11918720006942748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,2,power_law_1.01,0.11768319606781005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,16,power_law_1.2,0.20032639503479005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,2,power_law_1.01,0.12294399738311768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,2,power_law_1.01,0.12466559410095215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,2,power_law_1.01,0.13226879835128785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,16,power_law_1.2,0.20577919483184814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,2,power_law_1.01,0.13443200588226317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,2,power_law_1.01,0.14370559453964232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,2,power_law_1.01,0.1554111957550049
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,16,power_law_1.2,0.2859839916229248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,2,power_law_1.01,0.16637439727783204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,2,power_law_1.01,0.17984639406204223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,2,power_law_1.01,0.21461760997772217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,2,power_law_1.01,0.2452608108520508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,16,power_law_1.2,0.30792319774627686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,2,power_law_1.01,0.3059776067733765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,2,power_law_1.01,0.3598144054412842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,16,power_law_1.2,0.4245567798614502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,2,power_law_1.01,0.48663039207458497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,2,power_law_1.01,0.6302591800689697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,2,power_law_1.01,0.9213055610656739
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,16,power_law_1.2,0.4904895782470703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,2,power_law_1.01,1.0787455558776855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,16,power_law_1.2,0.674131202697754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,2,power_law_1.01,1.3788479804992675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,2,power_law_1.01,2.053401565551758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,16,power_law_1.2,0.828825569152832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,16,power_law_1.2,1.0486656188964845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,2,power_law_1.01,3.9752769470214844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,16,power_law_1.2,1.865331268310547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,16,power_law_1.2,3.1632320404052736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,8,1,balanced,0.035616000493367515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,8,1,balanced,0.06985599795977275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,8,1,balanced,0.08098666866620381
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,8,1,balanced,0.08137066662311554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,8,1,balanced,0.08229866623878479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,8,1,balanced,0.08381332953770955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,8,1,balanced,0.08374399940172832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,8,1,balanced,0.08349333206812541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,8,1,balanced,0.08342400193214417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,8,1,balanced,0.08290666838486989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,8,1,balanced,0.08690667152404785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,8,1,balanced,0.08557867010434468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,8,1,balanced,0.08505066235860188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,8,1,balanced,0.09087466200192769
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,8,1,balanced,0.09840533137321472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,8,1,balanced,0.10143466790517171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,8,1,balanced,0.10362666845321655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,8,1,balanced,0.12145066261291504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,8,1,balanced,0.12388267119725545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,8,1,balanced,0.15214932958285013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,8,1,balanced,0.1648319959640503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,8,1,balanced,0.22565333048502603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,8,1,balanced,0.26864532629648846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,8,1,balanced,0.3720746835072835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,8,1,balanced,0.47069867451985675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,8,1,balanced,0.6768426895141602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,8,1,balanced,0.873189369837443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,8,1,balanced,1.0797759691874187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,8,1,balanced,1.6924106280008953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,8,1,balanced,3.3100481033325195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,32,2,power_law_1.2,0.04821760058403015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,32,2,power_law_1.2,0.05204480290412903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,32,2,power_law_1.2,0.05428479909896851
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,32,2,power_law_1.2,0.06213120222091675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,32,2,power_law_1.2,0.0715071976184845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,32,2,power_law_1.2,0.08732159733772278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,4,power_law_1.2,0.18417919874191285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,32,2,power_law_1.2,0.10975359678268433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,32,2,power_law_1.2,0.11471359729766846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,4,power_law_1.2,0.15174399614334105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,32,2,power_law_1.2,0.1194111943244934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,4,power_law_1.2,0.15946240425109864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,32,2,power_law_1.2,0.12077440023422241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,4,power_law_1.2,0.23482239246368408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,32,2,power_law_1.2,0.12726399898529053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,32,2,power_law_1.2,0.12810239791870118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,4,power_law_1.2,0.3187328100204468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,32,2,power_law_1.2,0.13065600395202637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,4,power_law_1.2,0.6132031917572022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,32,2,power_law_1.2,0.13836159706115722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,4,power_law_1.2,0.6487167835235595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,32,2,power_law_1.2,0.145907199382782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,32,2,power_law_1.2,0.1509119987487793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,4,power_law_1.2,0.6928959846496582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,32,2,power_law_1.2,0.16412800550460815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,4,power_law_1.2,0.7029183864593506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,32,2,power_law_1.2,0.19095040559768678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,4,power_law_1.2,0.7261055946350098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,32,2,power_law_1.2,0.22208640575408936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,32,2,power_law_1.2,0.26245760917663574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,4,power_law_1.2,0.7699903964996337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,32,2,power_law_1.2,0.30750720500946044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,4,power_law_1.2,0.7743680000305175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,32,2,power_law_1.2,0.4065663814544678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,4,power_law_1.2,0.7839935779571533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,32,2,power_law_1.2,0.5125247955322265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,4,power_law_1.2,0.8097151756286621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,32,2,power_law_1.2,0.6861184120178223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,32,2,power_law_1.2,0.8769023895263672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,4,power_law_1.2,0.8585344314575195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,32,2,power_law_1.2,1.2660223960876464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,4,power_law_1.2,0.874944019317627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,32,2,power_law_1.2,1.527129554748535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,4,power_law_1.2,0.9506688117980957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,4,power_law_1.2,1.041990375518799
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,32,2,power_law_1.2,1.8997631072998047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,4,power_law_1.2,1.0559871673583985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,32,2,power_law_1.2,3.064179229736328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,4,power_law_1.2,1.2124544143676759
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,4,power_law_1.2,1.3210176467895507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,32,2,power_law_1.2,5.86767349243164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,4,power_law_1.2,1.4894783973693848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,4,power_law_1.2,1.6202432632446289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,4,power_law_1.2,1.993734359741211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,4,power_law_1.2,2.549331283569336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,4,power_law_1.2,3.244192123413086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,4,2,power_law_1.01,0.04580479860305786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,4,2,power_law_1.01,0.041465601325035094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,4,power_law_1.2,3.668889617919922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,4,2,power_law_1.01,0.04192639887332916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,4,2,power_law_1.01,0.04482559859752655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,4,2,power_law_1.01,0.04903680086135864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,4,2,power_law_1.01,0.05804799795150757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,1,4,power_law_1.2,4.603980636596679
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,4,2,power_law_1.01,0.058329600095748904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,4,2,power_law_1.01,0.058841598033905027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,4,2,power_law_1.01,0.06044160127639771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,4,2,power_law_1.01,0.060736000537872314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,4,2,power_law_1.01,0.062041598558425906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,4,2,power_law_1.01,0.06438400149345398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,4,2,power_law_1.01,0.0670527994632721
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,1,4,power_law_1.2,7.368326568603516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,4,2,power_law_1.01,0.0685375988483429
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,4,2,power_law_1.01,0.07327359914779663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,4,2,power_law_1.01,0.07590399980545044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,4,2,power_law_1.01,0.08316159844398499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,4,2,power_law_1.01,0.09304959774017334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,4,2,power_law_1.01,0.10894080400466918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,4,2,power_law_1.01,0.13725440502166747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,4,2,power_law_1.01,0.1633407950401306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,4,2,power_law_1.01,0.2086656093597412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,4,2,power_law_1.01,0.2516671895980835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,4,2,power_law_1.01,0.36985599994659424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,1,4,power_law_1.2,14.054464721679688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,4,2,power_law_1.01,0.4933568000793457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,4,2,power_law_1.01,0.6876031875610351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,4,2,power_law_1.01,0.928217601776123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,4,2,power_law_1.01,1.1532223701477051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,4,2,power_law_1.01,1.7531455993652343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,4,2,power_law_1.01,3.5441024780273436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,2,balanced,0.07436800003051758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,2,balanced,0.09737599889437358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,2,balanced,0.1460853318373362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,2,balanced,0.2449386715888977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,2,balanced,0.4389226833979289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,2,balanced,0.8333653608957926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,2,balanced,1.2215147018432617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,2,balanced,1.2254719734191895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,2,balanced,1.2236479918162029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,2,balanced,1.2249333063761394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,2,balanced,1.2297226587931316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,2,balanced,1.2355039914449055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,2,balanced,1.242250680923462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,2,balanced,1.2481013139088948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,2,balanced,1.2637279828389485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,2,balanced,1.2688159942626953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,2,balanced,1.2890933354695637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,2,balanced,1.316373348236084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,2,balanced,1.3460799853007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,2,balanced,1.4048372904459636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,2,balanced,1.4572800000508626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,2,balanced,1.569434642791748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,2,balanced,1.6797226270039876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,2,balanced,1.8794186909993489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,2,balanced,2.16976531346639
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,2,balanced,2.544816017150879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,2,balanced,3.361781438191732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,2,balanced,4.082293192545573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,2,balanced,5.831535975138347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,2,balanced,11.050773620605469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,1,power_law_1.2,0.04869759976863861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,1,power_law_1.2,0.05077760219573975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,1,power_law_1.2,0.05631999969482422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,1,power_law_1.2,0.06435199975967407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,1,power_law_1.2,0.07571840286254883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,1,power_law_1.2,0.10584319829940796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,1,power_law_1.2,0.11119359731674194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,1,power_law_1.2,0.11596800088882446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,1,power_law_1.2,0.11843199729919433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,1,power_law_1.2,0.12170239686965942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,1,power_law_1.2,0.12264959812164307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,1,power_law_1.2,0.12942080497741698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,1,power_law_1.2,0.13099520206451415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,1,power_law_1.2,0.13524479866027833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,1,power_law_1.2,0.14357759952545165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,1,power_law_1.2,0.1488767981529236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,1,power_law_1.2,0.15904639959335326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,1,power_law_1.2,0.18673280477523804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,1,power_law_1.2,0.20568320751190186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,1,power_law_1.2,0.25578880310058594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,1,power_law_1.2,0.3037440061569214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,1,power_law_1.2,0.4102784156799316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,1,power_law_1.2,0.5108416080474854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,1,power_law_1.2,0.7116799831390381
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,1,power_law_1.2,0.9075455665588379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,1,power_law_1.2,1.3025535583496093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,1,power_law_1.2,1.6923967361450196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,1,power_law_1.2,2.085094451904297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,1,power_law_1.2,3.253286361694336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,1,power_law_1.2,6.315244674682617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,8,power_law_1.01,0.10010240077972413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,8,power_law_1.01,0.12925440073013306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,8,power_law_1.01,0.0834879994392395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,8,power_law_1.01,0.09962239861488342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,8,power_law_1.01,0.12054400444030762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,8,power_law_1.01,0.16021759510040284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,8,power_law_1.01,0.15674240589141847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,8,power_law_1.01,0.15947519540786742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,8,power_law_1.01,0.17419519424438476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,8,power_law_1.01,0.16449919939041138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,4,balanced,0.049685334165891014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,4,balanced,0.054858664671579994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,4,balanced,0.07235733171304067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,8,power_law_1.01,0.17294080257415773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,4,balanced,0.1002453366915385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,4,balanced,0.14962666233380637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,4,balanced,0.2448479930559794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,4,balanced,0.3468159834543864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,4,balanced,0.34725332260131836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,8,power_law_1.01,0.18134399652481079
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,4,balanced,0.3487039804458618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,4,balanced,0.3476800123850505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,4,balanced,0.3505386511484782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,4,balanced,0.35336001714070636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,8,power_law_1.01,0.1841472029685974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,4,balanced,0.3569493293762207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,4,balanced,0.35682133833567303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,4,balanced,0.3656533161799113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,4,balanced,0.36903464794158936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,4,balanced,0.3790293137232463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,8,power_law_1.01,0.18795520067214966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,4,balanced,0.39608534177144367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,4,balanced,0.4058506488800049
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,4,balanced,0.43160001436869305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,4,balanced,0.464629332224528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,8,power_law_1.01,0.1979583978652954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,4,balanced,0.5200426578521729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,4,balanced,0.572165330251058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,4,balanced,0.6715253194173177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,8,power_law_1.01,0.20037760734558105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,4,balanced,0.8237973054250082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,4,balanced,1.0540640354156494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,8,power_law_1.01,0.22064640522003173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,4,balanced,1.3295679887135823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,4,balanced,1.6406985918680828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,8,power_law_1.01,0.2592832088470459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,4,balanced,2.3734025955200195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,8,power_law_1.01,0.2671295881271362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,4,balanced,4.502335866292317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,8,power_law_1.01,0.3146239995956421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,8,power_law_1.01,0.37941761016845704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,8,power_law_1.01,0.4657279968261719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,8,power_law_1.01,0.4658815860748291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,8,power_law_1.01,0.6361728191375733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,8,power_law_1.01,0.8086336135864258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,8,power_law_1.01,1.0897279739379884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,8,power_law_1.01,1.3072896003723145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,8,power_law_1.01,1.6610559463500976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,8,power_law_1.01,2.5153663635253904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,32,balanced,0.060677334666252136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,32,balanced,0.06256533165772755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,32,balanced,0.06081599990526835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,32,balanced,0.07016000151634216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,32,balanced,0.08760000268618266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,32,balanced,0.11936533451080322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,32,balanced,0.14698132872581482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,32,balanced,0.12706666191418967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,32,balanced,0.13260799646377563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,32,balanced,0.13636799653371176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,32,balanced,0.126309335231781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,32,balanced,0.13121066490809122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,32,balanced,0.1346879998842875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,32,balanced,0.1300159990787506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,32,balanced,0.13873599966367087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,32,balanced,0.13636799653371176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,32,balanced,0.14780799547831217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,32,balanced,0.15030399958292642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,32,balanced,0.16180266936620077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,32,balanced,0.17535465955734253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,32,balanced,0.18970666329065958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,32,balanced,0.2222399910291036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,32,balanced,0.2515679995218913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,8,power_law_1.01,5.151187133789063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,32,balanced,0.3150826692581177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,32,balanced,0.4110613266626994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,32,balanced,0.5308106740315756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,32,balanced,0.6850079695383707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,32,balanced,0.8641119798024496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,32,balanced,1.3075253168741863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,32,balanced,2.493920008341471
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,2,power_law_1.01,0.055212801694869994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,2,power_law_1.01,0.06785920262336731
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,4,2,power_law_1.01,0.05503360033035278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,4,2,power_law_1.01,0.05553280115127564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,2,power_law_1.01,0.06496000289916992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,4,2,power_law_1.01,0.06656000018119812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,2,power_law_1.01,0.08335360288619995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,4,2,power_law_1.01,0.07939839959144593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,4,2,power_law_1.01,0.07955840229988098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,2,power_law_1.01,0.10479999780654907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,4,2,power_law_1.01,0.08391680121421814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,4,2,power_law_1.01,0.09074559807777405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,2,power_law_1.01,0.14771840572357178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,4,2,power_law_1.01,0.09120000004768372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,4,2,power_law_1.01,0.0886784017086029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,2,power_law_1.01,0.160806405544281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,4,2,power_law_1.01,0.0921280026435852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,2,power_law_1.01,0.1672320008277893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,4,2,power_law_1.01,0.08930559754371643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,4,2,power_law_1.01,0.0938368022441864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,2,power_law_1.01,0.16640640497207643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,4,2,power_law_1.01,0.09408000111579895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,4,2,power_law_1.01,0.09857280254364013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,2,power_law_1.01,0.170905601978302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,4,2,power_law_1.01,0.11479040384292602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,4,2,power_law_1.01,0.11695359945297241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,2,power_law_1.01,0.17265280485153198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,4,2,power_law_1.01,0.1221119999885559
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,4,2,power_law_1.01,0.14948480129241942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,2,power_law_1.01,0.18120319843292237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,4,2,power_law_1.01,0.15553280115127563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,4,2,power_law_1.01,0.18076800107955932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,2,power_law_1.01,0.18134399652481079
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,4,2,power_law_1.01,0.23384320735931396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,4,2,power_law_1.01,0.32424321174621584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,2,power_law_1.01,0.18881280422210694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,4,2,power_law_1.01,0.359334397315979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,2,power_law_1.01,0.20084478855133056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,4,2,power_law_1.01,0.4464320182800293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,4,2,power_law_1.01,0.6529727935791015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,2,power_law_1.01,0.20502400398254395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,4,2,power_law_1.01,0.8703871726989746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,2,power_law_1.01,0.22204160690307617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,4,2,power_law_1.01,1.333561611175537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,2,power_law_1.01,0.2496959924697876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,4,2,power_law_1.01,1.692403221130371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,2,power_law_1.01,0.26549758911132815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,2,power_law_1.01,0.3171135902404785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,4,2,power_law_1.01,2.4140031814575194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,2,power_law_1.01,0.36780800819396975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,2,power_law_1.01,0.4724736213684082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,4,2,power_law_1.01,4.486796951293945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,2,power_law_1.01,0.5935616016387939
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,2,power_law_1.01,0.7769599914550781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,2,power_law_1.01,1.0156031608581544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,2,power_law_1.01,1.290982437133789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,2,power_law_1.01,1.6698432922363282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,2,power_law_1.01,1.9062143325805665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,2,power_law_1.01,3.1596031188964844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,16,2,power_law_1.2,0.03718400001525879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,16,2,power_law_1.2,0.039001598954200745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,16,2,power_law_1.2,0.04376319944858551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,2,power_law_1.01,5.493376159667969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,16,2,power_law_1.2,0.05289599895477295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,16,2,power_law_1.2,0.05319679975509643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,16,2,power_law_1.2,0.0562175989151001
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,16,2,power_law_1.2,0.05939840078353882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,16,2,power_law_1.2,0.05978879928588867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,16,2,power_law_1.2,0.05971199870109558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,16,2,power_law_1.2,0.060678398609161376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,16,2,power_law_1.2,0.06223999857902527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,16,2,power_law_1.2,0.06277120113372803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,16,2,power_law_1.2,0.061990398168563846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,16,2,power_law_1.2,0.06630399823188782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,16,2,power_law_1.2,0.0741375982761383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,16,2,power_law_1.2,0.07763839960098266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,16,2,power_law_1.2,0.08037760257720947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,16,2,power_law_1.2,0.09874560236930847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,16,2,power_law_1.2,0.10712959766387939
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,16,2,power_law_1.2,0.13767679929733276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,16,2,power_law_1.2,0.16832000017166138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,16,2,power_law_1.2,0.2139904022216797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,16,2,power_law_1.2,0.22967679500579835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,16,2,power_law_1.2,0.32568960189819335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,16,2,power_law_1.2,0.41328001022338867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,16,2,power_law_1.2,0.6046847820281982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,16,2,power_law_1.2,0.789305591583252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,16,2,power_law_1.2,0.9287808418273926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,16,2,power_law_1.2,1.5491456031799316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,16,2,power_law_1.2,2.640300750732422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,64,power_law_1.2,0.03704319894313812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,64,power_law_1.2,0.036883199214935304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,64,power_law_1.2,0.03371520042419433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,64,power_law_1.2,0.03374719917774201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,64,power_law_1.2,0.03400320112705231
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,64,power_law_1.2,0.034355199337005614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,64,power_law_1.2,0.03498240113258362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,64,power_law_1.2,0.0348800003528595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,64,power_law_1.2,0.03511039912700653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,64,power_law_1.2,0.03537279963493347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,64,power_law_1.2,0.03678080141544342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,64,power_law_1.2,0.037811198830604555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,64,power_law_1.2,0.03780480027198792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,64,power_law_1.2,0.038822400569915774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,64,power_law_1.2,0.04255360066890716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,64,power_law_1.2,0.04260480105876922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,64,power_law_1.2,0.046758401393890384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,64,power_law_1.2,0.05397120118141174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,64,power_law_1.2,0.05479679703712463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,64,power_law_1.2,0.06264960169792175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,64,power_law_1.2,0.07166720032691956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,64,power_law_1.2,0.09827839732170104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,64,power_law_1.2,0.11105279922485352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,8,power_law_1.01,0.059622400999069215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,64,power_law_1.2,0.18294399976730347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,64,power_law_1.2,0.22709119319915771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,64,power_law_1.2,0.41765761375427246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,8,power_law_1.01,0.06228479743003845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,64,power_law_1.2,0.4478400230407715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,8,power_law_1.01,0.055283200740814206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,64,power_law_1.2,0.5328959941864013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,8,power_law_1.01,0.05799679756164551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,64,power_law_1.2,0.8517696380615234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,8,power_law_1.01,0.059724801778793336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,64,power_law_1.2,1.8190143585205079
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,8,power_law_1.01,0.0670527994632721
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,8,power_law_1.01,0.06779519915580749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,8,power_law_1.01,0.06959360241889953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,8,power_law_1.01,0.07159680128097534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,8,power_law_1.01,0.07166720032691956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,8,power_law_1.01,0.0726527988910675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,8,power_law_1.01,0.0764415979385376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,8,power_law_1.01,0.07678080201148987
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,8,power_law_1.01,0.07984640002250672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,4,2,power_law_1.01,0.03776000142097473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,4,2,power_law_1.01,0.040870401263237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,8,power_law_1.01,0.08562560081481933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,4,2,power_law_1.01,0.038540801405906676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,4,2,power_law_1.01,0.042387199401855466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,4,2,power_law_1.01,0.04407039880752563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,4,2,power_law_1.01,0.04319359958171844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,4,2,power_law_1.01,0.04552319943904877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,4,2,power_law_1.01,0.045158401131629944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,4,2,power_law_1.01,0.04938240051269531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,8,power_law_1.01,0.08666239976882935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,4,2,power_law_1.01,0.04906879961490631
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,4,2,power_law_1.01,0.052832001447677614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,4,2,power_law_1.01,0.05435519814491272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,4,2,power_law_1.01,0.055353599786758426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,4,2,power_law_1.01,0.06399359703063964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,4,2,power_law_1.01,0.07183359861373902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,4,2,power_law_1.01,0.07617279887199402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,8,power_law_1.01,0.09724159836769104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,4,2,power_law_1.01,0.07672960162162781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,4,2,power_law_1.01,0.08394240140914917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,4,2,power_law_1.01,0.10207359790802002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,4,2,power_law_1.01,0.11479040384292602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,4,2,power_law_1.01,0.1466815948486328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,4,2,power_law_1.01,0.17870080471038818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,8,power_law_1.01,0.10995839834213257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,4,2,power_law_1.01,0.22997119426727294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,4,2,power_law_1.01,0.3190975904464722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,4,2,power_law_1.01,0.3729599952697754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,4,2,power_law_1.01,0.6110527992248536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,8,power_law_1.01,0.12624640464782716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,4,2,power_law_1.01,0.8143424034118653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,4,2,power_law_1.01,0.9550271987915039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,8,power_law_1.01,0.15612159967422484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,4,2,power_law_1.01,1.4726207733154297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,8,power_law_1.01,0.1804927945137024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,4,2,power_law_1.01,2.9173120498657226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,8,power_law_1.01,0.2323199987411499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,8,power_law_1.01,0.26871678829193113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,8,power_law_1.01,0.36168320178985597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,8,power_law_1.01,0.4651519775390625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,8,power_law_1.01,0.6422336101531982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,8,power_law_1.01,0.8222720146179199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,8,power_law_1.01,1.0145600318908692
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,2,power_law_1.01,0.09973120093345642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,8,power_law_1.01,1.6334463119506837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,2,power_law_1.01,0.14734079837799072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,2,power_law_1.01,0.14024319648742675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,2,power_law_1.01,0.19911680221557618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,2,power_law_1.01,0.2858304023742676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,2,power_law_1.01,0.43715839385986327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,8,power_law_1.01,3.2550785064697267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,2,power_law_1.01,0.4836927890777588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,2,power_law_1.01,0.5030399799346924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,2,power_law_1.01,0.5145023822784424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,2,power_law_1.01,0.5145472049713135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,2,power_law_1.01,0.5347263813018799
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,2,power_law_1.01,0.5520768165588379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,2,power_law_1.01,0.5582335948944092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,2,power_law_1.01,0.5716288089752197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,2,power_law_1.01,0.5955904006958008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,2,power_law_1.01,0.5796095848083496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,2,power_law_1.01,0.6552576065063477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,2,power_law_1.01,0.7476736068725586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,2,power_law_1.01,0.7035967826843261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,2,power_law_1.01,0.8169599533081054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,2,power_law_1.01,1.0167231559753418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,2,power_law_1.01,1.2080960273742676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,2,power_law_1.01,1.2076479911804199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,2,power_law_1.2,0.074099200963974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,2,power_law_1.01,1.5021247863769531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,2,power_law_1.2,0.08065279722213745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,2,power_law_1.01,1.7868223190307617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,2,power_law_1.2,0.08755199909210205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,2,power_law_1.2,0.11537920236587525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,2,power_law_1.01,2.4058816909790037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,2,power_law_1.2,0.1454208016395569
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,2,power_law_1.2,0.2476288080215454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,2,power_law_1.01,3.2562110900878904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,2,power_law_1.2,0.26759679317474366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,2,power_law_1.2,0.27954559326171874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,2,power_law_1.01,3.836492919921875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,2,power_law_1.2,0.29363839626312255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,2,power_law_1.2,0.29322240352630613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,2,power_law_1.01,5.606246566772461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,2,power_law_1.2,0.30268158912658694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,2,power_law_1.2,0.3161855936050415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,2,power_law_1.2,0.31445119380950926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,2,power_law_1.2,0.32734720706939696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,2,power_law_1.2,0.344268798828125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,2,power_law_1.01,11.033036804199218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,2,power_law_1.2,0.35136640071868896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,2,power_law_1.2,0.363756799697876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,2,power_law_1.2,0.4096640110015869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,2,power_law_1.2,0.4436800003051758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,2,power_law_1.2,0.5135615825653076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,2,power_law_1.2,0.5421823978424072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,2,power_law_1.2,0.6967040061950683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,2,power_law_1.2,0.8640447616577148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,2,power_law_1.2,1.1089920043945312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,2,power_law_1.2,1.5370752334594726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,2,power_law_1.2,2.063750457763672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,2,power_law_1.2,2.6111423492431642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,16,power_law_1.2,0.10143359899520873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,2,power_law_1.2,3.0860992431640626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,16,power_law_1.2,0.0775551974773407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,8,power_law_1.01,0.04161919951438904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,8,power_law_1.01,0.038764798641204835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,2,power_law_1.2,4.762905502319336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,16,power_law_1.2,0.06694399714469909
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,8,power_law_1.01,0.03649280071258545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,8,power_law_1.01,0.038873600959777835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,8,power_law_1.01,0.042412799596786496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,8,power_law_1.01,0.048851200938224794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,16,power_law_1.2,0.07905279994010925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,8,power_law_1.01,0.04679679870605469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,8,power_law_1.01,0.04902400076389313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,8,power_law_1.01,0.04862079918384552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,16,power_law_1.2,0.08636800050735474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,8,power_law_1.01,0.05118079781532288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,8,power_law_1.01,0.05015680193901062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,8,power_law_1.01,0.05335680246353149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,2,power_law_1.2,8.844096374511718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,16,power_law_1.2,0.10868480205535888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,8,power_law_1.01,0.05364480018615723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,8,power_law_1.01,0.05817599892616272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,8,power_law_1.01,0.0631168007850647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,16,power_law_1.2,0.11248639822006226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,8,power_law_1.01,0.06513280272483826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,8,power_law_1.01,0.06951680183410644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,8,power_law_1.01,0.08122239708900451
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,16,power_law_1.2,0.11665920019149781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,8,power_law_1.01,0.09358720183372497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,8,power_law_1.01,0.11032320261001587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,8,power_law_1.01,0.118995201587677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,2,power_law_1.2,0.04604159891605377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,16,power_law_1.2,0.11887359619140625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,8,power_law_1.01,0.15502079725265502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,8,power_law_1.01,0.18750720024108886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,2,power_law_1.2,0.05146880149841308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,16,power_law_1.2,0.12099839448928833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,8,power_law_1.01,0.25116159915924074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,2,power_law_1.2,0.05445759892463684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,8,power_law_1.01,0.3014336109161377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,16,power_law_1.2,0.12161279916763305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,8,power_law_1.01,0.4556543827056885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,2,power_law_1.2,0.06219519972801209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,8,power_law_1.01,0.5252352237701416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,16,power_law_1.2,0.12540160417556762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,2,power_law_1.2,0.07742720246315002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,8,power_law_1.01,0.7433152198791504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,16,power_law_1.2,0.1276800036430359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,2,power_law_1.2,0.09544320106506347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,1,balanced,0.10161067048708598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,1,balanced,0.16434666514396667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,1,balanced,0.19285867611567178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,1,balanced,0.3097653388977051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,8,power_law_1.01,1.0774527549743653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,1,balanced,0.5516746838887533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,2,power_law_1.2,0.10877439975738526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,16,power_law_1.2,0.13159680366516113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,1,balanced,1.0294346809387207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,1,balanced,1.0356640021006267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,1,balanced,1.0360853672027588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,1,balanced,1.040778636932373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,1,balanced,1.042143980662028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,1,balanced,1.0513386726379395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,2,power_law_1.2,0.12520960569381714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,1,balanced,1.0622186660766602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,1,balanced,1.0627093315124512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,16,power_law_1.2,0.13930879831314086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,1,balanced,1.0757439931233723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,1,balanced,1.0892159938812256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,1,balanced,1.097920020421346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,8,power_law_1.01,2.0214656829833983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,2,power_law_1.2,0.1333184003829956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,1,balanced,1.1204480330149333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,1,balanced,1.156176010767619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,1,balanced,1.1979893048604329
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,16,power_law_1.2,0.14168959856033325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,1,balanced,1.280618667602539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,1,balanced,1.386250654856364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,2,power_law_1.2,0.13640320301055908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,1,balanced,1.5462667147318523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,1,balanced,1.7109227180480957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,1,balanced,2.3322879473368325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,16,power_law_1.2,0.15240319967269897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,2,power_law_1.2,0.1467136025428772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,1,balanced,2.599616050720215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,1,balanced,3.772186597188314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,2,power_law_1.2,0.15082240104675293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,16,power_law_1.2,0.17314560413360597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,1,balanced,4.678970654805501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,2,power_law_1.2,0.15872000455856322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,1,balanced,5.770442962646484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,16,power_law_1.2,0.19279999732971193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,2,power_law_1.2,0.1737215995788574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,1,balanced,8.44541867574056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,16,power_law_1.2,0.2419071912765503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,2,power_law_1.2,0.1883455991744995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,2,power_law_1.2,0.19757440090179443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,16,power_law_1.2,0.28511359691619875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,1,balanced,15.956651051839193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,2,power_law_1.2,0.20054399967193604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,16,power_law_1.2,0.3593663930892944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,2,power_law_1.2,0.22894721031188964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,128,power_law_1.01,0.07580800056457519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,2,power_law_1.2,0.2573823928833008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,16,power_law_1.2,0.4213888168334961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,2,power_law_1.2,0.32947840690612795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,128,power_law_1.01,0.09653760194778442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,16,power_law_1.2,0.5244287967681884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,2,power_law_1.2,0.38341119289398196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,128,power_law_1.01,0.06147199869155884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,2,power_law_1.2,0.49422078132629393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,16,power_law_1.2,0.7210495948791504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,128,power_law_1.01,0.06906239986419678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,2,power_law_1.2,0.6099711894989014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,128,power_law_1.01,0.06997119784355163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,16,power_law_1.2,1.0136768341064453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,2,power_law_1.2,0.8294976234436036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,128,power_law_1.01,0.07653120160102844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,2,power_law_1.2,1.0698687553405761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,128,power_law_1.01,0.07549440264701843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,16,power_law_1.2,1.2465984344482421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,2,power_law_1.01,0.05230720043182373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,2,power_law_1.2,1.568563175201416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,128,power_law_1.01,0.07731840014457703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,2,power_law_1.01,0.07322880029678344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,128,power_law_1.01,0.0786624014377594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,2,power_law_1.2,1.9405311584472655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,1,16,power_law_1.2,1.5560768127441407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,128,power_law_1.01,0.07820159792900086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,2,power_law_1.01,0.06983680129051209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,2,power_law_1.2,2.437599945068359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,128,power_law_1.01,0.07960960268974304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,2,power_law_1.01,0.09279999732971192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,128,power_law_1.01,0.08307200074195861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,2,power_law_1.01,0.12342400550842285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,2,power_law_1.2,4.036371231079102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,1,16,power_law_1.2,2.5405952453613283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,128,power_law_1.01,0.0868224024772644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,2,power_law_1.01,0.15331200361251832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,128,power_law_1.01,0.08588160276412964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,2,power_law_1.01,0.19114240407943725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,2,power_law_1.01,0.20120320320129395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,128,power_law_1.01,0.0926912009716034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,2,power_law_1.2,7.474668884277344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,2,power_law_1.01,0.21162879467010498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,128,power_law_1.01,0.09423999786376953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,2,power_law_1.01,0.2189120054244995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,128,power_law_1.01,0.10049920082092285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,2,power_law_1.01,0.22525439262390137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,1,16,power_law_1.2,5.500627136230468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,128,power_law_1.01,0.12168320417404174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,2,power_law_1.01,0.2287168025970459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,128,power_law_1.01,0.12252800464630127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,2,power_law_1.01,0.2360703945159912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,2,power_law_1.01,0.25394558906555176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,128,power_law_1.01,0.15653760433197023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,2,power_law_1.01,0.257094407081604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,128,power_law_1.01,0.1641088008880615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,2,power_law_1.01,0.2694591999053955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,2,power_law_1.01,0.28501119613647463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,128,power_law_1.01,0.21238400936126708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,2,power_law_1.01,0.32452480792999266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,16,power_law_1.2,0.040703999996185306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,128,power_law_1.01,0.2266688108444214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,16,power_law_1.2,0.03518719971179962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,2,power_law_1.01,0.3507456064224243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,16,power_law_1.2,0.033555200695991515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,16,power_law_1.2,0.032678401470184325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,2,power_law_1.01,0.41242241859436035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,16,power_law_1.2,0.033983999490737916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,128,power_law_1.01,0.28615679740905764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,16,power_law_1.2,0.03513599932193756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,1,balanced,0.0788213312625885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,1,balanced,0.08337066570917766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,1,balanced,0.09880000352859497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,1,balanced,0.1325759987036387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,16,power_law_1.2,0.03524479866027832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,1,balanced,0.19698667526245117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,2,power_law_1.01,0.4716800212860107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,1,balanced,0.3057120045026143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,1,balanced,0.30871466795603436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,1,balanced,0.3123093247413635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,1,balanced,0.31134400765101117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,1,balanced,0.3130506674448649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,16,power_law_1.2,0.03500159978866577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,1,balanced,0.3168853322664897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,1,balanced,0.3224533398946126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,1,balanced,0.32497066259384155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,1,balanced,0.3323520024617513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,1,balanced,0.34008534749348956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,1,balanced,0.34600532054901123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,16,power_law_1.2,0.03543039858341217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,1,balanced,0.3603626489639282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,1,balanced,0.39825065930684406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,128,power_law_1.01,0.351692795753479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,1,balanced,0.4285653432210286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,2,power_law_1.01,0.6103807926177979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,1,balanced,0.5045919815699259
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,16,power_law_1.2,0.03609600067138672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,1,balanced,0.5692693392435709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,1,balanced,0.7261173725128174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,1,balanced,0.8739626407623291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,16,power_law_1.2,0.03641600012779236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,1,balanced,1.2670133113861084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,1,balanced,1.552901268005371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,16,power_law_1.2,0.03751679956912994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,2,power_law_1.01,0.7166143894195557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,1,balanced,2.2640053431193032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,16,power_law_1.2,0.03813120126724243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,1,balanced,2.970399856567383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,16,power_law_1.2,0.03906559944152832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,128,power_law_1.01,0.5485951900482178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,1,balanced,3.723519961039225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,16,power_law_1.2,0.04437119960784912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,2,power_law_1.01,0.941983985900879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,16,power_law_1.2,0.04535680115222931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,1,balanced,5.623818715413411
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,16,power_law_1.2,0.04694400131702423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,16,power_law_1.2,0.052767997980117796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,2,power_law_1.01,1.1972991943359375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,16,power_law_1.2,0.054841601848602296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,128,power_law_1.01,0.5679232120513916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,16,power_law_1.2,0.06735360026359558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,1,balanced,10.895066579182943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,16,power_law_1.2,0.07528319954872131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,2,power_law_1.01,1.6582975387573242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,16,power_law_1.2,0.09260799884796142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,16,power_law_1.2,0.1123136043548584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,16,power_law_1.2,0.1499776005744934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,2,power_law_1.01,2.078291130065918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,1,128,power_law_1.01,0.6943999767303467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,16,power_law_1.2,0.19288320541381837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,16,power_law_1.2,0.32235519886016845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,16,power_law_1.2,0.37582080364227294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,2,power_law_1.01,2.4858240127563476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,16,power_law_1.2,0.4735104084014893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,1,128,power_law_1.01,1.080409622192383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,16,power_law_1.2,0.721446418762207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,2,power_law_1.01,3.7408191680908205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,16,power_law_1.2,1.8611391067504883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,2,balanced,0.05003199974695841
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,2,balanced,0.05699733396371206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,2,balanced,0.07446933289368947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,2,balanced,0.10394133130709331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,2,balanced,0.15424533685048422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,2,balanced,0.2539946635564168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,2,balanced,0.3535626729329427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,2,balanced,0.3544693390528361
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,2,balanced,0.36842668056488037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,2,balanced,0.3617440064748128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,2,balanced,0.36557865142822266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,2,balanced,0.3736853202184041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,2,balanced,0.37753601868947345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,2,balanced,0.38251201311747235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,2,balanced,0.39313066005706787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,2,power_law_1.01,7.377849578857422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,2,balanced,0.39605867862701416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,2,balanced,0.4106293519337972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,2,balanced,0.4347466627756755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,2,balanced,0.4600906769434611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,2,balanced,0.5087466637293497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,2,balanced,0.5559306542078654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,2,balanced,0.6597226858139038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,2,balanced,0.780623992284139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,2,balanced,0.9842399756113688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,2,balanced,1.241866668065389
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,2,balanced,1.633248011271159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,1,128,power_law_1.01,2.1862016677856446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,2,balanced,2.170010725657145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,2,balanced,2.71941343943278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,2,balanced,3.9842185974121094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,2,balanced,7.631306966145833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,4,2,power_law_1.01,0.05015680193901062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,4,2,power_law_1.01,0.0658240020275116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,4,2,power_law_1.01,0.0588096022605896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,4,2,power_law_1.01,0.07643520236015319
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,4,2,power_law_1.01,0.07752320170402527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,4,2,power_law_1.01,0.07877119779586791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,4,2,power_law_1.01,0.08004480004310607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,4,2,power_law_1.01,0.08107519745826722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,4,2,power_law_1.01,0.08298879861831665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,4,2,power_law_1.01,0.08369280099868774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,4,2,power_law_1.01,0.08565760254859925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,4,2,power_law_1.01,0.08595200181007386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,4,2,power_law_1.01,0.0875711977481842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,4,2,power_law_1.01,0.09715840220451355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,4,power_law_1.01,0.0339711993932724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,4,2,power_law_1.01,0.10589439868927002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,4,2,power_law_1.01,0.11514879465103149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,4,2,power_law_1.01,0.12117120027542114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,4,2,power_law_1.01,0.15091840028762818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,4,2,power_law_1.01,0.15372159481048583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,4,power_law_1.01,0.03418239951133728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,4,2,power_law_1.01,0.20855679512023925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,4,2,power_law_1.01,0.2262592077255249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,4,2,power_law_1.01,0.33308799266815187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,4,power_law_1.01,0.04136320054531097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,4,2,power_law_1.01,0.41622400283813477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,4,power_law_1.01,0.032742398977279666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,4,2,power_law_1.01,0.5755328178405762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,4,power_law_1.01,0.04336000084877014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,4,2,power_law_1.01,0.7763840198516846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,4,power_law_1.01,0.03825919926166534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,4,power_law_1.01,0.036620798707008365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,4,2,power_law_1.01,0.941977596282959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,4,power_law_1.01,0.04778240025043488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,4,2,power_law_1.01,1.2968832015991212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,4,power_law_1.01,0.03948799967765808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,4,power_law_1.01,0.049472001194953916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,4,2,power_law_1.01,1.6210176467895507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,8,balanced,0.0521066685517629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,8,balanced,0.05387733379999796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,4,power_law_1.01,0.06266239881515503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,8,balanced,0.05605333546797434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,4,power_law_1.01,0.052742397785186766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,8,balanced,0.06105599800745646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,8,balanced,0.08618133266766866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,8,balanced,0.12154666582743327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,8,balanced,0.15157866477966309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,8,balanced,0.15230933825174967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,8,balanced,0.15293866395950317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,8,balanced,0.15451733271280924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,4,power_law_1.01,0.06610559821128845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,8,balanced,0.15330132842063904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,8,balanced,0.15821333726247153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,8,balanced,0.15843199690183005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,8,balanced,0.16156267126401266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,4,2,power_law_1.01,2.7933311462402344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,8,balanced,0.16893333196640015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,4,power_law_1.01,0.052134400606155394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,8,balanced,0.16801067193349203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,8,balanced,0.17802667617797852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,4,power_law_1.01,0.06738560199737549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,8,balanced,0.18709333737691244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,8,balanced,0.19867199659347534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,8,balanced,0.2210986614227295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,8,balanced,0.24266666173934937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,8,balanced,0.28969599803288776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,8,balanced,0.3330666621526082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,4,power_law_1.01,0.06649600267410279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,4,power_law_1.01,0.055641597509384154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,8,balanced,0.41461865107218426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,8,balanced,0.5256266593933105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,8,balanced,0.6918293635050455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,4,power_law_1.01,0.06871039867401123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,8,balanced,0.8994239966074625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,4,power_law_1.01,0.05567359924316406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,8,balanced,1.108138640721639
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,4,2,power_law_1.01,5.126105499267578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,4,power_law_1.01,0.06959360241889953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,8,balanced,1.6504213015238445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,4,power_law_1.01,0.05544319748878479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,4,power_law_1.01,0.06940799951553345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,8,balanced,3.1869331995646157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,4,power_law_1.01,0.07190399765968322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,4,power_law_1.01,0.0574400007724762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,4,power_law_1.01,0.07626240253448487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,4,power_law_1.01,0.05812479853630066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,4,power_law_1.01,0.08083840012550354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,4,power_law_1.01,0.08407679796218873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,4,power_law_1.01,0.0602944016456604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,4,power_law_1.01,0.08867200016975403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,4,power_law_1.01,0.06234880089759827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,4,power_law_1.01,0.09708799719810486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,4,power_law_1.01,0.0663424015045166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,4,power_law_1.01,0.10982400178909302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,4,power_law_1.01,0.07104640007019043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,4,power_law_1.01,0.12599680423736573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,4,power_law_1.01,0.14483200311660765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,4,power_law_1.01,0.07002239823341369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,4,power_law_1.01,0.1771456003189087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,4,power_law_1.01,0.07799040079116822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,4,power_law_1.01,0.2151103973388672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,4,power_law_1.01,0.0834879994392395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,4,power_law_1.01,0.2724031925201416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,4,power_law_1.01,0.0953984022140503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,4,power_law_1.01,0.34632320404052735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,4,power_law_1.01,0.10623359680175781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,4,power_law_1.01,0.46440958976745605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,4,power_law_1.01,0.12923519611358641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,4,power_law_1.01,0.609062385559082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,4,power_law_1.01,0.1561728000640869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,4,power_law_1.01,0.20738561153411866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,2688,1856,6,128,1,4,power_law_1.01,0.7261375904083252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,64,balanced,0.03573333223660787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,64,balanced,0.031210665901501972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,64,balanced,0.0295413335164388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,64,balanced,0.03129599988460541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,4,power_law_1.01,0.24610559940338134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,64,balanced,0.031290667752424874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,64,balanced,0.033359999457995095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,64,balanced,0.03172266731659571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,64,balanced,0.031258667508761086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,64,balanced,0.03346666693687439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,64,balanced,0.0334346666932106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,64,balanced,0.03323200096686681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,64,balanced,0.03327466547489166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,64,balanced,0.03332266708215078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,64,balanced,0.035616000493367515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,64,balanced,0.03770666569471359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,64,balanced,0.03755733370780945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,64,balanced,0.03771200031042099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,64,balanced,0.04001066585381826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,2688,1856,6,128,1,4,power_law_1.01,1.1872832298278808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,64,balanced,0.040661332507928215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,64,balanced,0.04394133388996124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,64,balanced,0.04598933458328247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,64,balanced,0.05202666421731313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,64,balanced,0.05428266525268555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,64,balanced,0.08168533444404602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,64,balanced,0.10019733508427937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,4,power_law_1.01,0.33540480136871337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,2,power_law_1.01,0.056979197263717654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,64,balanced,0.13429866234461466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,64,balanced,0.16288000345230103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,2,64,balanced,0.19417067368825278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,2,64,balanced,0.2842453320821126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,2,64,balanced,0.5142666498819987
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,2,power_law_1.01,0.06663680076599121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,4,power_law_1.01,0.43065600395202636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,2,power_law_1.01,0.0717311978340149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,2688,1856,6,128,1,4,power_law_1.01,2.4089279174804688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,2,power_law_1.01,0.08595839738845826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,2,4,power_law_1.01,0.5062655925750732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,2,power_law_1.01,0.10288000106811523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,2,power_law_1.01,0.13274240493774414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,2,4,power_law_1.01,0.7618815898895264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,2,power_law_1.01,0.13719040155410767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,2,power_law_1.01,0.14366079568862916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,2,power_law_1.01,0.14867199659347535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,2,power_law_1.01,0.15238399505615235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,2,4,power_law_1.01,1.576255989074707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,2,power_law_1.01,0.15367679595947265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,2,power_law_1.01,0.16149120330810546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,2,power_law_1.01,0.1628543972969055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,2,power_law_1.01,0.17585279941558837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,2,power_law_1.01,0.1875839948654175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,2,power_law_1.01,0.195360004901886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,2,power_law_1.01,0.19249279499053956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,2,power_law_1.01,0.2168191909790039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,2,power_law_1.01,0.24897279739379882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,2,power_law_1.01,0.2900223970413208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,32,2,balanced,0.04606399933497111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,32,2,balanced,0.04980800052483877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,32,2,balanced,0.051914667089780174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,32,2,balanced,0.06267199913660686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,32,2,balanced,0.09055466453234355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,32,2,balanced,0.11010133226712544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,32,2,balanced,0.11241066455841064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,2,power_law_1.01,0.3486016035079956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,32,2,balanced,0.10656000177065532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,32,2,balanced,0.10677867134412129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,32,2,balanced,0.11123733719189961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,32,2,balanced,0.10867200295130412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,32,2,balanced,0.11185600360234578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,32,2,balanced,0.11347200473149617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,32,2,balanced,0.1156213382879893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,32,2,balanced,0.11909866333007812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,32,2,balanced,0.12483200430870056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,32,2,balanced,0.12788266936937967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,32,2,balanced,0.15065067013104758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,32,2,balanced,0.1632319986820221
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,2,power_law_1.01,0.4590464115142822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,32,2,balanced,0.22882133722305298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,32,2,balanced,0.24422933657964072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,32,2,balanced,0.36189866065979004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,32,2,balanced,0.3885706663131714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,32,2,balanced,0.5624959866205851
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,32,2,balanced,0.6837226549784342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,32,2,balanced,1.0288320382436116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,2,power_law_1.01,0.5803647994995117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,32,2,balanced,1.2541279792785645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,32,2,balanced,1.511957327524821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,1,power_law_1.01,0.09723520278930664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,2,power_law_1.01,0.7823423862457275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,32,2,balanced,2.331461270650228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,32,2,balanced,4.481482823689778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,1,power_law_1.01,0.18595199584960936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,2,power_law_1.01,1.0169919967651366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,1,64,balanced,0.028629332780838013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,1,64,balanced,0.027221334477265675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,1,64,balanced,0.026949333647886913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,1,64,balanced,0.03032533327738444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,1,64,balanced,0.028901333610216778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,1,64,balanced,0.03230933348337809
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,1,64,balanced,0.03251733382542928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,1,64,balanced,0.03242666771014532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,1,64,balanced,0.032287999987602234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,1,power_law_1.01,0.24267520904541015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,2,power_law_1.01,1.3757696151733398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,1,64,balanced,0.03285333265860876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,1,64,balanced,0.0342399999499321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,1,64,balanced,0.032602667808532715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,1,64,balanced,0.03435733417669932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,1,64,balanced,0.032842665910720825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,1,64,balanced,0.03587199995915095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,1,64,balanced,0.03642666588226954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,1,64,balanced,0.03642666588226954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,1,64,balanced,0.03709866603215536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,64,balanced,0.03648533423741659
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,64,balanced,0.03920533259709676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,64,balanced,0.04032533367474874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,64,balanced,0.04042666653792063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,64,balanced,0.04278400043646494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,64,balanced,0.05046933392683665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,64,balanced,0.0551093320051829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,1,power_law_1.01,0.29009280204772947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,64,balanced,0.0674239993095398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,64,balanced,0.08051200211048126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,2,power_law_1.01,1.7458431243896484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,64,balanced,0.09284800291061401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,64,balanced,0.12338667114575703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,64,balanced,0.1839253306388855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,1,power_law_1.01,0.607699203491211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,2,power_law_1.01,2.1903615951538087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,1,power_law_1.01,0.9114560127258301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,2,power_law_1.01,3.281856155395508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,1,power_law_1.01,1.1086976051330566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,1,power_law_1.01,1.4753600120544434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,1,power_law_1.01,1.556230354309082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,2,power_law_1.01,6.727327728271485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,1,power_law_1.01,1.6001344680786134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,16,power_law_1.2,0.0686079978942871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,1,power_law_1.01,1.6800512313842773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,16,power_law_1.2,0.09624959826469422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,1,power_law_1.01,1.7813119888305664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,16,power_law_1.2,0.05139840245246887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,16,power_law_1.2,0.0627839982509613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,1,power_law_1.01,1.812723159790039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,16,power_law_1.2,0.06273279786109924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,16,power_law_1.2,0.07355520129203796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,1,power_law_1.01,1.8583040237426758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,16,power_law_1.2,0.08458240032196045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,1,power_law_1.01,1.9566400527954102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,16,power_law_1.2,0.08759040236473084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,16,power_law_1.2,0.08952320218086243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,1,power_law_1.01,1.965190315246582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,16,power_law_1.2,0.09109119772911071
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,8,balanced,0.02994133283694585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,8,balanced,0.031727999448776245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,8,balanced,0.03134933362404505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,8,balanced,0.03346133232116699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,8,balanced,0.03819733361403147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,8,balanced,0.039520000418027244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,8,balanced,0.04146666576464971
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,16,power_law_1.2,0.09198079705238342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,8,balanced,0.03953066716591517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,8,balanced,0.041093334555625916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,8,balanced,0.0399893323580424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,8,balanced,0.04153066625197729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,8,balanced,0.04204800228277842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,8,balanced,0.04163199911514918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,8,balanced,0.043968002001444496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,1,power_law_1.01,2.0596288681030273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,8,balanced,0.05176533261934916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,8,balanced,0.05385066568851471
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,8,balanced,0.051498666405677795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,8,balanced,0.05782400071620941
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,8,balanced,0.07473066449165344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,8,balanced,0.09140266974767049
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,8,balanced,0.09308800101280212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,8,balanced,0.12520000338554382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,8,balanced,0.15320000052452087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,16,power_law_1.2,0.09217919707298279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,8,balanced,0.19369065761566162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,8,balanced,0.24347732464472452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,8,balanced,0.3428533474604289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,8,balanced,0.4348213275273641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,8,balanced,0.5324853261311849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,8,balanced,0.8211092948913574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,16,power_law_1.2,0.09637119770050048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,1,power_law_1.01,2.194592094421387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,8,balanced,1.56769593556722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,16,power_law_1.2,0.0948095977306366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,1,power_law_1.01,2.2928255081176756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,16,power_law_1.2,0.104038405418396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,16,power_law_1.2,0.10968320369720459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,1,power_law_1.01,2.5509759902954103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,16,power_law_1.2,0.11563520431518555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,16,power_law_1.2,0.12766079902648925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,1,power_law_1.01,2.7190399169921875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,16,power_law_1.2,0.13538559675216674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,16,power_law_1.2,0.157151997089386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,1,power_law_1.01,3.1939903259277345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,16,power_law_1.2,0.17486720085144042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,16,power_law_1.2,0.22700800895690917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,1,power_law_1.01,3.329759979248047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,16,power_law_1.2,0.2639039993286133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,16,power_law_1.2,0.3662911891937256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,8,balanced,0.04458666841189066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,8,balanced,0.04637333254019419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,8,balanced,0.04974933465321859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,8,balanced,0.07026133437951405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,1,power_law_1.01,4.068409729003906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,8,balanced,0.09573333462079366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,8,balanced,0.1474506656328837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,8,balanced,0.14683733383814493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,8,balanced,0.15024532874425253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,8,balanced,0.14958399534225464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,8,balanced,0.15177067120869955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,8,balanced,0.1511893371740977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,8,balanced,0.15364799896876016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,8,balanced,0.15413332978884378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,8,balanced,0.15473066767056784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,16,power_law_1.2,0.4797760009765625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,8,balanced,0.16085333625475565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,8,balanced,0.16172266999880472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,8,balanced,0.16506666938463846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,8,balanced,0.1743733286857605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,8,balanced,0.17923200130462646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,8,balanced,0.19953600565592447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,8,balanced,0.22497600317001343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,8,balanced,0.25982399781545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,8,balanced,0.2834346691767375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,8,balanced,0.34829334417978924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,8,balanced,0.4089226722717285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,8,balanced,0.5674879948298136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,8,balanced,0.6938880284627279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,16,power_law_1.2,0.6742464065551758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,8,balanced,0.8608907063802084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,1,power_law_1.01,4.392307281494141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,8,balanced,1.2309280236562092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,8,balanced,2.3884533246358237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,16,power_law_1.2,0.8396672248840332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,1,power_law_1.01,5.65551986694336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,16,power_law_1.2,1.037452793121338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,16,power_law_1.2,1.5359040260314942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,1,power_law_1.01,6.56421127319336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,16,power_law_1.2,3.324153518676758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,1,power_law_1.01,7.8014976501464846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,4,power_law_1.2,0.06024320125579834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,1,power_law_1.01,11.44318084716797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,4,power_law_1.2,0.06447359919548035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,4,power_law_1.2,0.06593919992446899
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,4,power_law_1.2,0.06436480283737182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,4,power_law_1.2,0.07367680072784424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,4,power_law_1.2,0.08479999899864196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,4,power_law_1.2,0.0899392008781433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,1,power_law_1.01,20.96167755126953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,8,power_law_1.01,0.04733439981937408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,4,power_law_1.2,0.10550400018692016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,8,power_law_1.01,0.051571202278137204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,4,power_law_1.2,0.10955519676208496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,4,power_law_1.2,0.11272319555282592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,8,power_law_1.01,0.0770367980003357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,4,power_law_1.2,0.1171455979347229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,8,power_law_1.01,0.05235840082168579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,4,power_law_1.2,0.1212224006652832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,8,power_law_1.01,0.05913599729537964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,4,power_law_1.2,0.12272000312805176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,8,power_law_1.01,0.06954240202903747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,4,power_law_1.2,0.12551039457321167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,8,power_law_1.01,0.07550079822540283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,4,power_law_1.2,0.1355839967727661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,8,power_law_1.01,0.08158079981803894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,4,power_law_1.2,0.1414720058441162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,8,power_law_1.01,0.08449280261993408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,4,power_law_1.2,0.15025919675827026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,8,power_law_1.01,0.08737279772758484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,4,power_law_1.2,0.16787199974060057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,8,power_law_1.01,0.08659200072288513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,4,balanced,0.039733332892258964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,4,balanced,0.03836799909671148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,4,balanced,0.042026668787002563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,4,balanced,0.0462719996770223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,4,balanced,0.05526933570702871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,4,balanced,0.058746665716171265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,4,balanced,0.05813866853713989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,4,balanced,0.05840533475081126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,4,balanced,0.058389330903689064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,4,balanced,0.05866133173306783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,4,balanced,0.060175999999046326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,4,balanced,0.060677334666252136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,4,balanced,0.06251733501752217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,4,balanced,0.0651093324025472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,4,power_law_1.2,0.1862720012664795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,4,balanced,0.0703413337469101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,4,balanced,0.07236800094445546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,4,balanced,0.07658666869004567
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,4,balanced,0.08707732955614726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,4,balanced,0.09498133261998494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,4,balanced,0.11121599872907002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,4,balanced,0.12499733765920003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,8,power_law_1.01,0.09408640265464782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,4,balanced,0.1670773426691691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,4,balanced,0.18916799624760947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,4,balanced,0.26712532838185626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,4,balanced,0.34493335088094074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,4,balanced,0.4950186808904012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,4,balanced,0.6390933195749918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,4,balanced,0.7915306886037191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,4,power_law_1.2,0.21799681186676026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,4,balanced,1.1761386394500732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,8,power_law_1.01,0.0942911982536316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,4,balanced,2.2789173126220703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,4,power_law_1.2,0.2578495979309082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,8,power_law_1.01,0.10399359464645386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,4,power_law_1.2,0.330022406578064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,8,power_law_1.01,0.10796799659729003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,4,power_law_1.2,0.40268797874450685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,8,power_law_1.01,0.1159168004989624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,8,power_law_1.01,0.12275199890136719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,4,power_law_1.2,0.5402048110961915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,8,power_law_1.01,0.13804160356521605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,4,power_law_1.2,0.7264512062072754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,8,power_law_1.01,0.1548799991607666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,4,power_law_1.2,1.034438419342041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,8,power_law_1.01,0.18303359746932985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,8,power_law_1.01,0.21120638847351075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,4,power_law_1.2,1.341094398498535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,32,4,balanced,0.04243200023969015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,32,4,balanced,0.034202667574087776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,32,4,balanced,0.03387733300526937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,32,4,balanced,0.03212266663710276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,32,4,balanced,0.03378133227427801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,32,4,balanced,0.035802667339642845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,8,power_law_1.01,0.2744064092636108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,32,4,balanced,0.033904001116752625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,32,4,balanced,0.035760000348091125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,32,4,balanced,0.03399466723203659
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,32,4,balanced,0.034117333590984344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,32,4,balanced,0.03598399957021078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,32,4,balanced,0.03779733429352442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,32,4,balanced,0.03588266670703888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,32,4,balanced,0.03676799933115641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,32,4,balanced,0.04182399809360504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,32,4,balanced,0.04266133407751719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,32,4,balanced,0.043712000052134194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,32,4,balanced,0.04820266862710317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,32,4,balanced,0.050704002380371094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,32,4,balanced,0.05923733115196228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,32,4,balanced,0.06604266663392384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,32,4,balanced,0.08685333530108134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,32,4,balanced,0.09842133522033691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,4,power_law_1.2,1.6698368072509766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,32,4,balanced,0.130730668703715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,32,4,balanced,0.14803733428319296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,32,4,balanced,0.20737600326538086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,32,4,balanced,0.24512000878651938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,32,4,balanced,0.3034293254216512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,8,power_law_1.01,0.35571200847625734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,32,4,balanced,0.4354986747105916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,32,4,balanced,0.7921653588612875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,8,power_law_1.01,0.4610752105712891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,4,power_law_1.2,2.781177520751953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,8,power_law_1.01,0.6137152194976807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,8,power_law_1.01,0.8480447769165039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,4,power_law_1.2,5.195161437988281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,8,power_law_1.01,1.1101696014404296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,8,power_law_1.01,1.299084758758545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,8,power_law_1.01,1.970515251159668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,32,8,power_law_1.2,0.04676479995250702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,32,8,power_law_1.2,0.05120639801025391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,8,power_law_1.01,3.944224166870117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,32,8,power_law_1.2,0.04845440089702606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,32,8,power_law_1.2,0.046982398629188536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,32,8,power_law_1.2,0.050937598943710326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,8,1,power_law_1.01,0.035673600435256955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,8,1,power_law_1.01,0.03885439932346344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,8,1,power_law_1.01,0.04116480052471161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,8,1,power_law_1.01,0.05683839917182922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,8,1,power_law_1.01,0.05854079723358154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,32,8,power_law_1.2,0.055193597078323366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,8,1,power_law_1.01,0.058412802219390866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,8,1,power_law_1.01,0.057766401767730714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,8,1,power_law_1.01,0.0595583975315094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,8,1,power_law_1.01,0.0602944016456604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,8,1,power_law_1.01,0.06108160018920898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,8,1,power_law_1.01,0.06225280165672302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,8,1,power_law_1.01,0.06496639847755432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,32,8,power_law_1.2,0.05619840025901794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,8,1,power_law_1.01,0.06711040139198303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,8,1,power_law_1.01,0.07085440158843995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,8,1,power_law_1.01,0.0803264021873474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,8,1,power_law_1.01,0.08309119939804077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,8,1,power_law_1.01,0.08680959939956664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,8,1,power_law_1.01,0.1054527997970581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,8,1,power_law_1.01,0.11308799982070923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,32,8,power_law_1.2,0.06234239935874939
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,8,1,power_law_1.01,0.14149760007858275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,8,1,power_law_1.01,0.16743680238723754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,8,1,power_law_1.01,0.22081921100616456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,8,1,power_law_1.01,0.2635200023651123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,8,1,power_law_1.01,0.36631040573120116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,32,8,power_law_1.2,0.06314880251884461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,8,1,power_law_1.01,0.4600192070007324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,8,1,power_law_1.01,0.6570047855377197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,8,1,power_law_1.01,0.8482111930847168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,32,8,power_law_1.2,0.06349440217018128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,8,1,power_law_1.01,1.0392576217651368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,8,1,power_law_1.01,1.6090879440307617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,32,8,power_law_1.2,0.0679040014743805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,2,balanced,0.050437331199645996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,2,balanced,0.09156800309816997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,8,1,power_law_1.01,3.145894432067871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,2,balanced,0.10243200262387593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,2,balanced,0.15362133582433066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,32,8,power_law_1.2,0.07003520131111145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,2,balanced,0.25784534215927124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,2,balanced,0.31116799513498944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,2,balanced,0.31460799773534137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,2,balanced,0.31350932518641156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,2,balanced,0.31675734122594196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,2,balanced,0.31723199288050336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,2,balanced,0.3190453251202901
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,2,balanced,0.32261866331100464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,2,balanced,0.3261013428370158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,2,balanced,0.3287946581840515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,2,balanced,0.3357119957605998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,2,balanced,0.3388799826304118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,2,balanced,0.3493386507034302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,32,8,power_law_1.2,0.07468159794807434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,2,balanced,0.3874986569086711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,2,balanced,0.38604267438252765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,2,balanced,0.4562720060348511
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,2,balanced,0.4609333276748657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,2,balanced,0.6303519805272421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,2,balanced,0.6269439856211344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,2,balanced,0.9050133228302002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,2,balanced,0.994767983754476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,32,8,power_law_1.2,0.07678719758987426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,2,balanced,1.366762638092041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,2,balanced,1.8006292978922527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,2,balanced,2.079360008239746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,32,8,power_law_1.2,0.08574720025062561
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,2,balanced,3.1919307708740234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,32,8,power_law_1.2,0.0916159987449646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,2,balanced,6.089541117350261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,32,8,power_law_1.2,0.10201599597930908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,32,8,power_law_1.2,0.12282880544662475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,32,8,power_law_1.2,0.12952959537506104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,32,8,power_law_1.2,0.16353280544281007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,32,8,power_law_1.2,0.19325439929962157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,8,power_law_1.2,0.14789119958877564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,32,8,power_law_1.2,0.2540096044540405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,8,power_law_1.2,0.29811840057373046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,32,8,power_law_1.2,0.32433280944824217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,8,power_law_1.2,0.10305279493331909
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,8,power_law_1.2,0.14895999431610107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,32,8,power_law_1.2,0.4539008140563965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,8,power_law_1.2,0.18543360233306885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,16,balanced,0.03163733333349228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,16,balanced,0.02977066735426585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,16,balanced,0.029530666768550873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,16,balanced,0.031557333966096245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,16,balanced,0.031386665999889374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,16,balanced,0.03365866591533025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,16,balanced,0.033610666791598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,16,balanced,0.033615998923778534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,16,balanced,0.03370666752258936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,16,balanced,0.03345066557327906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,32,8,power_law_1.2,0.5861311912536621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,16,balanced,0.03528533379236857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,16,balanced,0.035717333356539406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,8,power_law_1.2,0.22869760990142823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,16,balanced,0.03583466758330663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,16,balanced,0.035386666655540466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,16,balanced,0.03972266614437103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,16,balanced,0.03965866565704346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,16,balanced,0.039594667653242745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,16,balanced,0.04159466673930486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,16,balanced,0.04363200068473816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,16,balanced,0.0479360024134318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,16,balanced,0.051914667089780174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,16,balanced,0.062319998939832054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,16,balanced,0.07083733379840851
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,16,balanced,0.09858666857083638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,16,balanced,0.12577066818873087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,16,balanced,0.16925867398579916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,16,balanced,0.21057599782943726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,16,balanced,0.2516319950421651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,8,power_law_1.2,0.33671040534973146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,16,balanced,0.36478400230407715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,16,balanced,0.6971840063730875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,32,8,power_law_1.2,0.8752448081970214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,8,power_law_1.2,0.34727680683135986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,8,power_law_1.2,0.3769216060638428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,32,8,power_law_1.2,1.0208640098571777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,8,power_law_1.2,0.3599999904632568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,8,power_law_1.2,0.3703232049942017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,8,power_law_1.2,0.39755520820617674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,32,8,power_law_1.2,1.4266624450683594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,8,power_law_1.2,0.40474882125854494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,8,power_law_1.2,0.4065279960632324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,32,8,power_law_1.2,2.084441566467285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,8,power_law_1.2,0.4117119789123535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,8,power_law_1.2,0.4196608066558838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,8,power_law_1.2,0.48348798751831057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,8,power_law_1.2,0.5331776142120361
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,32,8,power_law_1.2,4.293894577026367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,8,power_law_1.2,0.5280384063720703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,32,8,power_law_1.2,0.029606398940086365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,32,8,power_law_1.2,0.029625600576400755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,32,8,power_law_1.2,0.028479999303817748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,32,8,power_law_1.2,0.02810879945755005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,32,8,power_law_1.2,0.028384000062942505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,8,power_law_1.2,0.6199552059173584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,32,8,power_law_1.2,0.029356798529624938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,32,8,power_law_1.2,0.029471999406814574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,32,8,power_law_1.2,0.02951039969921112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,32,8,power_law_1.2,0.030368000268936157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,32,8,power_law_1.2,0.030112001299858093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,32,8,power_law_1.2,0.030662399530410767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,8,power_law_1.2,0.6076032161712647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,32,8,power_law_1.2,0.03198719918727875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,32,8,power_law_1.2,0.03213439881801605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,32,8,power_law_1.2,0.031744000315666196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,32,8,power_law_1.2,0.03871999979019165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,32,8,power_law_1.2,0.03854719996452331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,32,8,power_law_1.2,0.03893119990825653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,8,power_law_1.2,0.7362688064575196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,32,8,power_law_1.2,0.04642559885978699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,32,8,power_law_1.2,0.05242879986763001
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,32,8,power_law_1.2,0.06216319799423218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,32,8,power_law_1.2,0.06165760159492493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,32,8,power_law_1.2,0.08358399868011475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,8,power_law_1.2,0.7873151779174805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,32,8,power_law_1.2,0.1014847993850708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,32,8,power_law_1.2,0.1443519949913025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,32,8,power_law_1.2,0.17651840448379516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,32,8,power_law_1.2,0.24678399562835693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,8,power_law_1.2,0.9893119812011719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,32,8,power_law_1.2,0.286572790145874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,32,8,power_law_1.2,0.3674432039260864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,32,8,power_law_1.2,0.5499328136444092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,8,power_law_1.2,1.2987968444824218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,32,8,power_law_1.2,0.9877504348754883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,8,power_law_1.2,1.6855680465698242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,8,power_law_1.2,2.2015167236328126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,32,power_law_1.2,0.03600000143051148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,8,power_law_1.2,2.515763282775879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,32,power_law_1.2,0.02998400032520294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,32,power_law_1.2,0.03036159873008728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,32,power_law_1.2,0.029862400889396668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,32,power_law_1.2,0.030777600407600404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,32,power_law_1.2,0.03190400004386902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,8,power_law_1.2,3.56561279296875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,32,power_law_1.2,0.03223679959774017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,32,power_law_1.2,0.03307519853115082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,32,power_law_1.2,0.03364480137825012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,32,power_law_1.2,0.0341376006603241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,32,power_law_1.2,0.03543039858341217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,32,power_law_1.2,0.03592959940433502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,32,power_law_1.2,0.03759360015392303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,32,power_law_1.2,0.03806720077991486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,8,power_law_1.2,8.316595458984375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,32,power_law_1.2,0.04213759899139404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,32,power_law_1.2,0.04413439929485321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,32,power_law_1.2,0.04557439982891083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,16,power_law_1.01,0.1518272042274475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,32,power_law_1.2,0.05000320076942444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,32,power_law_1.2,0.05731199979782105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,32,power_law_1.2,0.06788480281829834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,16,power_law_1.01,0.24206080436706542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,32,power_law_1.2,0.08198400139808655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,32,power_law_1.2,0.11818239688873292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,16,power_law_1.01,0.34445440769195557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,32,power_law_1.2,0.14855040311813356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,32,power_law_1.2,0.21569280624389647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,16,power_law_1.01,0.10808320045471191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,32,power_law_1.2,0.2752831935882568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,32,power_law_1.2,0.3710207939147949
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,16,power_law_1.01,0.1453376054763794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,32,power_law_1.2,0.587500810623169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,16,power_law_1.01,0.19267200231552123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,32,power_law_1.2,0.6918591976165771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,16,power_law_1.01,0.21314558982849122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,32,power_law_1.2,1.1750528335571289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,16,power_law_1.01,0.2523200035095215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,16,power_law_1.01,0.2563584089279175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,16,power_law_1.01,0.273036789894104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,32,power_law_1.2,2.7345855712890623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,16,power_law_1.01,0.27453439235687255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,16,power_law_1.01,0.2853760004043579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,16,power_law_1.01,0.3001215934753418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,16,power_law_1.01,0.298688006401062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,16,power_law_1.01,0.3197376012802124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,16,power_law_1.01,0.3166208028793335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,16,power_law_1.01,0.32189440727233887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,16,power_law_1.01,0.35434880256652834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,16,power_law_1.01,0.3784127950668335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,4,power_law_1.01,0.03309440016746521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,4,power_law_1.01,0.031430399417877196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,16,power_law_1.01,0.4569536209106445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,4,power_law_1.01,0.03289600014686585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,4,power_law_1.01,0.03701759874820709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,16,power_law_1.01,0.4417280197143555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,4,power_law_1.01,0.03875199854373932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,4,power_law_1.01,0.04398080110549927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,4,power_law_1.01,0.044947201013565065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,16,power_law_1.01,0.5345471858978271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,4,power_law_1.01,0.04609920084476471
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,4,power_law_1.01,0.0472896009683609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,16,power_law_1.01,0.6105088233947754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,4,power_law_1.01,0.04746240079402923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,4,power_law_1.01,0.05214080214500427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,4,power_law_1.01,0.05425919890403748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,16,power_law_1.01,0.8221504211425781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,4,power_law_1.01,0.05642240047454834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,4,power_law_1.01,0.056364798545837404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,4,power_law_1.01,0.06391040086746216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,16,power_law_1.01,0.8502528190612793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,4,power_law_1.01,0.0652351975440979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,4,power_law_1.01,0.06873599886894226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,4,power_law_1.01,0.08107519745826722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,16,power_law_1.01,1.083129596710205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,4,power_law_1.01,0.09596800208091735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,4,power_law_1.01,0.11763839721679688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,4,power_law_1.01,0.13800959587097167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,16,power_law_1.01,1.2599424362182616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,4,power_law_1.01,0.19215359687805175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,4,power_law_1.01,0.22880001068115235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,4,power_law_1.01,0.32755839824676514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,16,power_law_1.01,1.636390495300293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,4,power_law_1.01,0.4111487865447998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,4,power_law_1.01,0.6390207767486572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,4,power_law_1.01,0.8104191780090332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,16,power_law_1.01,2.2511104583740233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,4,power_law_1.01,0.9690560340881348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,4,power_law_1.01,1.5402303695678712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,16,power_law_1.01,4.27757453918457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,4,power_law_1.01,3.181715202331543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,4,balanced,0.04827199876308441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,4,balanced,0.050101334849993386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,4,balanced,0.07211733361085255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,4,balanced,0.10016000270843506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,4,balanced,0.14865066607793173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,4,balanced,0.24730134010314941
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,4,balanced,0.24564266204833984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,4,balanced,0.25072532892227173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,4,balanced,0.2508053382237752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,4,balanced,0.2516266703605652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,4,balanced,0.2550933361053467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,4,balanced,0.25961599747339886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,4,balanced,0.26154132684071857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,4,balanced,0.2608960072199504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,4,balanced,0.2738719979921977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,4,balanced,0.2776426672935486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,4,balanced,0.2825973431269328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,4,balanced,0.29867732524871826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,4,balanced,0.31497599681218463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,4,balanced,0.34542401631673175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,4,balanced,0.3900373379389445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,4,balanced,0.45132267475128174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,4,balanced,0.5180746714274088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,16,power_law_1.01,0.09925119876861573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,4,balanced,0.663152019182841
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,4,balanced,0.7798133691151937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,4,balanced,1.1029226779937744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,4,balanced,1.3665013313293457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,4,balanced,1.7094772656758626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,16,power_law_1.01,0.14603519439697266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,4,balanced,2.5492053031921387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,16,power_law_1.01,0.1991871953010559
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,4,balanced,4.953914642333984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,16,power_law_1.01,0.07902079820632935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,16,power_law_1.01,0.09685760140419006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,16,power_law_1.01,0.12600959539413453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,16,power_law_1.01,0.12981120347976685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,16,power_law_1.01,0.15667200088500977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,16,power_law_1.01,0.1580415964126587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,16,power_law_1.01,0.16157439947128296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,16,power_law_1.01,0.1586624026298523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,16,power_law_1.01,0.1690816044807434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,4,balanced,0.05362133185068766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,4,balanced,0.07868800063927968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,4,balanced,0.08940266569455464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,4,balanced,0.12913599610328674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,4,balanced,0.19484800100326538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,4,balanced,0.254533330599467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,4,balanced,0.2551413377126058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,4,balanced,0.25789332389831543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,4,balanced,0.2600853244463603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,4,balanced,0.2612053354581197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,4,balanced,0.2619360089302063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,4,balanced,0.2638346751530965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,4,balanced,0.26713067293167114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,16,power_law_1.01,0.1798784017562866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,4,balanced,0.26816000541051227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,4,balanced,0.2781066695849101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,4,balanced,0.2813013394673665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,4,balanced,0.2897333304087321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,4,balanced,0.3470453421274821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,4,balanced,0.32898666461308795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,4,balanced,0.4245599905649821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,4,balanced,0.4169013500213623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,4,balanced,0.5718933343887329
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,4,balanced,0.5696693261464437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,4,balanced,0.7766666412353516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,16,power_law_1.01,0.17774720191955568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,4,balanced,0.879034678141276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,4,balanced,1.3131840229034424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,4,balanced,1.5223093032836914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,16,power_law_1.01,0.18729599714279174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,4,4,balanced,1.9199412663777669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,4,4,balanced,2.90067195892334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,16,power_law_1.01,0.19264639616012574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,4,4,balanced,5.566682815551758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,16,power_law_1.01,0.19864319562911986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,16,power_law_1.01,0.21466879844665526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,16,power_law_1.01,0.23619840145111085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,16,power_law_1.01,0.268940806388855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,16,power_law_1.01,0.2657984018325806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,16,power_law_1.01,0.32816638946533205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,16,power_law_1.01,0.36868479251861574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,16,power_law_1.01,0.4932544231414795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,8,8,balanced,0.05005866785844167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,8,8,balanced,0.05035733183224996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,8,8,balanced,0.05232533315817515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,8,8,balanced,0.053861334919929504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,8,8,balanced,0.05881066620349884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,16,power_law_1.01,0.5617728233337402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,8,8,balanced,0.0643093337615331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,8,8,balanced,0.06487999856472015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,8,8,balanced,0.0642986645301183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,8,8,balanced,0.06676266590754192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,8,8,balanced,0.06513600051403046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,8,8,balanced,0.06678399940331776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,8,8,balanced,0.0683840016523997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,8,8,balanced,0.06880000233650208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,8,8,balanced,0.07008533179759979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,8,8,balanced,0.07486933469772339
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,8,8,balanced,0.07787199815114339
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,8,8,balanced,0.08086933195590973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,8,8,balanced,0.08916266759236653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,8,8,balanced,0.09914666414260864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,8,8,balanced,0.13524799545605978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,8,8,balanced,0.1342080036799113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,8,8,balanced,0.17773866653442383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,8,8,balanced,0.20071999231974283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,8,8,balanced,0.27900266647338867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,8,8,balanced,0.33450667063395184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,8,8,balanced,0.46427734692891437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,16,power_law_1.01,0.7349567890167237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,8,8,balanced,0.5952320098876953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,8,8,balanced,0.7115519841512045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,8,8,balanced,1.1031893094380696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,8,8,balanced,2.1399307250976562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,16,power_law_1.01,0.9409983634948731
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,16,power_law_1.01,1.1064191818237306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,16,power_law_1.01,1.665407943725586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,16,power_law_1.01,3.041702461242676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,2,8,power_law_1.2,0.027590399980545043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,2,8,power_law_1.2,0.024102400243282317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,2,8,power_law_1.2,0.023705600202083586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,2,8,power_law_1.2,0.02439039945602417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,2,8,power_law_1.2,0.025484800338745117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,2,8,power_law_1.2,0.026924800872802735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,2,8,power_law_1.2,0.027884799242019653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,2,8,power_law_1.2,0.030700799822807313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,2,8,power_law_1.2,0.031116798520088196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,2,8,power_law_1.2,0.03243519961833954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,2,8,power_law_1.2,0.03617919981479645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,2,8,power_law_1.2,0.03866240084171295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,2,8,power_law_1.2,0.04042240083217621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,2,8,power_law_1.2,0.04119040071964264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,2,8,power_law_1.2,0.0455808013677597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,2,8,power_law_1.2,0.0456063985824585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,2,8,power_law_1.2,0.0507968008518219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,1,balanced,0.056143999099731445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,1,balanced,0.08945066730181377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,1,balanced,0.10549867153167725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,1,balanced,0.15648000439008078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,1,balanced,0.25729066133499146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,1,balanced,0.4526453415552775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,2,8,power_law_1.2,0.057055997848510745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,1,balanced,0.45875732103983563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,1,balanced,0.46433067321777344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,1,balanced,0.4695039987564087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,1,balanced,0.4703199863433838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,1,balanced,0.4739733139673869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,8,power_law_1.2,0.06653439998626709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,1,balanced,0.48392001787821454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,1,balanced,0.48775466283162433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,1,balanced,0.4925813277562459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,1,balanced,0.5089493195215861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,1,balanced,0.5151093403498331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,8,power_law_1.2,0.0762943983078003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,1,balanced,0.5362613201141357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,1,balanced,0.5754613478978475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,1,balanced,0.607093334197998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,1,balanced,0.6903519630432129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,8,power_law_1.2,0.07731840014457703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,1,balanced,0.7682879765828451
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,1,balanced,0.9777493476867676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,1,balanced,1.1020159721374512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,8,power_law_1.2,0.09238399863243103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,1,balanced,1.5398613611857097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,1,balanced,1.8056373596191406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,8,power_law_1.2,0.12378879785537719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,1,balanced,2.6141173044840493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,8,power_law_1.2,0.15674240589141847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,1,balanced,3.2893333435058594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,8,power_law_1.2,0.20584321022033691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,1,balanced,4.107845306396484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,8,power_law_1.2,0.2933504104614258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,1,balanced,6.16653315226237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,8,power_law_1.2,0.33112320899963377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,8,power_law_1.2,0.4414720058441162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,1,balanced,11.878592173258463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,8,power_law_1.2,0.6868607997894287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,8,power_law_1.2,1.5507007598876954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,1,power_law_1.01,0.03282560110092163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,1,power_law_1.01,0.036447998881340024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,1,power_law_1.01,0.04237439930438995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,1,power_law_1.01,0.05172479748725891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,1,power_law_1.01,0.061887997388839724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,1,power_law_1.01,0.08766720294952393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,1,power_law_1.01,0.09128320217132568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,8,power_law_1.2,0.05809280276298523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,1,power_law_1.01,0.0936959981918335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,1,power_law_1.01,0.09681280255317688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,8,power_law_1.2,0.053446400165557864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,1,power_law_1.01,0.09815679788589478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,8,power_law_1.2,0.051712000370025636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,1,power_law_1.01,0.10089600086212158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,8,power_law_1.2,0.056524801254272464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,1,power_law_1.01,0.10369280576705933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,8,power_law_1.2,0.06005120277404785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,1,power_law_1.01,0.10649600028991699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,8,power_law_1.2,0.07027199864387512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,1,power_law_1.01,0.12420480251312256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,1,power_law_1.01,0.1359488010406494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,8,power_law_1.2,0.07449600100517273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,1,power_law_1.01,0.1400704026222229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,8,power_law_1.2,0.07687680125236511
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,1,power_law_1.01,0.15127040147781373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,8,power_law_1.2,0.0788096010684967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,1,power_law_1.01,0.1751871943473816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,8,power_law_1.2,0.07763199806213379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,1,power_law_1.01,0.18439040184020997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,8,power_law_1.2,0.08091520071029663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,1,power_law_1.01,0.23605759143829347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,8,power_law_1.2,0.08229759931564332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,1,power_law_1.01,0.2826047897338867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,8,power_law_1.2,0.08417919874191285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,1,power_law_1.01,0.37867519855499265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,8,power_law_1.2,0.08707839846611024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,1,power_law_1.01,0.5042880058288575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,8,power_law_1.2,0.09395840167999267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,1,power_law_1.01,0.7175424098968506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,8,power_law_1.2,0.09818879961967468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,1,power_law_1.01,0.8181568145751953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,8,power_law_1.2,0.10176639556884766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,8,power_law_1.2,0.11689599752426147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,1,power_law_1.01,1.1758272171020507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,8,power_law_1.2,0.13328640460968016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,8,power_law_1.2,0.1678272008895874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,1,power_law_1.01,1.4845312118530274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,8,power_law_1.2,0.1790336012840271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,8,power_law_1.2,0.23425281047821045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,1,power_law_1.01,1.8428031921386718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,8,power_law_1.2,0.2943936109542847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,8,power_law_1.2,0.4351679801940918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,1,power_law_1.01,2.9105663299560547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,8,power_law_1.2,0.4580671787261963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,8,power_law_1.2,0.713209581375122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,8,power_law_1.2,0.9705344200134277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,1,power_law_1.01,5.6094718933105465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,8,power_law_1.2,1.178700828552246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,8,power_law_1.2,1.9051904678344727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,8,power_law_1.2,3.846303939819336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,4,balanced,0.06983466446399689
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,4,balanced,0.10830400387446086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,4,balanced,0.12801600495974222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,4,balanced,0.20231467485427856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,4,balanced,0.33508801460266113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,4,balanced,0.44887999693552655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,4,balanced,0.4485333363215129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,4,balanced,0.4521600008010864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,4,balanced,0.4525013367335002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,4,balanced,0.45424532890319824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,4,balanced,0.4578080177307129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,4,balanced,0.4606879949569702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,4,balanced,0.46616001923878986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,4,balanced,0.46834667523701984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,4,balanced,0.47630401452382404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,4,balanced,0.48124265670776367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,4,balanced,0.4925866524378459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,4,balanced,0.5144160191218058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,4,balanced,0.533898671468099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,4,balanced,0.6806506315867106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,4,balanced,0.6198293368021647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,4,balanced,0.9561440149943033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,4,balanced,0.7795893351236979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,4,balanced,1.1691946983337402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,4,balanced,1.1513013044993083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,4,balanced,1.7290347417195637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,4,balanced,1.8833120663960774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,2,4,balanced,2.3909759521484375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,2,4,balanced,3.421152114868164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,8,balanced,0.0354720006386439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,8,balanced,0.03576533248027166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,8,balanced,0.03570666660865148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,8,balanced,0.03571200122435888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,8,balanced,0.04018666595220566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,8,balanced,0.03956266740957896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,8,balanced,0.04151466737190882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,8,balanced,0.04159466673930486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,8,balanced,0.0396373321612676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,8,balanced,0.04125866790612539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,8,balanced,0.041290665666262306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,8,balanced,0.04357333481311798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,8,balanced,0.04208533465862274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,8,balanced,0.04331199824810028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,8,balanced,0.04779199759165446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,8,balanced,0.04770133395989736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,8,balanced,0.05000533163547516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,8,balanced,0.051781331499417625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,2,4,balanced,6.866170883178711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,8,balanced,0.05805333455403646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,8,balanced,0.06195199986298879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,8,balanced,0.06644266843795776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,8,balanced,0.08249600231647491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,8,balanced,0.09546132882436116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,8,balanced,0.1273973286151886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,8,balanced,0.16196800271670023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,8,balanced,0.2178773283958435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,8,balanced,0.2671999931335449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,8,balanced,0.32630934317906696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,8,balanced,0.4949920177459717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,8,balanced,0.9472693602244059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,1,balanced,0.037861332297325134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,1,balanced,0.056346664826075234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,1,balanced,0.06658133367697398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,1,balanced,0.10355200370152791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,1,balanced,0.16780799627304077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,1,balanced,0.21501866976420084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,1,balanced,0.21767467260360718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,1,balanced,0.21868799130121866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,1,balanced,0.21702400843302408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,1,balanced,0.21917333205540976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,1,balanced,0.2213653326034546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,1,balanced,0.22206399838129678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,1,balanced,0.22336000204086304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,1,balanced,0.22633065780003866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,1,balanced,0.23180800676345825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,1,balanced,0.23518933852513632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,1,balanced,0.2391306757926941
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,1,balanced,0.28059200445810956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,1,balanced,0.2627893288930257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,1,balanced,0.35101866722106934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,1,balanced,0.30567467212677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,1,balanced,0.41228799025217694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,1,balanced,0.4150293270746867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,1,balanced,0.5963626702626547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,1,balanced,0.6333813269933065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,64,power_law_1.2,0.063155198097229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,1,balanced,0.9121812979380289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,1,balanced,1.0728639761606853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,2,1,balanced,1.3394773801167805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,2,1,balanced,2.056015968322754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,2,1,balanced,3.916442553202311
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,64,power_law_1.2,0.062438398599624634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,64,power_law_1.2,0.055078399181365964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,64,power_law_1.2,0.055622398853302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,64,power_law_1.2,0.05877760052680969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,64,power_law_1.2,0.06103039979934692
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,64,power_law_1.2,0.06023039817810059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,64,power_law_1.2,0.061919999122619626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,4,16,power_law_1.2,0.03919360041618347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,64,power_law_1.2,0.06128000020980835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,4,16,power_law_1.2,0.03553279936313629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,4,16,power_law_1.2,0.03320960104465485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,64,power_law_1.2,0.061952000856399535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,4,16,power_law_1.2,0.033164799213409424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,4,16,power_law_1.2,0.03342719972133636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,4,16,power_law_1.2,0.03568640053272247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,64,power_law_1.2,0.0629696011543274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,4,16,power_law_1.2,0.03566080033779144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,4,16,power_law_1.2,0.03573119938373566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,4,16,power_law_1.2,0.03565439879894257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,64,power_law_1.2,0.06488959789276123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,4,16,power_law_1.2,0.03583999872207642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,4,16,power_law_1.2,0.037049600481987
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,4,16,power_law_1.2,0.037503999471664426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,64,power_law_1.2,0.06724479794502258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,4,16,power_law_1.2,0.037964800000190736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,4,16,power_law_1.2,0.039801600575447085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,4,16,power_law_1.2,0.044249600172042845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,64,power_law_1.2,0.06653439998626709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,4,16,power_law_1.2,0.04540160000324249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,4,16,power_law_1.2,0.046758401393890384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,4,16,power_law_1.2,0.049420800805091855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,64,power_law_1.2,0.07279360294342041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,4,16,power_law_1.2,0.05441920161247253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,4,16,power_law_1.2,0.06265599727630615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,4,16,power_law_1.2,0.07191680073738098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,64,power_law_1.2,0.07410560250282287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,4,16,power_law_1.2,0.0902463972568512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,4,16,power_law_1.2,0.10968320369720459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,64,power_law_1.2,0.08008319735527039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,4,16,power_law_1.2,0.14396159648895263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,4,16,power_law_1.2,0.18094079494476317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,64,power_law_1.2,0.10042239427566528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,4,16,power_law_1.2,0.365664005279541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,4,16,power_law_1.2,0.359116792678833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,64,power_law_1.2,0.10971519947052003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,4,16,power_law_1.2,0.4652095794677734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,64,power_law_1.2,0.14135680198669434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,4,16,power_law_1.2,0.7379456043243409
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,64,power_law_1.2,0.16727039813995362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,4,16,power_law_1.2,1.6989824295043945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,64,power_law_1.2,0.22615039348602295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,64,power_law_1.2,0.24510080814361573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,64,power_law_1.2,0.3240895986557007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,64,power_law_1.2,0.40970239639282224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,64,power_law_1.2,0.5955647945404052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,1,power_law_1.2,0.09678080081939697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,64,power_law_1.2,0.7787712097167969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,1,power_law_1.2,0.18032000064849854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,1,power_law_1.2,0.16732159852981568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,1,power_law_1.2,0.3025536060333252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,1,power_law_1.2,0.46215038299560546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,4,64,power_law_1.2,0.9789312362670899
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,1,power_law_1.2,0.6593088150024414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,1,power_law_1.2,1.0660096168518067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,1,power_law_1.2,1.0825920104980469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,1,power_law_1.2,1.1913536071777344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,4,64,power_law_1.2,1.633145523071289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,1,power_law_1.2,1.2223615646362305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,1,power_law_1.2,1.2906047821044921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,1,power_law_1.2,1.3807359695434571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,1,power_law_1.2,1.3890624046325684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,1,power_law_1.2,1.4474944114685058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,1,power_law_1.2,1.5348031997680665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,4,64,power_law_1.2,3.4108097076416017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,1,power_law_1.2,1.5179264068603515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,1,power_law_1.2,1.6267520904541015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,1,power_law_1.2,1.7779455184936523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,1,power_law_1.2,1.826963233947754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,1,power_law_1.2,2.088857650756836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,1,power_law_1.2,2.0975744247436525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,1,power_law_1.2,2.454534339904785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,1,power_law_1.2,2.739481544494629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,1,power_law_1.2,3.346681594848633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,2,balanced,0.0684746652841568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,2,balanced,0.07248533268769582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,2,balanced,0.07929599781831105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,2,balanced,0.10703466335932414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,2,balanced,0.14940266807874045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,2,balanced,0.22684266169865927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,2,balanced,0.3063039978345235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,2,balanced,0.31019200881322223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,2,balanced,0.3116319974263509
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,2,balanced,0.3166026671727498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,2,balanced,0.3148053288459778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,2,balanced,0.31621332963307697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,2,balanced,0.3194933334986369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,2,balanced,0.3205066720644633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,2,balanced,0.3299093246459961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,2,balanced,0.33399999141693115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,2,balanced,0.34603198369344074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,2,balanced,0.36903464794158936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,1,power_law_1.2,3.753683090209961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,2,balanced,0.3894666830698649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,2,balanced,0.43490668137868244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,2,balanced,0.47332266966501874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,2,balanced,0.5636213223139445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,2,balanced,0.65611732006073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,2,balanced,0.8483839829762777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,2,balanced,1.108357350031535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,2,balanced,1.4688000679016113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,2,balanced,1.9646080334981282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,1,power_law_1.2,4.875263977050781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,2,balanced,2.4486133257548013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,2,balanced,3.703141212463379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,2,balanced,7.060965220133464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,1,power_law_1.2,5.963724899291992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,1,power_law_1.2,7.105356597900391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,1,power_law_1.2,10.503961944580078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,32,power_law_1.2,0.19584640264511108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,1,power_law_1.2,19.186277770996092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,32,power_law_1.2,0.2563456058502197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,32,power_law_1.2,0.20477440357208251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,32,power_law_1.2,0.10391680002212525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,32,power_law_1.2,0.14644479751586914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,32,power_law_1.2,0.17264000177383423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,32,power_law_1.2,0.18638720512390136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,32,power_law_1.2,0.2072511911392212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,32,power_law_1.2,0.2005824089050293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,32,power_law_1.2,0.21875839233398436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,16,power_law_1.01,0.032339200377464294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,16,power_law_1.01,0.030604800581932066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,32,power_law_1.2,0.22121601104736327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,16,power_law_1.01,0.029388800263404846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,16,power_law_1.01,0.029139199852943422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,32,power_law_1.2,0.22695679664611818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,16,power_law_1.01,0.030086401104927062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,16,power_law_1.01,0.031276801228523256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,32,power_law_1.2,0.21724159717559816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,16,power_law_1.01,0.03185920119285583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,16,power_law_1.01,0.031839999556541446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,16,power_law_1.01,0.03198080062866211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,32,power_law_1.2,0.22842879295349122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,16,power_law_1.01,0.03215999901294708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,16,power_law_1.01,0.032390400767326355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,32,power_law_1.2,0.24039039611816407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,16,power_law_1.01,0.03441280126571655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,16,power_law_1.01,0.03446399867534637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,32,power_law_1.01,0.03294720053672791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,16,power_law_1.01,0.03619199991226196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,32,power_law_1.2,0.23299839496612548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,32,power_law_1.01,0.028512001037597656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,16,power_law_1.01,0.039212799072265624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,32,power_law_1.01,0.028537601232528687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,16,power_law_1.01,0.039526399970054624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,32,power_law_1.2,0.2533951997756958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,32,power_law_1.01,0.02781440019607544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,16,power_law_1.01,0.04426240026950836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,32,power_law_1.01,0.029440000653266907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,16,power_law_1.01,0.04733439981937408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,32,power_law_1.01,0.029811200499534608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,16,power_law_1.01,0.0549888014793396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,32,power_law_1.2,0.2628607988357544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,32,power_law_1.01,0.03020159900188446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,16,power_law_1.01,0.05799040198326111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,32,power_law_1.01,0.03049600124359131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,16,power_law_1.01,0.0655232012271881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,32,power_law_1.2,0.2842751979827881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,32,power_law_1.01,0.03049600124359131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,16,power_law_1.01,0.08616960048675537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,32,power_law_1.01,0.030937600135803222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,32,power_law_1.01,0.031385600566864014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,16,power_law_1.01,0.09874560236930847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,32,power_law_1.2,0.33019518852233887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,32,power_law_1.01,0.031667199730873105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,16,power_law_1.01,0.14410879611968994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,32,power_law_1.01,0.032671999931335446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,16,power_law_1.01,0.16660480499267577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,32,power_law_1.01,0.03402880132198334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,32,power_law_1.2,0.36663041114807127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,32,power_law_1.01,0.03726080060005188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,16,power_law_1.01,0.22654080390930176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,32,power_law_1.01,0.03771519958972931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,32,power_law_1.2,0.45281281471252444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,32,power_law_1.01,0.039673599600791934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,16,power_law_1.01,0.3202431917190552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,32,power_law_1.01,0.04447999894618988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,32,power_law_1.01,0.04462080001831055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,32,power_law_1.2,0.5562943935394287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,16,power_law_1.01,0.3857088088989258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,32,power_law_1.01,0.048895999789237976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,32,power_law_1.01,0.05440000295639038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,32,power_law_1.01,0.06930559873580933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,32,power_law_1.2,0.6582208156585694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,16,power_law_1.01,0.5261631965637207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,32,power_law_1.01,0.0809664011001587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,32,power_law_1.01,0.11779199838638306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,32,power_law_1.2,0.8897279739379883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,32,power_law_1.01,0.14376959800720215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,16,power_law_1.01,1.1995264053344727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,32,power_law_1.01,0.2020672082901001
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,32,power_law_1.2,1.1652416229248046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,32,power_law_1.01,0.26654078960418703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,32,power_law_1.01,0.3161344051361084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,32,power_law_1.2,1.6298944473266601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,32,power_law_1.01,0.4860352039337158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,32,power_law_1.2,2.0266048431396486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,32,power_law_1.01,1.0029312133789063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,32,power_law_1.2,3.2339454650878907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,1,2,power_law_1.2,0.03540480136871338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,32,power_law_1.2,5.695660781860352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,1,2,power_law_1.2,0.03562879860401154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,1,2,power_law_1.2,0.042368000745773314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,1,2,power_law_1.2,0.05182719826698303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,1,2,power_law_1.2,0.06910079717636108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,1,2,power_law_1.2,0.10533759593963624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,1,2,power_law_1.2,0.11089279651641845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,1,2,power_law_1.2,0.11163519620895386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,1,2,power_law_1.2,0.11607040166854858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,1,2,power_law_1.2,0.11358079910278321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,1,2,power_law_1.2,0.11475839614868164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,1,2,power_law_1.2,0.12247680425643921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,1,2,power_law_1.2,0.11942399740219116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,1,2,power_law_1.2,0.12526079416275024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,1,2,power_law_1.2,0.13013119697570802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,1,power_law_1.2,0.0695743978023529
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,1,2,power_law_1.2,0.13346560001373292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,1,2,power_law_1.2,0.13878400325775148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,1,2,power_law_1.2,0.15141760110855101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,1,power_law_1.2,0.09159039855003356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,2,power_law_1.2,0.16810879707336426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,2,power_law_1.2,0.2028736114501953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,2,power_law_1.2,0.20549120903015136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,1,power_law_1.2,0.10228480100631714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,2,power_law_1.2,0.25798399448394777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,2,power_law_1.2,0.3203520059585571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,4,power_law_1.2,0.05053439736366272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,1,power_law_1.2,0.13932160139083863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,2,power_law_1.2,0.3936448097229004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,2,power_law_1.2,0.543455982208252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,4,power_law_1.2,0.05260159969329834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,1,power_law_1.2,0.1927296042442322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,2,power_law_1.2,0.711897611618042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,4,power_law_1.2,0.056806397438049314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,2,power_law_1.2,0.8617024421691895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,1,power_law_1.2,0.3306175947189331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,4,power_law_1.2,0.0593280017375946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,2,power_law_1.2,1.0297344207763672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,1,power_law_1.2,0.36179840564727783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,4,power_law_1.2,0.06780800223350525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,2,power_law_1.2,1.7511680603027344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,1,power_law_1.2,0.3879807949066162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,4,power_law_1.2,0.07985280156135559
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,4,power_law_1.2,0.09104639887809754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,1,power_law_1.2,0.4031231880187988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,2,power_law_1.2,3.597772979736328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,4,power_law_1.2,0.1063040018081665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,1,power_law_1.2,0.40417280197143557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,4,power_law_1.2,0.11346559524536133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,1,power_law_1.2,0.42947840690612793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,4,power_law_1.2,0.11656320095062256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,1,power_law_1.2,0.44219517707824707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,4,power_law_1.2,0.1229632019996643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,1,power_law_1.2,0.4600704193115234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,4,power_law_1.2,0.12317440509796143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,4,power_law_1.2,0.1302783966064453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,1,power_law_1.2,0.4825471878051758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,4,power_law_1.2,0.1412927985191345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,1,power_law_1.2,0.49601922035217283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,4,power_law_1.2,0.1528447985649109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,1,power_law_1.2,0.51561598777771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,4,power_law_1.2,0.16121599674224854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,4,power_law_1.2,0.17436800003051758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,1,power_law_1.2,0.5517951965332031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,32,power_law_1.01,0.07512320280075073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,4,power_law_1.2,0.2025536060333252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,1,power_law_1.2,0.6451200008392334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,4,power_law_1.2,0.2114880084991455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,32,power_law_1.01,0.09017599821090698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,1,power_law_1.2,0.6941567897796631
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,4,power_law_1.2,0.25421440601348877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,32,power_law_1.01,0.05375999808311462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,4,power_law_1.2,0.29767038822174074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,1,power_law_1.2,0.8478079795837402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,32,power_law_1.01,0.06079999804496765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,4,power_law_1.2,0.3945280075073242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,1,power_law_1.2,1.0587008476257325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,32,power_law_1.01,0.06459519863128663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,4,power_law_1.2,0.5078527927398682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,1,power_law_1.2,1.3472064018249512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,32,power_law_1.01,0.07169280052185059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,4,power_law_1.2,0.6554175853729248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,32,power_law_1.01,0.07630079984664917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,4,power_law_1.2,0.8206336021423339
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,1,power_law_1.2,1.647750473022461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,32,power_law_1.01,0.07681919932365418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,4,power_law_1.2,1.1782336235046387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,1,power_law_1.2,2.2153472900390625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,32,power_law_1.01,0.07964159846305847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,4,power_law_1.2,1.4714112281799316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,32,power_law_1.01,0.07960960268974304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,1,power_law_1.2,2.641119956970215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,32,power_law_1.01,0.08005759716033936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,4,power_law_1.2,1.823219108581543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,32,power_law_1.01,0.08498560190200806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,1,power_law_1.2,3.774739074707031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,4,power_law_1.2,2.835315132141113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,32,power_law_1.01,0.08453760147094727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,32,power_law_1.01,0.08872960209846496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,32,power_law_1.01,0.09464319944381713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,1,power_law_1.2,4.954195022583008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,4,power_law_1.2,5.342073440551758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,32,power_law_1.01,0.09730560183525086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,32,power_law_1.01,0.10364799499511719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,8,1,power_law_1.2,5.979935836791992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,32,power_law_1.01,0.11864320039749146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,32,power_law_1.01,0.12743040323257446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,32,power_law_1.01,0.15459840297698973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,32,power_law_1.01,0.18546559810638427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,8,1,power_law_1.2,9.015449523925781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,32,power_law_1.01,0.23297278881072997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,8,4,power_law_1.01,0.03820799887180328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,8,4,power_law_1.01,0.035795199871063235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,8,4,power_law_1.01,0.03454079926013946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,8,4,power_law_1.01,0.033888000249862674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,8,4,power_law_1.01,0.03468160033226013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,8,4,power_law_1.01,0.035097599029541016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,32,power_law_1.01,0.250220799446106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,8,4,power_law_1.01,0.035872000455856326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,8,4,power_law_1.01,0.03544319868087768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,8,4,power_law_1.01,0.036185601353645326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,8,4,power_law_1.01,0.03622399866580963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,8,4,power_law_1.01,0.037574398517608645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,8,4,power_law_1.01,0.03880319893360138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,8,4,power_law_1.01,0.03984639942646027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,8,4,power_law_1.01,0.04240640103816986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,8,4,power_law_1.01,0.048076799511909483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,8,4,power_law_1.01,0.05185920000076294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,8,4,power_law_1.01,0.051507198810577394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,32,power_law_1.01,0.3283071994781494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,8,4,power_law_1.01,0.061343997716903687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,8,4,power_law_1.01,0.0645632028579712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,8,4,power_law_1.01,0.07772799730300903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,8,4,power_law_1.01,0.09538559913635254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,8,4,power_law_1.01,0.13320959806442262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,8,4,power_law_1.01,0.14817919731140136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,8,4,power_law_1.01,0.21453440189361572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,8,4,power_law_1.01,0.21824638843536376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,8,4,power_law_1.01,0.4385216236114502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,32,power_law_1.01,0.39407360553741455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,8,4,power_law_1.01,0.5570559978485108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,8,4,power_law_1.01,0.5826432228088378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,8,4,power_law_1.01,0.9557184219360352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,8,4,power_law_1.01,1.6594240188598632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,8,1,power_law_1.2,16.988018798828126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,32,power_law_1.01,0.5620927810668945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,32,power_law_1.01,0.6715583801269531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,2,32,power_law_1.01,0.8396415710449219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,2,32,power_law_1.01,1.2574848175048827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,1,2,power_law_1.01,0.1351744055747986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,1,2,power_law_1.01,0.18963199853897095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,1,2,power_law_1.01,0.17103999853134155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,1,2,power_law_1.01,0.2330751895904541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,1,power_law_1.2,0.05443840026855469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,1,2,power_law_1.01,0.23276159763336182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,1,2,power_law_1.01,0.23676159381866455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,1,2,power_law_1.01,0.24546558856964112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,1,2,power_law_1.01,0.27861120700836184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,1,2,power_law_1.01,0.29687039852142333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,1,2,power_law_1.01,0.3073663949966431
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,1,power_law_1.2,0.08716800212860107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,1,2,power_law_1.01,0.27956480979919435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,1,2,power_law_1.01,0.2854720115661621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,1,2,power_law_1.01,0.2887743949890137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,1,2,power_law_1.01,0.29687039852142333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,1,2,power_law_1.01,0.3036288022994995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,1,2,power_law_1.01,0.3182463884353638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,1,power_law_1.2,0.08487679958343505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,1,2,power_law_1.01,0.37292799949645994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,1,2,power_law_1.01,0.4428864002227783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,1,2,power_law_1.01,0.48137598037719725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,1,2,power_law_1.01,0.6343423843383789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,1,2,power_law_1.01,0.716102409362793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,1,power_law_1.2,0.12223999500274658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,1,2,power_law_1.01,0.9488960266113281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,1,2,power_law_1.01,1.1367679595947267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,2,32,power_law_1.01,2.681452751159668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,1,power_law_1.2,0.182259202003479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,1,2,power_law_1.01,2.0323968887329102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,1,2,power_law_1.01,2.5934015274047852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,1,power_law_1.2,0.24161279201507568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,1,2,power_law_1.01,3.4479679107666015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,1,power_law_1.2,0.29536640644073486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,1,2,power_law_1.01,5.027084732055664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,1,power_law_1.2,0.4021183967590332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,1,2,power_law_1.01,5.856032180786133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,1,power_law_1.2,0.4336063861846924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,1,power_law_1.2,0.44423680305480956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,1,2,power_law_1.01,9.858048248291016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,1,power_law_1.2,0.47342720031738283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,1,power_law_1.2,0.49584641456604006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,1,power_law_1.2,0.5141503810882568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,16,balanced,0.04594666759173075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,16,balanced,0.04785599807898203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,16,balanced,0.04711466530958811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,16,balanced,0.04987200101216634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,16,balanced,0.07239999870459239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,1,2,power_law_1.01,18.853395080566408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,16,balanced,0.09934932986895244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,16,balanced,0.10050666332244873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,16,balanced,0.10281599561373393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,16,balanced,0.10128000378608704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,16,balanced,0.10310399532318115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,16,balanced,0.10429867108662923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,16,balanced,0.10681600371996562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,16,balanced,0.10737599929173787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,16,balanced,0.1091306706269582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,16,balanced,0.11500799655914307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,16,balanced,0.11508267124493916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,1,power_law_1.2,0.5396287918090821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,16,balanced,0.12198932965596516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,16,balanced,0.13034666577974954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,16,balanced,0.13920533657073975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,16,balanced,0.15921599666277567
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,16,balanced,0.18099733193715414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,16,balanced,0.21976532538731894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,16,balanced,0.25391467412312824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,16,balanced,0.33875731627146405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,16,balanced,0.41198933124542236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,16,balanced,0.5726293325424194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,16,balanced,0.7199467023213705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,1,power_law_1.2,0.5645952224731445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,16,balanced,0.8907413482666016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,16,balanced,1.3253119786580403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,16,balanced,2.584437370300293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,1,power_law_1.2,0.595359992980957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,1,power_law_1.2,0.6307775974273682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,1,power_law_1.2,0.6923264026641845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,1,power_law_1.2,0.7569983959197998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,1,power_law_1.2,0.8749695777893066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,1,power_law_1.2,0.9778112411499024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,1,4,power_law_1.2,0.08655359745025634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,1,power_law_1.2,1.1904959678649902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,1,4,power_law_1.2,0.07253760099411011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,1,4,power_law_1.2,0.08972799777984619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,1,4,power_law_1.2,0.08727040290832519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,1,4,power_law_1.2,0.08644480109214783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,1,4,power_law_1.2,0.08933759927749634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,1,4,power_law_1.2,0.09132800102233887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,1,4,power_law_1.2,0.0942463994026184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,1,4,power_law_1.2,0.0991807997226715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,1,power_law_1.2,1.4483327865600586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,1,4,power_law_1.2,0.10083839893341065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,1,4,power_law_1.2,0.099481600522995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,1,4,power_law_1.2,0.10675200223922729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,1,4,power_law_1.2,0.10840959548950195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,1,4,power_law_1.2,0.1170240044593811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,1,4,power_law_1.2,0.13653759956359862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,1,4,power_law_1.2,0.14229120016098024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,1,4,power_law_1.2,0.1670591950416565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,4,8,power_law_1.2,0.039027199149131775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,1,4,power_law_1.2,0.17713919878005982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,1,power_law_1.2,1.8346239089965821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,1,4,power_law_1.2,0.24055678844451905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,1,4,power_law_1.2,0.3187839984893799
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,4,8,power_law_1.2,0.03425920009613037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,1,4,power_law_1.2,0.37043840885162355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,1,4,power_law_1.2,0.5620480060577393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,4,8,power_law_1.2,0.03329919874668121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,1,4,power_law_1.2,0.7330751895904541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,4,8,power_law_1.2,0.035020801424980166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,1,4,power_law_1.2,0.8733759880065918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,1,power_law_1.2,2.1101247787475588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,4,8,power_law_1.2,0.03619840145111084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,1,4,power_law_1.2,1.1807040214538573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,4,8,power_law_1.2,0.03925119936466217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,1,4,power_law_1.2,2.2499584197998046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,4,8,power_law_1.2,0.03924480080604553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,4,8,power_law_1.2,0.03916159868240356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,1,4,power_law_1.2,3.3009342193603515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,4,8,power_law_1.2,0.03947519958019256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,1,power_law_1.2,2.761740875244141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,4,8,power_law_1.2,0.04007680118083954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,1,4,power_law_1.2,3.6506622314453123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,4,8,power_law_1.2,0.04090240001678467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,4,8,power_law_1.2,0.04158720076084137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,4,8,power_law_1.2,0.04301440119743347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,1,4,power_law_1.2,5.849318313598633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,1,power_law_1.2,3.5499649047851562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,4,8,power_law_1.2,0.044147199392318724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,4,8,power_law_1.2,0.04787200093269348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,4,8,power_law_1.2,0.04935039877891541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,4,8,power_law_1.2,0.052928000688552856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,4,8,power_law_1.2,0.06257280111312866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,1,power_law_1.2,4.2780609130859375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,4,8,power_law_1.2,0.07028480172157288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,1,4,power_law_1.2,11.591654205322266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,4,8,power_law_1.2,0.07651839852333069
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,4,8,power_law_1.2,0.0861952006816864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,4,8,power_law_1.2,0.10736000537872314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,4,8,power_law_1.2,0.13684480190277098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,4,8,power_law_1.2,0.18894720077514648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,1,power_law_1.2,6.264793777465821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,4,8,power_law_1.2,0.26908159255981445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,4,8,power_law_1.2,0.38289918899536135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,4,8,power_law_1.2,0.5333888053894043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,4,8,power_law_1.2,0.5604159832000732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,1,power_law_1.2,11.605267333984376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,4,8,power_law_1.2,0.9510784149169922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,4,8,power_law_1.2,2.3182144165039062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,16,balanced,0.05026133358478546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,16,balanced,0.06050133208433787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,16,balanced,0.07794133325417836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,16,balanced,0.10552533467610677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,16,balanced,0.13505599896113077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,16,balanced,0.15294399857521057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,16,balanced,0.15465066830317178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,16,balanced,0.1549013356367747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,16,balanced,0.15590932965278625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,16,balanced,0.15546666582425436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,16,balanced,0.1567520002524058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,16,balanced,0.15833600362141928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,16,balanced,0.15934933225313822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,16,balanced,0.16022933522860208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,16,balanced,0.16552000244458517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,16,balanced,0.16698133945465088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,16,balanced,0.17058134078979492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,16,balanced,0.1978773276011149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,16,balanced,0.19079999128977457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,16,balanced,0.23683732748031616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,16,balanced,0.23315733671188354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,16,balanced,0.3489919900894165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,16,balanced,0.2972426613171895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,16,balanced,0.42502399285634357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,16,balanced,0.41894932587941486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,16,balanced,0.5973600149154663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,16,balanced,0.664576013882955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,2,16,balanced,0.8334346612294515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,2,16,balanced,1.1858932971954346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,2,16,balanced,2.398245334625244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,4,balanced,0.040234667559464775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,4,balanced,0.03844800094763438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,4,balanced,0.04228266576925913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,4,balanced,0.05273066461086273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,4,balanced,0.0728000005086263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,4,balanced,0.07459733386834462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,4,balanced,0.07473599910736084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,4,balanced,0.07494933406511943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,4,balanced,0.07562666634718578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,4,balanced,0.07649066547552745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,4,balanced,0.07606400052706401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,4,balanced,0.07766399780909221
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,4,balanced,0.07964799801508586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,4,balanced,0.08123733103275299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,4,balanced,0.0862506628036499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,4,balanced,0.08887466788291931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,4,balanced,0.09055466453234355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,4,balanced,0.09758399923642476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,4,balanced,0.10605333248774211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,4,balanced,0.12261866529782613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,4,balanced,0.13462932904561362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,4,balanced,0.17637866735458374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,4,balanced,0.19725332657496134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,4,balanced,0.2773120005925496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,4,balanced,0.3407680193583171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,4,balanced,0.49165332317352295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,4,balanced,0.6351999839146932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,4,balanced,0.7798026402791342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,4,balanced,1.1848906675974529
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,4,balanced,2.2851200103759766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,32,power_law_1.01,0.044275200366973876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,32,power_law_1.01,0.03430399894714355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,32,power_law_1.01,0.030681601166725157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,32,power_law_1.01,0.028339201211929323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,2,balanced,0.033514666060606636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,2,balanced,0.03554133325815201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,2,balanced,0.04570133487383524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,2,balanced,0.07029866675535838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,2,balanced,0.10402133067448933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,2,balanced,0.12822932998339334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,32,power_law_1.01,0.03020159900188446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,2,balanced,0.12917332847913107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,2,balanced,0.12963199615478516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,2,balanced,0.12900267044703165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,2,balanced,0.12925333778063455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,2,balanced,0.1309760014216105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,2,balanced,0.1316426694393158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,2,balanced,0.13410666584968567
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,2,balanced,0.13270399967829385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,2,balanced,0.13846400380134583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,2,balanced,0.14642133315404257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,32,power_law_1.01,0.03112959861755371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,2,balanced,0.14607999722162882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,2,balanced,0.15408000349998474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,2,balanced,0.1577173372109731
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,2,balanced,0.19050133228302002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,2,balanced,0.17932265996932983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,2,balanced,0.24862933158874512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,2,balanced,0.2485333283742269
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,32,power_law_1.01,0.031711998581886294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,2,balanced,0.34815998872121173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,2,balanced,0.3843573331832886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,2,balanced,0.5514026482899984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,2,balanced,0.6345920165379842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,32,power_law_1.01,0.03193599879741669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,1,2,balanced,0.7866666316986084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,1,2,balanced,1.193013350168864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,32,power_law_1.01,0.03111039996147156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,1,2,balanced,2.2561813990275064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,32,power_law_1.01,0.03203200101852417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,32,power_law_1.01,0.03255679905414581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,32,power_law_1.01,0.032601600885391234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,32,power_law_1.01,0.033292800188064575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,32,power_law_1.01,0.03431040048599243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,32,power_law_1.01,0.039001598954200745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,32,power_law_1.01,0.04017280042171478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,32,power_law_1.01,0.04312320053577423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,32,power_law_1.01,0.04547199904918671
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,32,power_law_1.01,0.04835200011730194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,4,2,power_law_1.2,0.03928320109844208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,32,power_law_1.01,0.05367680191993714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,4,2,power_law_1.2,0.03935999870300293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,4,2,power_law_1.2,0.04172160029411316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,4,2,power_law_1.2,0.0449535995721817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,32,power_law_1.01,0.057068800926208495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,4,2,power_law_1.2,0.04609920084476471
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,4,2,power_law_1.2,0.04635519981384277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,4,2,power_law_1.2,0.04669440090656281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,32,power_law_1.01,0.06995199918746949
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,4,2,power_law_1.2,0.04661760032176972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,4,2,power_law_1.2,0.051052802801132204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,4,2,power_law_1.2,0.051622402667999265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,32,power_law_1.01,0.08423680067062378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,4,2,power_law_1.2,0.05231999754905701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,4,2,power_law_1.2,0.05575680136680603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,4,2,power_law_1.2,0.057036799192428586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,32,power_law_1.01,0.1021888017654419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,4,2,power_law_1.2,0.06568959951400757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,4,2,power_law_1.2,0.0726144015789032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,4,2,power_law_1.2,0.0782912015914917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,4,2,power_law_1.2,0.07912960052490234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,32,power_law_1.01,0.12027519941329956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,4,2,power_law_1.2,0.08985599875450134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,4,2,power_law_1.2,0.09493759870529175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,4,2,power_law_1.2,0.12106239795684814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,4,2,power_law_1.2,0.15633920431137086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,32,power_law_1.01,0.15616639852523803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,4,2,power_law_1.2,0.21264638900756835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,4,2,power_law_1.2,0.2514944076538086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,4,2,power_law_1.2,0.3613312005996704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,32,power_law_1.01,0.17644799947738649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,4,2,power_law_1.2,0.4293951988220215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,4,2,power_law_1.2,0.5428671836853027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,16,balanced,0.045610666275024414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,16,balanced,0.0484799991051356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,16,balanced,0.04790933430194855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,16,balanced,0.07444799939791362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,16,balanced,0.1011946698029836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,16,balanced,0.11266666650772095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,16,balanced,0.11504000425338745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,4,2,power_law_1.2,0.6430912017822266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,2,32,power_law_1.01,0.20556159019470216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,16,balanced,0.11477333307266235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,16,balanced,0.11699199676513672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,16,balanced,0.11459733049074809
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,16,balanced,0.11637866497039795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,16,balanced,0.11654933293660481
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,16,balanced,0.11745599905649821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,16,balanced,0.11755200227101643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,16,balanced,0.12390933434168498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,4,2,power_law_1.2,0.9109503746032714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,16,balanced,0.12287466724713643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,16,balanced,0.125082661708196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,16,balanced,0.14083199699719748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,16,balanced,0.13402666648228964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,16,balanced,0.15967466433842978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,16,balanced,0.15499732891718546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,16,balanced,0.21686933437983194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,16,balanced,0.2001919945081075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,16,balanced,0.3429280122121175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,2,32,power_law_1.01,0.34637439250946045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,16,balanced,0.29585067431132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,4,2,power_law_1.2,1.602739143371582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,16,balanced,0.43193598588307697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,16,balanced,0.49092264970143634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,16,balanced,0.5616960128148397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,16,balanced,0.8440799713134766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,16,balanced,1.5871680577596028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,4,2,power_law_1.2,3.1332544326782226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,2,32,power_law_1.01,0.6204800128936767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,2,balanced,0.043951998154322304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,2,balanced,0.04979733129342397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,2,balanced,0.06901866694291432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,2,balanced,0.09925333658854167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,2,balanced,0.15036267042160034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,2,balanced,0.15385066469510397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,2,balanced,0.1549066702524821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,2,balanced,0.15492266416549683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,4,power_law_1.2,0.04620159864425659
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,2,balanced,0.15734400351842245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,2,balanced,0.15967466433842978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,2,balanced,0.16014400124549866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,2,balanced,0.1621226668357849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,2,balanced,0.16383999586105347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,2,balanced,0.16753600041071573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,2,balanced,0.17278399070103964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,8,power_law_1.2,0.08327680230140685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,2,balanced,0.17648533980051676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,2,balanced,0.1861120065053304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,2,balanced,0.20428266127904257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,2,balanced,0.22317866484324136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,2,balanced,0.2543626626332601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,2,balanced,0.2840106685956319
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,4,power_law_1.2,0.05554559826850891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,2,balanced,0.3821440140406291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,2,balanced,0.4278186559677124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,2,balanced,0.6077226797739664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,2,balanced,0.7239733537038168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,2,balanced,1.0323893229166667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,4,power_law_1.2,0.051635199785232545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,2,balanced,1.347589333852132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,8,power_law_1.2,0.08399360179901123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,2,balanced,1.665770689646403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,4,power_law_1.2,0.05550079941749573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,2,balanced,2.604981263478597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,8,power_law_1.2,0.06281599998474122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,4,power_law_1.2,0.06476799845695495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,2,balanced,4.983776092529297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,4,power_law_1.2,0.0744704008102417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,8,power_law_1.2,0.078847998380661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,4,power_law_1.2,0.09171839952468872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,4,power_law_1.2,0.09747200012207032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,8,power_law_1.2,0.09188479781150818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,4,power_law_1.2,0.09774079918861389
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,4,power_law_1.2,0.10046720504760742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,8,power_law_1.2,0.1273792028427124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,4,power_law_1.2,0.10454399585723877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,8,power_law_1.2,0.13553279638290405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,4,power_law_1.2,0.10709120035171509
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,4,power_law_1.2,0.11259520053863525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,8,power_law_1.2,0.13119360208511352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,4,power_law_1.2,0.1189568042755127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,4,power_law_1.2,0.12566399574279785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,8,power_law_1.2,0.1431615948677063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,4,power_law_1.2,0.13524479866027833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,8,power_law_1.2,0.14092799425125122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,4,power_law_1.2,0.140556800365448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,4,power_law_1.2,0.15667200088500977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,8,power_law_1.2,0.14451839923858642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,4,power_law_1.01,0.09481599926948547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,4,power_law_1.2,0.17614079713821412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,8,power_law_1.2,0.15047680139541625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,4,power_law_1.2,0.21052799224853516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,4,power_law_1.01,0.15544320344924928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,4,power_law_1.2,0.2453632116317749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,8,power_law_1.2,0.15210239887237548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,4,power_law_1.01,0.24348800182342528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,4,power_law_1.2,0.30373759269714357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,8,power_law_1.2,0.15843839645385743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,4,power_law_1.2,0.3807487964630127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,4,power_law_1.01,0.12559360265731812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,4,power_law_1.2,0.4984447956085205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,8,power_law_1.2,0.16813440322875978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,4,power_law_1.01,0.1904960036277771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,4,power_law_1.2,0.6799039840698242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,4,power_law_1.01,0.2846719980239868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,8,power_law_1.2,0.17280639410018922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,4,power_law_1.2,0.9103360176086426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,4,power_law_1.01,0.3312896013259888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,8,power_law_1.2,0.18106240034103394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,4,power_law_1.01,0.4441408157348633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,4,power_law_1.2,1.1093759536743164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,8,power_law_1.2,0.20728321075439454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,4,power_law_1.01,0.44255361557006834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,4,power_law_1.2,1.526803207397461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,4,power_law_1.01,0.4580927848815918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,8,power_law_1.2,0.2304447889328003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,4,power_law_1.01,0.4823296070098877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,4,power_law_1.2,2.16376953125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,8,power_law_1.2,0.2857215881347656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,4,power_law_1.01,0.49770240783691405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,8,power_law_1.2,0.30757761001586914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,4,power_law_1.01,0.48920321464538574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,4,power_law_1.2,4.517382431030273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,4,power_law_1.01,0.5292928218841553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,8,power_law_1.2,0.3881151914596558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,4,power_law_1.01,0.5411839962005616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,8,power_law_1.2,0.45151357650756835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,4,power_law_1.01,0.5581759929656982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,4,power_law_1.01,0.5729919910430908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,8,power_law_1.2,0.5877632141113281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,4,power_law_1.01,0.6068607807159424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,8,power_law_1.2,0.7294271945953369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,4,power_law_1.01,0.6719295978546143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,4,power_law_1.01,0.8031935691833496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,8,power_law_1.2,1.040608024597168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,4,power_law_1.01,0.7984384059906006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,4,power_law_1.01,0.9459327697753906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,4,4,power_law_1.2,0.04320000112056732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,4,4,power_law_1.2,0.040300801396369934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,4,4,power_law_1.2,0.04063999950885773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,4,4,power_law_1.2,0.041606399416923526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,8,power_law_1.2,1.1908224105834961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,4,4,power_law_1.2,0.04135040044784546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,4,power_law_1.01,1.0670592308044433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,4,4,power_law_1.2,0.04161919951438904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,4,4,power_law_1.2,0.04196479916572571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,4,4,power_law_1.2,0.043731200695037845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,4,4,power_law_1.2,0.04757120013237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,4,4,power_law_1.2,0.05059840083122254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,4,4,power_law_1.2,0.051507198810577394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,4,4,power_law_1.2,0.05676800012588501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,4,4,power_law_1.2,0.055155199766159055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,4,4,power_law_1.2,0.06357759833335877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,4,power_law_1.01,1.348198413848877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,4,4,power_law_1.2,0.06849920153617858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,4,4,power_law_1.2,0.07069439888000488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,4,4,power_law_1.2,0.0837119996547699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,4,4,power_law_1.2,0.09658880233764648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,4,4,power_law_1.2,0.11347839832305909
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,4,4,power_law_1.2,0.1471168041229248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,4,4,power_law_1.2,0.178656005859375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,4,4,power_law_1.2,0.23873279094696045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,4,8,power_law_1.2,1.5847423553466797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,4,power_law_1.01,1.4654144287109374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,4,4,power_law_1.2,0.25892479419708253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,4,4,power_law_1.2,0.3390271902084351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,4,4,power_law_1.2,0.4757887840270996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,4,4,power_law_1.2,0.6209983825683594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,4,4,power_law_1.2,0.8459136009216308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,4,power_law_1.01,1.8864063262939452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,4,4,power_law_1.2,1.179961585998535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,4,4,power_law_1.2,1.735647964477539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,4,power_law_1.01,2.3030527114868162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,4,4,power_law_1.2,3.6375873565673826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,4,8,power_law_1.2,2.486783981323242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,4,power_law_1.01,2.8136383056640626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,4,power_law_1.01,4.160044860839844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,4,8,power_law_1.2,5.093529510498047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,4,power_law_1.01,7.6995903015136715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,4,4,balanced,0.052101333936055504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,4,4,balanced,0.04178133110205332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,4,4,balanced,0.05829333265622457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,4,4,balanced,0.05894400179386139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,4,4,balanced,0.05877333382765452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,4,4,balanced,0.05875733494758606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,4,4,balanced,0.05977066854635874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,4,4,balanced,0.05816000203291575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,4,4,balanced,0.05970133344332377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,4,4,balanced,0.060133333007494606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,4,4,balanced,0.059936001896858215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,4,4,balanced,0.060640002290407814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,4,4,balanced,0.06001600126425425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,4,4,balanced,0.06452266871929169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,4,4,balanced,0.07420800129572551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,4,4,balanced,0.07167466481526692
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,4,4,balanced,0.07512533167997996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,4,4,balanced,0.08089066545168559
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,4,4,balanced,0.08715732892354329
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,4,4,balanced,0.1037600040435791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,4,4,balanced,0.10730666915575664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,4,4,balanced,0.14009599884351095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,4,4,balanced,0.1588746706644694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,4,4,balanced,0.21178134282430014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,4,4,balanced,0.25062400102615356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,4,4,balanced,0.3532533248265584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,4,4,balanced,0.44943467775980633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,4,4,balanced,0.5543200174967448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,4,4,balanced,0.8463413715362549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,4,4,balanced,1.6093120574951172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,64,power_law_1.01,0.047225600481033324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,64,power_law_1.01,0.03521920144557953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,64,power_law_1.01,0.028147199749946596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,4,16,power_law_1.01,0.026976001262664796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,64,power_law_1.01,0.030796799063682555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,4,16,power_law_1.01,0.022489599883556366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,64,power_law_1.01,0.030265599489212036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,4,16,power_law_1.01,0.023206399381160737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,64,power_law_1.01,0.030579200387001036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,4,16,power_law_1.01,0.022668799757957457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,64,power_law_1.01,0.030880001187324525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,4,16,power_law_1.01,0.022899200022220612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,64,power_law_1.01,0.03110400140285492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,4,16,power_law_1.01,0.02418559938669205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,64,power_law_1.01,0.030585598945617676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,4,16,power_law_1.01,0.024447999894618988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,64,power_law_1.01,0.031200000643730165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,8,1,power_law_1.01,0.048230400681495665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,4,16,power_law_1.01,0.026047998666763307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,64,power_law_1.01,0.03108479976654053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,8,1,power_law_1.01,0.05123839974403381
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,4,16,power_law_1.01,0.0302592009305954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,64,power_law_1.01,0.032793599367141726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,8,1,power_law_1.01,0.05910400152206421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,64,power_law_1.01,0.033529600501060484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,4,16,power_law_1.01,0.029798400402069092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,8,1,power_law_1.01,0.06944640278816223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,64,power_law_1.01,0.03596160113811493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,4,16,power_law_1.01,0.030444800853729248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,64,power_law_1.01,0.039827200770378116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,8,1,power_law_1.01,0.0795199990272522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,4,16,power_law_1.01,0.031091201305389404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,64,power_law_1.01,0.04103040099143982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,8,1,power_law_1.01,0.11119359731674194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,4,16,power_law_1.01,0.031891199946403506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,64,power_law_1.01,0.0417279988527298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,8,1,power_law_1.01,0.11594239473342896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,4,16,power_law_1.01,0.030777600407600404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,64,power_law_1.01,0.04652799963951111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,8,1,power_law_1.01,0.11741440296173096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,4,16,power_law_1.01,0.03356800079345703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,64,power_law_1.01,0.045561599731445315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,8,1,power_law_1.01,0.11945600509643554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,4,16,power_law_1.01,0.03514240086078644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,64,power_law_1.01,0.051123201847076416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,8,1,power_law_1.01,0.12139519453048705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,4,16,power_law_1.01,0.03617919981479645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,64,power_law_1.01,0.056543999910354616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,8,1,power_law_1.01,0.12778879404067994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,4,16,power_law_1.01,0.04022400081157684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,8,1,power_law_1.01,0.13084160089492797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,64,power_law_1.01,0.07024639844894409
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,16,power_law_1.01,0.044844800233840944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,8,1,power_law_1.01,0.13343360424041747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,64,power_law_1.01,0.08135039806365967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,16,power_law_1.01,0.04812160134315491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,8,1,power_law_1.01,0.13719680309295654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,64,power_law_1.01,0.10832639932632446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,16,power_law_1.01,0.05235840082168579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,8,1,power_law_1.01,0.14554879665374756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,16,power_law_1.01,0.06934400200843811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,64,power_law_1.01,0.12723840475082399
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,8,1,power_law_1.01,0.15128320455551147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,16,power_law_1.01,0.07580159902572632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,64,power_law_1.01,0.18391040563583375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,8,1,power_law_1.01,0.16210559606552125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,16,power_law_1.01,0.09672319889068604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,8,1,power_law_1.01,0.1958847999572754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,64,power_law_1.01,0.17830400466918944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,8,1,power_law_1.01,0.22470400333404542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,16,power_law_1.01,0.1236799955368042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,8,1,power_law_1.01,0.28847360610961914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,1,64,power_law_1.01,0.24812159538269044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,16,power_law_1.01,0.1531775951385498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,8,1,power_law_1.01,0.34615039825439453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,8,1,power_law_1.01,0.47870721817016604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,1,64,power_law_1.01,0.32337279319763185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,16,power_law_1.01,0.23140480518341064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,8,1,power_law_1.01,0.6004032135009766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,16,power_law_1.01,0.22946560382843018
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,8,1,power_law_1.01,0.8514240264892579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,1,64,power_law_1.01,0.6335552215576172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,8,1,power_law_1.01,1.092915153503418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,16,power_law_1.01,0.3773247957229614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,8,1,power_law_1.01,1.575872039794922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,16,power_law_1.01,0.7921535968780518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,8,1,power_law_1.01,2.0776512145996096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,8,1,power_law_1.01,2.5584896087646483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,8,1,power_law_1.01,3.9923774719238283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,8,1,power_law_1.01,7.803692626953125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,1,power_law_1.2,0.09941120147705078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,1,power_law_1.2,0.11650559902191163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,1,power_law_1.2,0.19292800426483153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,1,power_law_1.2,0.29192960262298584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,1,power_law_1.2,0.42817277908325196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,1,power_law_1.2,0.8141568183898926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,16,power_law_1.01,0.06182399988174438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,1,power_law_1.2,0.9058239936828614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,16,power_law_1.01,0.06451839804649354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,1,power_law_1.2,0.9373056411743164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,1,power_law_1.2,0.9691328048706055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,16,power_law_1.01,0.06626560091972351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,1,power_law_1.2,1.0013567924499511
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,16,power_law_1.01,0.053299200534820554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,1,power_law_1.2,1.0380800247192383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,1,power_law_1.2,1.0406271934509277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,16,power_law_1.01,0.05488640069961548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,1,power_law_1.2,1.0972864151000976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,16,power_law_1.01,0.058771198987960814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,1,power_law_1.2,1.1277055740356445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,1,power_law_1.2,1.1425600051879883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,64,power_law_1.01,0.06650879979133606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,16,power_law_1.01,0.06233599781990051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,1,power_law_1.2,1.1831616401672362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,64,power_law_1.01,0.06989439725875854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,16,power_law_1.01,0.06357759833335877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,1,power_law_1.2,1.1948415756225585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,64,power_law_1.01,0.06252800226211548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,1,power_law_1.2,1.2858367919921876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,16,power_law_1.01,0.06672639846801758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,64,power_law_1.01,0.0621504008769989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,1,power_law_1.2,1.3643903732299805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,64,power_law_1.01,0.061843198537826535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,16,power_law_1.01,0.06627200245857238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,1,power_law_1.2,1.5471936225891114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,64,power_law_1.01,0.06371840238571166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,16,power_law_1.01,0.06898559927940369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,1,power_law_1.2,1.7640447616577148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,64,power_law_1.01,0.06592000126838685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,16,power_law_1.01,0.06963840126991272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,1,power_law_1.2,2.1433151245117186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,64,power_law_1.01,0.06552960276603699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,64,power_law_1.01,0.06615039706230164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,1,power_law_1.2,2.232703971862793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,16,power_law_1.01,0.07471359968185425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,64,power_law_1.01,0.06614400148391723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,16,power_law_1.01,0.07428479790687562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,1,power_law_1.2,2.7861055374145507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,64,power_law_1.01,0.06707839965820313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,16,power_law_1.01,0.0812991976737976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,64,power_law_1.01,0.0664192020893097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,1,power_law_1.2,3.6896640777587892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,64,power_law_1.01,0.06910719871520996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,16,power_law_1.01,0.0828544020652771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,64,power_law_1.01,0.06789119839668274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,1,power_law_1.2,4.7304832458496096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,16,power_law_1.01,0.08932480216026306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,64,power_law_1.01,0.07478399872779846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,16,power_law_1.01,0.10254080295562744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,64,power_law_1.01,0.07472000122070313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,1,power_law_1.2,5.852934265136719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,64,power_law_1.01,0.07797759771347046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,16,power_law_1.01,0.11535999774932862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,64,power_law_1.01,0.08498560190200806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,16,power_law_1.01,0.14134399890899657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,64,power_law_1.01,0.09644160270690919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,1,power_law_1.2,6.9127037048339846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,64,power_law_1.01,0.11420799493789673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,16,power_law_1.01,0.15952639579772948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,64,power_law_1.01,0.1370751976966858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,16,power_law_1.01,0.2105151891708374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,64,power_law_1.01,0.17350399494171143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,16,power_law_1.01,0.24131200313568116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,1,power_law_1.2,9.880384063720703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,64,power_law_1.01,0.2027967929840088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,64,power_law_1.01,0.26727681159973143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,16,power_law_1.01,0.31902079582214354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,64,power_law_1.01,0.3382848024368286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,16,power_law_1.01,0.4018496036529541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,64,power_law_1.01,0.4829567909240723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,16,power_law_1.01,0.5633855819702148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,64,power_law_1.01,0.6124927997589111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,1,power_law_1.2,18.225439453125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,16,power_law_1.01,0.7515840053558349
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,64,power_law_1.01,0.8084287643432617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,16,power_law_1.01,0.9197248458862305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,64,power_law_1.01,1.320518398284912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,16,power_law_1.01,1.4397952079772949
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,64,power_law_1.01,2.4776384353637697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,16,power_law_1.01,2.7158336639404297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,1,power_law_1.01,0.03281919956207276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,1,power_law_1.01,0.03752320110797882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,1,power_law_1.01,0.04442879855632782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,1,power_law_1.01,0.05138559937477112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,1,power_law_1.01,0.06554239988327026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,1,power_law_1.01,0.08984959721565247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,1,power_law_1.01,0.0929472029209137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,1,power_law_1.01,0.09667840003967285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,1,power_law_1.01,0.09781119823455811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,1,power_law_1.01,0.10190080404281616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,1,power_law_1.01,0.10898560285568237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,1,power_law_1.01,0.11263359785079956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,1,power_law_1.01,0.1191167950630188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,1,power_law_1.01,0.12076159715652465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,2,power_law_1.01,0.042559999227523806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,1,power_law_1.01,0.12877440452575684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,2,power_law_1.01,0.0392192006111145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,1,power_law_1.01,0.13566720485687256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,2,power_law_1.01,0.04067200124263763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,1,power_law_1.01,0.1486143946647644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,2,power_law_1.01,0.04505600035190582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,1,power_law_1.01,0.1821887969970703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,1,power_law_1.01,0.20814080238342286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,2,power_law_1.01,0.04930559992790222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,1,power_law_1.01,0.26640000343322756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,2,power_law_1.01,0.058822399377822875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,2,power_law_1.01,0.05976960062980652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,1,power_law_1.01,0.32340478897094727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,2,power_law_1.01,0.060249602794647215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,1,power_law_1.01,0.4371647834777832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,2,power_law_1.01,0.06222079992294312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,1,power_law_1.01,0.5508992195129394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,2,power_law_1.01,0.06284800171852112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,2,power_law_1.01,0.06627839803695679
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,1,power_law_1.01,0.7736063957214355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,2,power_law_1.01,0.06857600212097167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,1,power_law_1.01,0.9954431533813477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,2,power_law_1.01,0.07089920043945312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,2,power_law_1.01,0.0719488024711609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,1,power_law_1.01,1.4311936378479004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,2,power_law_1.01,0.07836800217628478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,2,power_law_1.01,0.0807807981967926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,2,power_law_1.01,0.0883072018623352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,1,power_law_1.01,1.8479488372802735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,2,power_law_1.01,0.10217599868774414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,2,power_law_1.01,0.11323519945144653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,1,power_law_1.01,2.2936447143554686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,2,power_law_1.01,0.1413632035255432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,2,power_law_1.01,0.17335679531097412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,2,power_law_1.01,0.22541439533233643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,2,power_law_1.01,0.27231359481811523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,1,power_law_1.01,3.5924606323242188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,2,power_law_1.01,0.39680640697479247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,2,power_law_1.01,0.5251584053039551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,2,power_law_1.01,0.7639679908752441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,2,power_law_1.01,1.0032768249511719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,1,power_law_1.01,6.984294128417969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,2,power_law_1.01,1.2079232215881348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,2,power_law_1.01,1.9320512771606446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,2,power_law_1.01,3.422739028930664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,16,power_law_1.01,0.03919360041618347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,16,power_law_1.01,0.03226880133152008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,16,power_law_1.01,0.030508801341056824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,16,power_law_1.01,0.03073279857635498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,16,power_law_1.01,0.03209599852561951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,16,power_law_1.01,0.034643200039863584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,16,power_law_1.01,0.035231998562812804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,16,power_law_1.01,0.03573119938373566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,16,power_law_1.01,0.03594880104064942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,16,power_law_1.01,0.037625598907470706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,16,power_law_1.01,0.04095360040664673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,16,balanced,0.03306666761636734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,16,balanced,0.03160533308982849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,16,balanced,0.03296533226966858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,16,balanced,0.033615998923778534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,16,balanced,0.03540800015131632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,16,balanced,0.03741333385308584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,16,balanced,0.037647999823093414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,16,power_law_1.01,0.04257279932498932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,16,balanced,0.037658666570981346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,16,balanced,0.03772266705830892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,16,balanced,0.03748266647259394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,16,balanced,0.03976533313592275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,16,balanced,0.03949866692225138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,16,balanced,0.039488000174363456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,16,balanced,0.04171200096607208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,16,balanced,0.044112001856168113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,16,balanced,0.04659200211366018
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,16,balanced,0.04774933556715647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,16,balanced,0.054117331902186074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,16,power_law_1.01,0.045126399397850035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,16,balanced,0.06108800073464712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,16,balanced,0.07679466903209686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,16,balanced,0.08478400111198425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,16,balanced,0.10948266585667928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,16,balanced,0.12573867042859396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,16,balanced,0.15687466661135355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,16,balanced,0.2020853360493978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,16,power_law_1.01,0.04314880073070526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,16,balanced,0.27157866954803467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,16,balanced,0.33692800998687744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,16,balanced,0.40979735056559247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,16,balanced,0.621018648147583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,16,power_law_1.01,0.052857601642608644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,16,balanced,1.18885334332784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,16,power_law_1.01,0.05562880039215088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,16,power_law_1.01,0.05440000295639038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,16,power_law_1.01,0.06696320176124573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,16,power_law_1.01,0.07416960000991821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,16,power_law_1.01,0.09497600197792053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,16,power_law_1.01,0.1094208002090454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,16,power_law_1.01,0.14375679492950438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,16,power_law_1.01,0.166758394241333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,16,power_law_1.01,0.22174720764160155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,16,power_law_1.01,0.30847361087799074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,16,power_law_1.01,0.40023040771484375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,16,power_law_1.01,0.5102528095245361
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,16,power_law_1.01,0.653875207901001
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,16,power_law_1.01,1.006809616088867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,1,power_law_1.01,0.03375999927520752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,1,power_law_1.01,0.033107200264930726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,1,power_law_1.01,0.03692159950733185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,1,power_law_1.01,0.041075199842453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,16,power_law_1.01,2.1059263229370115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,1,power_law_1.01,0.04826880097389221
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,1,power_law_1.01,0.060288000106811526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,1,power_law_1.01,0.0634112000465393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,1,power_law_1.01,0.06542080044746398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,1,power_law_1.01,0.06815999746322632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,1,power_law_1.01,0.06913920044898987
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,1,power_law_1.01,0.0725823998451233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,1,power_law_1.01,0.07767040133476258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,1,power_law_1.01,0.08187519907951354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,1,power_law_1.01,0.08247680068016053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,1,power_law_1.01,0.09189760088920593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,1,power_law_1.01,0.09549440145492553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,1,power_law_1.01,0.10325119495391846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,1,power_law_1.01,0.12201600074768067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,1,power_law_1.01,0.14433280229568482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,1,power_law_1.01,0.1816831946372986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,1,power_law_1.01,0.2170367956161499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,1,power_law_1.01,0.29972479343414304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,1,power_law_1.01,0.38341119289398196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,1,power_law_1.01,0.5490496158599854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,1,power_law_1.01,0.7101503849029541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,2,power_law_1.2,0.0689087986946106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,1,power_law_1.01,1.046828842163086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,1,power_law_1.01,1.3677887916564941
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,2,power_law_1.2,0.06387199759483338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,1,power_law_1.01,1.6877056121826173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,2,power_law_1.2,0.07898880243301391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,1,power_law_1.01,2.6553600311279295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,2,power_law_1.2,0.10382080078125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,2,power_law_1.2,0.12558079957962037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,1,power_law_1.01,5.251379013061523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,2,power_law_1.2,0.19715839624404907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,2,power_law_1.2,0.21490559577941895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,2,power_law_1.2,0.22181758880615235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,2,power_law_1.2,0.23179519176483154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,2,power_law_1.2,0.24007039070129393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,2,power_law_1.2,0.24539520740509033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,2,power_law_1.2,0.25830399990081787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,2,4,power_law_1.2,0.06504960060119629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,2,power_law_1.2,0.26080639362335206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,2,4,power_law_1.2,0.06241919994354248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,2,4,power_law_1.2,0.06580479741096497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,2,4,power_law_1.2,0.06584320068359376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,2,4,power_law_1.2,0.06663039922714234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,2,4,power_law_1.2,0.06684160232543945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,2,4,power_law_1.2,0.06711040139198303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,2,4,power_law_1.2,0.06800000071525573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,2,4,power_law_1.2,0.06977919936180114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,2,power_law_1.2,0.2738624095916748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,2,4,power_law_1.2,0.06908159852027893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,2,4,power_law_1.2,0.07244160175323486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,2,4,power_law_1.2,0.07779840230941773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,2,4,power_law_1.2,0.07853440046310425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,2,4,power_law_1.2,0.08214399814605713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,2,4,power_law_1.2,0.08903679847717286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,2,4,power_law_1.2,0.09604480266571044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,2,power_law_1.2,0.28510079383850095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,2,4,power_law_1.2,0.10805120468139648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,2,4,power_law_1.2,0.1260480046272278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,2,4,power_law_1.2,0.15336960554122925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,2,4,power_law_1.2,0.20246400833129882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,2,4,power_law_1.2,0.2225600004196167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,2,4,power_law_1.2,0.34565119743347167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,2,4,power_law_1.2,0.4339776039123535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,2,power_law_1.2,0.29569919109344484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,2,4,power_law_1.2,0.6374271869659424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,2,4,power_law_1.2,0.7662591934204102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,2,4,power_law_1.2,1.062508773803711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,2,4,power_law_1.2,1.3501824378967284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,2,power_law_1.2,0.311078405380249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,2,4,power_law_1.2,2.201651191711426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,2,power_law_1.2,0.36428160667419435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,2,4,power_law_1.2,2.969990348815918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,2,power_law_1.2,0.3938623905181885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,2,4,power_law_1.2,5.560697555541992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,2,power_law_1.2,0.4971776008605957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,2,power_law_1.2,0.6011968135833741
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,2,power_law_1.2,0.7479872226715087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,2,power_law_1.2,0.8462783813476562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,2,power_law_1.2,1.1736576080322265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,16,4,power_law_1.01,0.03255679905414581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,16,4,power_law_1.01,0.03028480112552643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,16,4,power_law_1.01,0.02974080145359039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,2,power_law_1.2,1.4366208076477052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,16,4,power_law_1.01,0.030406400561332703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,16,4,power_law_1.01,0.03022719919681549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,16,4,power_law_1.01,0.031667199730873105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,16,4,power_law_1.01,0.03158400058746338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,16,4,power_law_1.01,0.033369600772857666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,16,4,power_law_1.01,0.0342848002910614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,16,4,power_law_1.01,0.03468160033226013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,16,4,power_law_1.01,0.034745600819587705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,16,4,power_law_1.01,0.03594239950180054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,16,4,power_law_1.01,0.036262398958206175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,16,4,power_law_1.01,0.03731200098991394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,16,4,power_law_1.01,0.045440000295639035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,16,4,power_law_1.01,0.04674560129642487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,16,4,power_law_1.01,0.04583680033683777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,16,4,power_law_1.01,0.056435197591781616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,16,4,power_law_1.01,0.06736639738082886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,2,power_law_1.2,1.989900779724121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,16,4,power_law_1.01,0.07861120104789734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,16,4,power_law_1.01,0.08878080248832702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,16,4,power_law_1.01,0.12614400386810304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,16,4,power_law_1.01,0.16787840127944947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,16,4,power_law_1.01,0.20554239749908448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,16,4,power_law_1.01,0.29191040992736816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,16,4,power_law_1.01,0.4304448127746582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,16,4,power_law_1.01,0.608780813217163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,16,4,power_law_1.01,0.7226111888885498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,16,4,power_law_1.01,1.1050880432128907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,2,power_law_1.2,2.596569633483887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,16,4,power_law_1.01,1.9960319519042968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,4,2,power_law_1.2,3.025267219543457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,4,2,power_law_1.2,4.836019134521484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,4,power_law_1.01,0.046675199270248414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,4,power_law_1.01,0.051744002103805545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,4,power_law_1.01,0.07390080094337463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,4,2,power_law_1.2,9.957273864746094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,4,power_law_1.01,0.05668479800224304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,4,power_law_1.01,0.07086719870567322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,4,power_law_1.01,0.08536319732666016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,4,power_law_1.01,0.09440000057220459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,4,power_law_1.01,0.10779520273208618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,4,power_law_1.01,0.11185920238494873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,4,power_law_1.01,0.11381759643554687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,4,power_law_1.01,0.1199679970741272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,4,power_law_1.01,0.1250048041343689
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,1,power_law_1.2,0.030444800853729248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,4,power_law_1.01,0.12688640356063843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,1,power_law_1.2,0.035667198896408084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,1,power_law_1.2,0.04053759872913361
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,4,power_law_1.01,0.13214720487594606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,1,power_law_1.2,0.05268480181694031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,1,power_law_1.2,0.06392319798469544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,4,power_law_1.01,0.14843519926071166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,1,power_law_1.2,0.10056960582733154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,1,power_law_1.2,0.10408320426940917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,1,power_law_1.2,0.11034239530563354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,4,power_law_1.01,0.15701119899749755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,1,power_law_1.2,0.11575679779052735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,1,power_law_1.2,0.11997439861297607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,4,power_law_1.01,0.16022399663925171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,1,power_law_1.2,0.12268799543380737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,1,power_law_1.2,0.12502399682998658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,4,power_law_1.01,0.17863039970397948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,1,power_law_1.2,0.1273151993751526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,1,power_law_1.2,0.1299072027206421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,1,power_law_1.2,0.14494080543518068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,4,power_law_1.01,0.19335039854049682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,1,power_law_1.2,0.14921599626541138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,1,power_law_1.2,0.15318399667739868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,4,power_law_1.01,0.22665600776672362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,1,power_law_1.2,0.17567360401153564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,1,power_law_1.2,0.17564159631729126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,1,power_law_1.2,0.20888960361480713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,4,power_law_1.01,0.2552639961242676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,1,power_law_1.2,0.23781120777130127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,1,power_law_1.2,0.2939136028289795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,4,power_law_1.01,0.3439872026443481
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,1,power_law_1.2,0.36445438861846924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,1,power_law_1.2,0.4676799774169922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,4,power_law_1.01,0.42101120948791504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,1,power_law_1.2,0.5303487777709961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,1,power_law_1.2,0.733574390411377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,4,power_law_1.01,0.5383232116699219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,1,power_law_1.2,0.9435839653015137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,4,power_law_1.01,0.6698368072509766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,32,1,power_law_1.01,0.05182719826698303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,4,1,power_law_1.2,1.1528384208679199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,32,1,power_law_1.01,0.05613440275192261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,4,power_law_1.01,0.9219903945922852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,4,1,power_law_1.2,1.7492288589477538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,32,1,power_law_1.01,0.06416640281677247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,4,power_law_1.01,1.1700991630554198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,4,1,power_law_1.2,3.3743167877197267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,32,1,power_law_1.01,0.07861120104789734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,4,power_law_1.01,1.4196991920471191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,32,1,power_law_1.01,0.11520639657974244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,4,power_law_1.01,2.2000448226928713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,32,1,power_law_1.01,0.14788479804992677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,32,1,power_law_1.01,0.1627071976661682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,32,1,power_law_1.01,0.20131840705871581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,4,power_law_1.01,4.231564712524414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,32,1,power_law_1.01,0.21227519512176513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,32,1,power_law_1.01,0.21634559631347655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,4,power_law_1.01,0.056883198022842404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,32,1,power_law_1.01,0.22529919147491456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,4,power_law_1.01,0.06220160126686096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,32,1,power_law_1.01,0.2367487907409668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,4,power_law_1.01,0.057843202352523805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,4,power_law_1.01,0.06213120222091675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,32,1,power_law_1.01,0.24303359985351564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,4,power_law_1.01,0.07169920206069946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,4,power_law_1.01,0.07959679961204529
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,32,1,power_law_1.01,0.25467519760131835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,4,power_law_1.01,0.09002879858016968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,4,power_law_1.01,0.09199360013008118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,32,1,power_law_1.01,0.27130880355834963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,4,power_law_1.01,0.09354239702224731
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,32,1,power_law_1.01,0.2771199941635132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,4,power_law_1.01,0.09478399753570557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,1,power_law_1.01,0.11880960464477539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,4,power_law_1.01,0.09790719747543335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,32,1,power_law_1.01,0.2937855958938599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,4,power_law_1.01,0.10020480155944825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,1,power_law_1.01,0.18817280530929564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,4,power_law_1.01,0.10247039794921875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,32,1,power_law_1.01,0.34307839870452883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,4,power_law_1.01,0.10984959602355956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,1,power_law_1.01,0.2369663953781128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,4,power_law_1.01,0.11665279865264892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,32,1,power_law_1.01,0.3863039970397949
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,4,power_law_1.01,0.12135039567947388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,1,power_law_1.01,0.450105619430542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,4,power_law_1.01,0.12548480033874512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,32,1,power_law_1.01,0.4725632190704346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,4,power_law_1.01,0.1400704026222229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,1,power_law_1.01,0.6363711833953858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,4,power_law_1.01,0.15852799415588378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,32,1,power_law_1.01,0.5694143772125244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,4,power_law_1.01,0.19130239486694336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,1,power_law_1.01,1.1163776397705079
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,4,power_law_1.01,0.2263551950454712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,32,1,power_law_1.01,0.7405183792114258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,4,power_law_1.01,0.28394880294799807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,1,power_law_1.01,1.2386752128601075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,4,power_law_1.01,0.34535040855407717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,32,1,power_law_1.01,0.921241569519043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,1,power_law_1.01,1.303097629547119
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,4,power_law_1.01,0.4697472095489502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,4,power_law_1.01,0.6129792213439942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,32,1,power_law_1.01,1.257049560546875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,1,power_law_1.01,1.3449983596801758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,4,power_law_1.01,0.8743359565734863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,32,1,power_law_1.01,1.479148769378662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,1,power_law_1.01,1.3788607597351075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,4,power_law_1.01,1.109875202178955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,1,power_law_1.01,1.4115903854370118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,32,1,power_law_1.01,2.1124031066894533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,4,power_law_1.01,1.3558336257934571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,1,power_law_1.01,1.4410623550415038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,4,power_law_1.01,2.2684736251831055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,32,1,power_law_1.01,2.719615936279297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,1,power_law_1.01,1.5021375656127929
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,32,2,power_law_1.2,0.035648000240325925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,32,2,power_law_1.2,0.03641600012779236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,32,2,power_law_1.2,0.035071998834609985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,32,2,power_law_1.2,0.03661440014839172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,32,2,power_law_1.2,0.03713279962539673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,32,2,power_law_1.2,0.037990400195121767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,32,2,power_law_1.2,0.03900800049304962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,32,2,power_law_1.2,0.038764798641204835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,1,power_law_1.01,1.5560256004333497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,32,2,power_law_1.2,0.03940480053424835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,32,2,power_law_1.2,0.039980798959732056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,32,2,power_law_1.2,0.040403199195861814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,32,1,power_law_1.01,3.338163375854492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,32,2,power_law_1.2,0.0416703999042511
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,4,power_law_1.01,4.135718536376953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,32,2,power_law_1.2,0.042419201135635375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,32,2,power_law_1.2,0.043609601259231565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,32,2,power_law_1.2,0.04917759895324707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,32,2,power_law_1.2,0.05139840245246887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,32,2,power_law_1.2,0.05345919728279114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,32,2,power_law_1.2,0.06883199810981751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,1,power_law_1.01,1.58919677734375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,32,2,power_law_1.2,0.07343999743461609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,32,2,power_law_1.2,0.08762879967689514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,32,2,power_law_1.2,0.10577919483184814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,32,2,power_law_1.2,0.13159680366516113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,32,2,power_law_1.2,0.16944639682769774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,32,2,power_law_1.2,0.24447360038757324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,32,2,power_law_1.2,0.2639616012573242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,1,power_law_1.01,1.6330495834350587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,32,2,power_law_1.2,0.4119103908538818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,32,1,power_law_1.01,5.311404800415039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,32,2,power_law_1.2,0.5129856109619141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,32,2,power_law_1.2,0.6890304088592529
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,32,2,power_law_1.2,1.0440640449523926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,1,power_law_1.01,1.706719970703125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,32,2,power_law_1.2,2.069811248779297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,1,power_law_1.01,1.8857599258422852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,32,1,power_law_1.01,10.112351989746093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,1,power_law_1.01,1.8995264053344727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,1,power_law_1.01,2.1680320739746093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,32,power_law_1.01,0.06455680131912231
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,1,power_law_1.01,2.3891584396362306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,32,power_law_1.01,0.056979197263717654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,32,power_law_1.01,0.049568000435829165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,32,power_law_1.01,0.049542400240898135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,1,power_law_1.01,2.936844825744629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,32,power_law_1.01,0.05173119902610779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,32,power_law_1.01,0.051526397466659546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,32,power_law_1.01,0.05384320020675659
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,8,power_law_1.2,0.05013120174407959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,1,power_law_1.01,2.95034236907959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,32,power_law_1.01,0.054099202156066895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,8,power_law_1.2,0.07002239823341369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,32,power_law_1.01,0.05457280278205871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,4,1,balanced,0.05994666616121928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,8,power_law_1.2,0.06537600159645081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,4,1,balanced,0.11803733309110005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,32,power_law_1.01,0.055846399068832396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,4,1,balanced,0.14008532961209616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,4,1,balanced,0.13167466719945273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,4,1,balanced,0.13540266950925192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,4,1,balanced,0.13578133781750998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,4,1,balanced,0.1363200048605601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,1,power_law_1.01,3.619481658935547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,4,1,balanced,0.13714667161305746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,4,1,balanced,0.13493333260218301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,4,1,balanced,0.13660800457000732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,4,1,balanced,0.14123200376828512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,32,power_law_1.01,0.056390398740768434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,4,1,balanced,0.13991999626159668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,4,1,balanced,0.141594668229421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,4,1,balanced,0.1456160048643748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,8,power_law_1.2,0.05624319911003113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,4,1,balanced,0.15691199898719788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,4,1,balanced,0.16077867150306702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,4,1,balanced,0.16404267152150473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,4,1,balanced,0.19107200702031454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,4,1,balanced,0.20534932613372803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,4,1,balanced,0.2470560073852539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,32,power_law_1.01,0.057183998823165896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,4,1,balanced,0.27561066548029584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,4,1,balanced,0.3600320021311442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,4,1,balanced,0.44380799929300946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,4,1,balanced,0.6318773428599039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,8,power_law_1.2,0.06469759941101075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,4,1,balanced,0.8098239898681641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,32,power_law_1.01,0.05861120223999024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,4,1,balanced,1.1884586811065674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,4,1,balanced,1.5494346618652344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,4,1,balanced,1.9173493385314941
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,32,power_law_1.01,0.05920640230178833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,1,power_law_1.01,4.3146625518798825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,8,power_law_1.2,0.08110719919204712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,4,1,balanced,3.0369440714518228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,32,power_law_1.01,0.06471679806709289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,8,power_law_1.2,0.08706560134887695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,4,1,balanced,5.998645146687825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,32,power_law_1.01,0.06550400257110596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,32,power_law_1.01,0.07249280214309692
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,8,power_law_1.2,0.10051840543746948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,32,power_law_1.01,0.08044800162315369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,1,power_law_1.01,5.853171157836914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,8,power_law_1.2,0.10060160160064698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,32,power_law_1.01,0.09155840277671815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,8,power_law_1.2,0.102457594871521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,32,power_law_1.01,0.10904959440231324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,32,power_law_1.01,0.12684799432754518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,8,power_law_1.2,0.1055616021156311
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,32,power_law_1.01,0.1582208037376404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,8,power_law_1.2,0.11387519836425782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,1,power_law_1.01,7.082790374755859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,32,power_law_1.01,0.18905600309371948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,8,power_law_1.2,0.11536639928817749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,32,power_law_1.01,0.2477952003479004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,8,power_law_1.2,0.12212480306625366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,32,power_law_1.01,0.3118655920028687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,8,power_law_1.2,0.1313024044036865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,32,power_law_1.01,0.4283328056335449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,8,power_law_1.2,0.1327936053276062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,2,1,power_law_1.01,8.502732849121093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,32,power_law_1.01,0.5733823776245117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,8,power_law_1.2,0.14011520147323608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,32,power_law_1.01,0.6974720001220703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,8,power_law_1.2,0.15808000564575195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,64,power_law_1.2,0.10983680486679077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,8,power_law_1.2,0.1728767991065979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,32,power_law_1.01,1.0242815971374513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,8,power_law_1.2,0.20551040172576904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,64,power_law_1.2,0.10698239803314209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,8,power_law_1.2,0.23502719402313232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,32,power_law_1.01,2.1064191818237306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,2,1,power_law_1.01,13.504396057128906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,8,power_law_1.2,0.2876607894897461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,64,power_law_1.2,0.0974399983882904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,8,power_law_1.2,0.34997119903564455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,64,power_law_1.2,0.07943040132522583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,8,power_law_1.2,0.43500800132751466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,8,power_law_1.2,0.5980544090270996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,64,power_law_1.2,0.08639360070228577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,8,power_law_1.2,0.8494208335876465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,64,power_law_1.2,0.09155200123786926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,8,power_law_1.2,0.9528063774108887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,64,power_law_1.2,0.09245439767837524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,8,power_law_1.2,1.1857791900634767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,64,power_law_1.2,0.09902080297470092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,32,balanced,0.03822399924198786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,32,balanced,0.036042665441830955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,32,balanced,0.033973333736260734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,32,balanced,0.035402665535608925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,32,balanced,0.038165333370367684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,2,1,power_law_1.01,24.72852478027344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,32,balanced,0.0379573330283165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,32,balanced,0.038319999972979225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,32,balanced,0.03788266579310099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,32,balanced,0.03766933331886927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,32,balanced,0.03851733356714249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,32,balanced,0.03781333317359289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,32,balanced,0.03806933263937632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,32,balanced,0.03980266551176707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,32,balanced,0.03982399900754293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,32,balanced,0.04411733150482178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,32,balanced,0.04219200213750204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,32,balanced,0.04588800172011057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,32,balanced,0.04618666569391886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,32,balanced,0.0490880012512207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,32,balanced,0.05153599878152212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,32,balanced,0.05473066866397858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,32,balanced,0.06884799897670746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,32,balanced,0.07856533428033192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,32,balanced,0.09734400113423665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,32,balanced,0.11772800485293071
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,32,balanced,0.14934933185577393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,8,power_law_1.2,1.8125823974609374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,32,balanced,0.18477867046991983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,64,power_law_1.2,0.10117119550704956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,32,balanced,0.21575466791788736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,32,balanced,0.31706132491429645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,32,balanced,0.5808106660842896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,64,power_law_1.2,0.1028480052947998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,64,power_law_1.2,0.10517120361328125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,8,power_law_1.2,3.729030227661133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,64,power_law_1.2,0.10348800420761109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,64,power_law_1.2,0.09796479940414429
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,64,power_law_1.2,0.10648319721221924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,64,power_law_1.2,0.10929280519485474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,64,power_law_1.2,0.10786559581756591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,128,balanced,0.07880533238252004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,128,balanced,0.08213866750399272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,128,balanced,0.08018133540948232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,128,balanced,0.07926400005817413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,128,balanced,0.07855466504891713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,128,balanced,0.09359467029571533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,128,balanced,0.12008000413576762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,128,balanced,0.1050986647605896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,128,balanced,0.11688533425331116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,128,balanced,0.11079466342926025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,128,balanced,0.11961066722869873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,128,balanced,0.12392533818880717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,128,balanced,0.11531200011571248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,128,balanced,0.12009599804878235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,128,balanced,0.12325867017110188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,128,balanced,0.12763200203577676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,128,balanced,0.12970667084058127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,128,balanced,0.13214932878812155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,128,balanced,0.1402773360411326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,128,balanced,0.15347199638684592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,64,power_law_1.2,0.1146239995956421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,128,balanced,0.16245333353678384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,128,balanced,0.18402665853500366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,128,balanced,0.20755199591318765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,128,balanced,0.2564000089963277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,128,balanced,0.3365386724472046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,128,balanced,0.42469334602355957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,128,balanced,0.5289599895477295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,1,power_law_1.01,0.05129600167274475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,128,balanced,0.6700533231099447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,128,balanced,0.9948053359985352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,1,power_law_1.01,0.06215680241584778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,128,balanced,1.8731625874837239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,64,power_law_1.2,0.12817280292510985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,1,power_law_1.01,0.08612480163574218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,1,power_law_1.01,0.11567360162734985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,1,power_law_1.01,0.14179199934005737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,64,power_law_1.2,0.13177599906921386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,1,power_law_1.01,0.22058238983154296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,128,power_law_1.2,0.06331520080566407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,1,power_law_1.01,0.24021120071411134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,1,power_law_1.01,0.2498431921005249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,128,power_law_1.2,0.0677183985710144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,64,power_law_1.2,0.16397440433502197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,1,power_law_1.01,0.25347840785980225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,128,power_law_1.2,0.057817602157592775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,1,power_law_1.01,0.2587199926376343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,1,power_law_1.01,0.2649280071258545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,64,power_law_1.2,0.19141119718551636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,128,power_law_1.2,0.06102399826049805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,1,power_law_1.01,0.2735167980194092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,1,power_law_1.01,0.27141120433807375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,128,power_law_1.2,0.05963519811630249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,1,power_law_1.01,0.2839103937149048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,64,power_law_1.2,0.23057920932769777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,128,power_law_1.2,0.06330879926681518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,1,power_law_1.01,0.29870080947875977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,1,power_law_1.01,0.3106368064880371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,128,power_law_1.2,0.06578559875488281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,1,power_law_1.01,0.32490880489349366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,64,power_law_1.2,0.2757567882537842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,1,power_law_1.01,0.3766400098800659
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,128,power_law_1.2,0.06472319960594178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,1,power_law_1.01,0.42335357666015627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,128,power_law_1.2,0.06458879709243774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,1,power_law_1.01,0.5073023796081543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,128,power_law_1.2,0.06394240260124207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,1,power_law_1.01,0.5740863800048828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,64,power_law_1.2,0.3826688051223755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,1,power_law_1.01,0.7338687896728515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,128,power_law_1.2,0.06673280000686646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,8,8,power_law_1.2,0.03854719996452331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,1,power_law_1.01,0.8894335746765136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,128,power_law_1.2,0.06519039869308471
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,8,8,power_law_1.2,0.0355648010969162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,64,power_law_1.2,0.474073600769043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,8,8,power_law_1.2,0.035846400260925296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,1,power_law_1.01,1.2040960311889648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,128,power_law_1.2,0.0666815996170044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,8,8,power_law_1.2,0.038489601016044615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,128,power_law_1.2,0.06674559712409973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,8,8,power_law_1.2,0.03917439877986908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,1,power_law_1.01,1.5008383750915528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,8,8,power_law_1.2,0.04203520119190216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,128,power_law_1.2,0.07137280106544494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,64,power_law_1.2,0.6905983924865723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,8,8,power_law_1.2,0.042483198642730716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,1,power_law_1.01,2.1010368347167967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,8,8,power_law_1.2,0.04377599954605103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,128,power_law_1.2,0.07292159795761108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,8,8,power_law_1.2,0.04385280013084412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,128,power_law_1.2,0.07923840284347534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,8,8,power_law_1.2,0.04412800073623657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,1,power_law_1.01,2.7005887985229493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,64,power_law_1.2,0.9397120475769043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,8,8,power_law_1.2,0.044844800233840944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,128,power_law_1.2,0.09104639887809754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,8,8,power_law_1.2,0.04838399887084961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,8,8,power_law_1.2,0.048416000604629514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,128,power_law_1.2,0.1017024040222168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,1,power_law_1.01,3.332896041870117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,8,8,power_law_1.2,0.05080320239067078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,128,power_law_1.2,0.12791039943695068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,64,power_law_1.2,1.105510425567627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,8,8,power_law_1.2,0.0559935986995697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,8,8,power_law_1.2,0.05742719769477844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,128,power_law_1.2,0.14094079732894899
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,8,8,power_law_1.2,0.06385279893875122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,1,power_law_1.01,5.131609725952148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,8,8,power_law_1.2,0.07409279942512512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,128,power_law_1.2,0.18520959615707397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,8,8,power_law_1.2,0.07909759879112244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,8,8,power_law_1.2,0.0965503990650177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,128,power_law_1.2,0.23824000358581543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,64,power_law_1.2,1.7369087219238282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,8,8,power_law_1.2,0.11004159450531006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,128,power_law_1.2,0.3199295997619629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,8,8,power_law_1.2,0.14563839435577391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,8,8,power_law_1.2,0.1668544054031372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,128,power_law_1.2,0.4190336227416992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,8,8,power_law_1.2,0.2652672052383423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,1,power_law_1.01,9.753606414794922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,8,8,power_law_1.2,0.3006655931472778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,128,power_law_1.2,0.6237184047698975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,64,power_law_1.2,3.3713470458984376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,8,8,power_law_1.2,0.5027904033660888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,128,power_law_1.2,0.8778623580932617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,8,8,power_law_1.2,0.7173183917999267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,128,power_law_1.2,1.0363200187683106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,8,8,power_law_1.2,0.8934911727905274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,8,8,power_law_1.2,1.4329024314880372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,128,power_law_1.2,1.5915583610534667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,8,8,power_law_1.2,2.7270015716552733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,128,power_law_1.2,3.663840103149414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,1,balanced,0.048341333866119385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,1,balanced,0.06534400085608165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,1,balanced,0.10005333026250203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,1,balanced,0.1662773291269938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,1,balanced,0.2951573332150777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,1,balanced,0.3907466729482015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,1,balanced,0.3928000132242839
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,1,balanced,0.3949386676152547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,1,balanced,0.39741865793863934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,1,balanced,0.3965546687444051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,1,balanced,0.39789334932963055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,1,balanced,0.40221333503723145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,1,balanced,0.40248000621795654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,1,balanced,0.40750932693481445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,1,2,balanced,0.06669333577156067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,1,balanced,0.4119199911753337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,1,2,balanced,0.09221333265304565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,1,balanced,0.4149920145670573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,1,2,balanced,0.14342400431632996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,1,balanced,0.4230773448944092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,1,2,balanced,0.140474667151769
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,1,balanced,0.4898666540781657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,1,2,balanced,0.13954133788744608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,1,balanced,0.4490346511205037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,1,2,balanced,0.1443893313407898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,1,2,balanced,0.14336533347765604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,1,balanced,0.5531573295593262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,1,2,balanced,0.14350933829943338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,1,balanced,0.5071146488189697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,1,2,balanced,0.14644799629847208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,1,2,balanced,0.14474667112032572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,1,balanced,0.7723519802093506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,1,2,balanced,0.14858133594195047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,1,balanced,0.6274453401565552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,1,2,balanced,0.14869866768519083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,1,2,balanced,0.14840533336003622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,1,balanced,0.9648213386535645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,1,2,balanced,0.15542399883270264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,1,2,balanced,0.1718239982922872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,1,balanced,0.9404640197753906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,1,2,balanced,0.1750826636950175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,1,2,balanced,0.18280533949534097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,1,2,balanced,0.21311465899149576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,1,balanced,1.3323307037353516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,1,2,balanced,0.2223200003306071
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,1,2,balanced,0.2913279930750529
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,1,balanced,1.4434399604797363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,1,2,balanced,0.32262933254241943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,1,2,balanced,0.4366613229115804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,1,2,balanced,0.5432746807734171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,1,1,balanced,1.8144480387369792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,1,2,balanced,0.7735146681467692
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,1,2,balanced,0.9785760243733724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,1,1,balanced,2.7301225662231445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,1,2,balanced,1.4432800610860188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,1,2,balanced,1.8894826571146648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,1,1,balanced,5.3773549397786455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,1,2,balanced,2.327616055806478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,1,2,balanced,3.69598388671875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,1,2,balanced,7.529258728027344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,16,power_law_1.01,0.06432639956474304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,32,power_law_1.01,0.09646720290184022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,16,power_law_1.01,0.0948415994644165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,32,power_law_1.01,0.12888959646224976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,16,power_law_1.01,0.10524159669876099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,32,power_law_1.01,0.10455039739608765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,16,power_law_1.01,0.06033920049667359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,32,power_law_1.01,0.07592319846153259
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,16,power_law_1.01,0.07163519859313965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,32,power_law_1.01,0.08320000171661376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,16,balanced,0.04980800052483877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,16,balanced,0.05221866567929586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,16,balanced,0.05176533261934916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,16,balanced,0.05395199855168661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,16,balanced,0.062309334675470986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,16,balanced,0.08479467034339905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,16,balanced,0.10280533631642659
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,16,balanced,0.10271466771761577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,16,balanced,0.10406399766604106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,16,balanced,0.10582933823267619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,16,balanced,0.105813334385554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,16,balanced,0.10914666453997295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,16,balanced,0.10748799641927083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,16,power_law_1.01,0.07770879864692688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,16,balanced,0.10920533537864685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,16,balanced,0.11668800314267476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,16,balanced,0.11733866731325786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,32,power_law_1.01,0.08821120262145996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,16,balanced,0.12258133292198181
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,16,balanced,0.1306933363278707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,16,balanced,0.13900267084439596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,16,balanced,0.15726932883262634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,16,balanced,0.17431465784708658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,16,balanced,0.20909333229064941
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,16,balanced,0.24267733097076416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,16,balanced,0.30868266026179
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,16,power_law_1.01,0.08291839957237243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,16,balanced,0.3856426477432251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,32,power_law_1.01,0.09297279715538025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,16,balanced,0.5182666778564453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,16,balanced,0.6631573438644409
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,16,balanced,0.8176213105519613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,16,balanced,1.2189760208129883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,16,power_law_1.01,0.0894976019859314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,32,power_law_1.01,0.0944703996181488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,16,balanced,2.3721866607666016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,16,power_law_1.01,0.08872320055961609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,32,power_law_1.01,0.09650560021400452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,2,power_law_1.01,0.12093440294265748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,16,power_law_1.01,0.08975359797477722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,32,power_law_1.01,0.09670400023460388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,16,power_law_1.01,0.09631999731063842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,32,power_law_1.01,0.09684479832649232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,32,power_law_1.01,0.09719679951667785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,16,power_law_1.01,0.09475200176239014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,2,power_law_1.01,0.18567680120468139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,32,power_law_1.01,0.0965503990650177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,16,power_law_1.01,0.10017280578613282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,32,power_law_1.01,0.10204160213470459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,16,power_law_1.01,0.1031999945640564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,2,power_law_1.01,0.13521920442581176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,32,power_law_1.01,0.10806399583816528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,16,power_law_1.01,0.10675840377807617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,32,power_law_1.01,0.1097216010093689
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,16,power_law_1.01,0.11000959873199463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,32,power_law_1.01,0.1128000020980835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,2,power_law_1.01,0.24788479804992675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,16,power_law_1.01,0.12003200054168701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,128,power_law_1.01,0.06783360242843628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,32,power_law_1.01,0.12353279590606689
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,16,power_law_1.01,0.13330559730529784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,128,power_law_1.01,0.06755200028419495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,32,power_law_1.01,0.12891520261764527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,128,power_law_1.01,0.06431360244750976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,2,power_law_1.01,0.364300799369812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,16,power_law_1.01,0.14652800559997559
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,128,power_law_1.01,0.06669440269470214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,32,power_law_1.01,0.14796160459518432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,16,power_law_1.01,0.17167999744415283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,128,power_law_1.01,0.061267197132110596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,32,power_law_1.01,0.16691839694976807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,2,power_law_1.01,0.5947519779205322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,128,power_law_1.01,0.06488320231437683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,16,power_law_1.01,0.19089280366897582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,32,power_law_1.01,0.21588480472564697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,128,power_law_1.01,0.06308479905128479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,16,power_law_1.01,0.2440704107284546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,128,power_law_1.01,0.06390399932861328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,32,power_law_1.01,0.25106561183929443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,2,power_law_1.01,0.6521791934967041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,128,power_law_1.01,0.0648576021194458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,16,power_law_1.01,0.27204480171203616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,128,power_law_1.01,0.06308479905128479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,32,power_law_1.01,0.31660799980163573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,16,power_law_1.01,0.36728320121765134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,128,power_law_1.01,0.06317440271377564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,2,power_law_1.01,0.6817024230957032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,32,power_law_1.01,0.389683198928833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,128,power_law_1.01,0.0650111973285675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,16,power_law_1.01,0.4746367931365967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,128,power_law_1.01,0.0692351996898651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,32,power_law_1.01,0.5732351779937744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,128,power_law_1.01,0.06690559983253479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,2,power_law_1.01,0.7078847885131836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,16,power_law_1.01,0.6217408180236816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,128,power_law_1.01,0.07086079716682434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,32,power_law_1.01,0.7002560138702393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,128,power_law_1.01,0.0720192015171051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,16,power_law_1.01,0.8218624114990234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,2,power_law_1.01,0.7123072147369385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,128,power_law_1.01,0.07401599884033203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,32,power_law_1.01,0.8292927742004395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,128,power_law_1.01,0.08553599715232849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,16,power_law_1.01,0.9709823608398438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,128,power_law_1.01,0.09456639885902404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,2,power_law_1.01,0.7434304237365723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,128,power_law_1.01,0.11521279811859131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,32,power_law_1.01,1.3582592010498047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,128,power_law_1.01,0.13415679931640626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,16,power_law_1.01,1.4831616401672363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,2,power_law_1.01,0.7658559799194335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,128,power_law_1.01,0.17091200351715088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,128,power_law_1.01,0.20181119441986084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,32,power_law_1.01,2.5815935134887695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,128,power_law_1.01,0.27219200134277344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,2,power_law_1.01,0.7634175777435303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,16,power_law_1.01,2.981158447265625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,128,power_law_1.01,0.3629312038421631
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,2,power_law_1.01,0.8019519805908203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,128,power_law_1.01,0.502675199508667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,128,power_law_1.01,0.6252543926239014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,2,power_law_1.01,0.8157888412475586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,128,power_law_1.01,0.8019519805908203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,2,power_law_1.01,0.8510784149169922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,128,power_law_1.01,1.3091391563415526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,2,power_law_1.01,0.8720064163208008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,128,power_law_1.01,2.556038475036621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,2,power_law_1.01,0.9688192367553711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,1,2,power_law_1.2,0.11942399740219116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,1,2,power_law_1.2,0.16235519647598268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,1,2,power_law_1.2,0.14238719940185546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,1,2,power_law_1.2,0.21189119815826415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,2,power_law_1.01,1.0024831771850586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,1,2,power_law_1.2,0.2021183967590332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,1,2,power_law_1.2,0.21509759426116942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,1,2,power_law_1.2,0.2240000009536743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,1,2,power_law_1.2,0.22050559520721436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,1,2,power_law_1.2,0.22167680263519288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,1,2,power_law_1.2,0.2306879997253418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,1,2,power_law_1.2,0.23519999980926515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,1,2,power_law_1.2,0.24532480239868165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,1,2,power_law_1.2,0.24361600875854492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,1,2,power_law_1.2,0.2585088014602661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,1,2,power_law_1.2,0.2890624046325684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,2,power_law_1.01,1.1541824340820312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,1,2,power_law_1.2,0.3116735935211182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,1,2,power_law_1.2,0.3621567964553833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,1,2,power_law_1.2,0.39073278903961184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,1,2,power_law_1.2,0.46166400909423827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,1,2,power_law_1.2,0.577235221862793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,1,2,power_law_1.2,0.7576576232910156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,1,2,power_law_1.2,0.9576191902160645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,1,2,power_law_1.2,1.1453632354736327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,2,power_law_1.01,1.2600383758544922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,1,2,power_law_1.2,1.7892736434936523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,1,2,power_law_1.2,2.538163185119629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,1,2,power_law_1.2,3.5495872497558594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,2,power_law_1.01,1.5308799743652344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,1,2,power_law_1.2,5.4202880859375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,1,power_law_1.01,0.08977280259132385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,1,2,power_law_1.2,6.202521514892578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,1,power_law_1.01,0.09824640154838563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,2,power_law_1.01,1.648089599609375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,1,power_law_1.01,0.17825920581817628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,1,power_law_1.01,0.27615358829498293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,1,power_law_1.01,0.36412160396575927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,1,2,power_law_1.2,9.813689422607421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,1,power_law_1.01,0.6214144229888916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,2,power_law_1.01,2.0099008560180662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,1,power_law_1.01,0.6692863941192627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,1,power_law_1.01,0.6918208122253418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,1,power_law_1.01,0.7001664161682128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,1,power_law_1.01,0.7097536087036133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,1,power_law_1.01,0.7242112159729004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,2,power_law_1.01,2.5059839248657227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,1,power_law_1.01,0.7345536231994629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,1,2,power_law_1.2,20.8682373046875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,1,power_law_1.01,0.7439487934112549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,1,power_law_1.01,0.7709824085235596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,1,power_law_1.01,0.7781055927276611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,1,power_law_1.01,0.8071231842041016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,1,power_law_1.01,0.8050047874450683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,2,power_law_1.01,3.533599853515625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,1,power_law_1.01,0.9069696426391601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,1,power_law_1.01,0.9235903739929199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,1,power_law_1.01,1.029759979248047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,1,power_law_1.01,1.1463616371154786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,1,power_law_1.01,1.385587215423584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,2,power_law_1.01,3.847711944580078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,1,power_law_1.01,1.580851173400879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,1,power_law_1.01,2.0058303833007813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,1,power_law_1.01,2.442348861694336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,1,power_law_1.01,3.312249755859375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,2,2,power_law_1.01,4.679052734375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,32,power_law_1.01,0.036723199486732486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,32,power_law_1.01,0.028672000765800475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,32,power_law_1.01,0.028575998544692994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,1,power_law_1.01,4.213471984863281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,32,power_law_1.01,0.028857600688934327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,32,power_law_1.01,0.029977598786354066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,32,power_law_1.01,0.030720001459121703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,32,power_law_1.01,0.03134079873561859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,1,power_law_1.01,5.1171520233154295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,32,power_law_1.01,0.031116798520088196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,32,power_law_1.01,0.031539198756217954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,32,power_law_1.01,0.03198719918727875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,32,power_law_1.01,0.03258880078792572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,32,power_law_1.01,0.033983999490737916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,2,2,power_law_1.01,7.4036094665527346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,32,power_law_1.01,0.0347135990858078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,1,power_law_1.01,9.975308990478515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,32,power_law_1.01,0.036108800768852235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,32,power_law_1.01,0.038380798697471616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,32,power_law_1.01,0.038899201154708865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,32,power_law_1.01,0.03973119854927063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,32,power_law_1.01,0.04156160056591034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,32,power_law_1.01,0.04655359983444214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,32,power_law_1.01,0.051520001888275144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,32,power_law_1.01,0.06007680296897888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,32,power_law_1.01,0.07334399819374085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,32,power_law_1.01,0.09102720022201538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,32,power_law_1.01,0.11495039463043213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,1,power_law_1.01,20.17406005859375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,32,power_law_1.01,0.1407487988471985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,32,power_law_1.01,0.21068160533905028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,32,power_law_1.01,0.2753024101257324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,2,2,power_law_1.01,13.69054718017578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,32,power_law_1.01,0.36532480716705323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,32,power_law_1.01,0.47535362243652346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,32,power_law_1.01,1.0856191635131835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,16,balanced,0.043893332282702126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,16,balanced,0.04572799801826477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,16,balanced,0.04572266836961111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,16,balanced,0.047930667797724404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,16,balanced,0.048138668139775596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,16,balanced,0.054234668612480164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,16,balanced,0.07468266785144806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,16,balanced,0.07261866827805837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,16,balanced,0.07471466561158498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,16,balanced,0.07479466497898102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,16,balanced,0.07664533456166585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,16,balanced,0.07786666850248973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,16,balanced,0.07871999839941661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,16,balanced,0.08096533517042796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,16,balanced,0.08852266271909077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,16,balanced,0.08864532907803853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,16,balanced,0.09751466910044353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,16,balanced,0.10769066214561462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,16,balanced,0.11755733688672383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,16,balanced,0.1402400036652883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,16,balanced,0.1588053305943807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,16,balanced,0.20117332537968954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,16,balanced,0.25270400444666546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,16,balanced,0.33243733644485474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,16,balanced,0.4182933171590169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,16,balanced,0.5948266585667928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,16,balanced,0.7559786637624105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,16,balanced,0.9431146780649821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,16,balanced,1.4376266797383626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,2,4,balanced,0.09318400422732036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,2,4,balanced,0.10532266894976298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,2,4,balanced,0.13061333696047464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,2,4,balanced,0.1220853328704834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,2,4,balanced,0.12177600463231404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,16,balanced,2.8150507609049478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,2,4,balanced,0.12322133779525757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,2,4,balanced,0.12454400459925334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,2,4,balanced,0.12668266892433167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,2,4,balanced,0.1274186670780182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,2,4,balanced,0.12648533781369528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,2,4,balanced,0.13179733355840048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,2,4,balanced,0.13668800393740335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,2,4,balanced,0.13869866728782654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,2,4,balanced,0.13453333576520285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,2,4,balanced,0.15365866820017496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,2,4,balanced,0.15537066260973612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,2,4,balanced,0.14511467019716898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,2,4,balanced,0.18556799491246542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,2,4,balanced,0.1795039971669515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,2,4,balanced,0.23568000396092734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,2,4,balanced,0.2529866695404053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,2,4,balanced,0.3282613356908162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,2,4,balanced,0.3747680187225342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,2,4,balanced,0.5273973147074381
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,2,4,balanced,0.6901066303253174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,2,4,balanced,0.9781386852264404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,2,4,balanced,1.2772106329600017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,2,4,balanced,1.5775413513183594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,2,4,balanced,2.528144041697184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,2,4,balanced,4.981893221537272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,2,power_law_1.2,0.06024320125579834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,2,power_law_1.2,0.07311360239982605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,2,power_law_1.2,0.06847360134124755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,2,power_law_1.2,0.08698239922523499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,2,power_law_1.2,0.11345920562744141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,2,power_law_1.2,0.12595839500427247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,2,power_law_1.2,0.16906239986419677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,8,power_law_1.01,0.06214399933815003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,2,power_law_1.2,0.17884160280227662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,2,power_law_1.2,0.1879680037498474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,8,power_law_1.01,0.07109119892120361
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,2,power_law_1.2,0.18968960046768188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,2,power_law_1.2,0.19927040338516236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,8,power_law_1.01,0.055852800607681274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,2,power_law_1.2,0.21303679943084716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,2,power_law_1.2,0.2160640001296997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,8,power_law_1.01,0.057094401121139525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,2,power_law_1.2,0.22211201190948487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,8,power_law_1.01,0.06815360188484192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,2,power_law_1.2,0.22958080768585204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,2,power_law_1.2,0.24195199012756347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,8,power_law_1.01,0.07420160174369812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,2,power_law_1.2,0.25470719337463377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,2,power_law_1.2,0.28334081172943115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,8,power_law_1.01,0.08877440094947815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,2,power_law_1.2,0.3076224088668823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,8,power_law_1.01,0.08923519849777221
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,2,power_law_1.2,0.3510976076126099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,8,power_law_1.01,0.09427199959754944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,2,power_law_1.2,0.40836482048034667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,2,power_law_1.2,0.5143360137939453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,8,power_law_1.01,0.09448320269584656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,2,power_law_1.2,0.6503232002258301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,8,power_law_1.01,0.0955456018447876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,2,power_law_1.2,0.8685888290405274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,8,power_law_1.01,0.09713280200958252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,2,power_law_1.2,1.1412991523742675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,8,power_law_1.01,0.10060800313949585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,2,power_law_1.2,1.4712448120117188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,8,power_law_1.01,0.10201599597930908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,8,power_law_1.01,0.11011840105056762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,2,power_law_1.2,1.9158912658691407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,8,power_law_1.01,0.10913280248641968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,2,power_law_1.2,2.4476415634155275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,8,power_law_1.01,0.12126079797744752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,8,power_law_1.01,0.13549439907073973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,2,power_law_1.2,3.795801544189453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,8,power_law_1.01,0.13976320028305053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,8,power_law_1.01,0.16323839426040648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,8,power_law_1.01,0.19271039962768555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,2,power_law_1.2,7.099814605712891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,8,power_law_1.01,0.23379840850830078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,8,power_law_1.01,0.30824320316314696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,8,power_law_1.01,0.3680704116821289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,8,power_law_1.01,0.4615808010101318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,8,power_law_1.01,0.6619071960449219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,8,power_law_1.01,0.7654784202575684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,8,power_law_1.01,0.9574975967407227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,8,power_law_1.01,1.4704511642456055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,4,power_law_1.01,0.07052159905433655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,4,power_law_1.01,0.10453120470046998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,8,power_law_1.01,2.845427131652832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,4,power_law_1.01,0.14796160459518432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,4,power_law_1.01,0.0841152012348175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,4,power_law_1.01,0.12382080554962158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,4,power_law_1.01,0.1724544048309326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,4,power_law_1.01,0.1903807997703552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,4,power_law_1.01,0.2447360038757324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,4,power_law_1.01,0.2474816083908081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,4,power_law_1.01,0.24839038848876954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,4,power_law_1.01,0.26451199054718016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,4,power_law_1.01,0.281062388420105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,4,power_law_1.01,0.2871167898178101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,64,power_law_1.2,0.11642240285873413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,4,power_law_1.01,0.3038784027099609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,4,power_law_1.01,0.31402881145477296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,64,power_law_1.2,0.06421759724617004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,4,power_law_1.01,0.319321608543396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,4,power_law_1.01,0.32782080173492434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,64,power_law_1.2,0.0640447974205017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,4,power_law_1.01,0.3623487949371338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,4,power_law_1.01,0.3889024019241333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,64,power_law_1.2,0.051545602083206174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,4,power_law_1.01,0.4516160011291504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,64,power_law_1.2,0.05358719825744629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,4,power_law_1.01,0.4904319763183594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,4,power_law_1.01,0.613862419128418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,64,power_law_1.2,0.057683199644088745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,4,power_law_1.01,0.695136022567749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,64,power_law_1.2,0.057043200731277464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,4,power_law_1.01,0.9190848350524903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,64,power_law_1.2,0.06478719711303711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,4,power_law_1.01,1.0691967964172364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,64,power_law_1.2,0.06390399932861328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,4,power_law_1.01,1.4272319793701171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,64,power_law_1.2,0.060083198547363284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,4,power_law_1.01,1.789459228515625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,64,power_law_1.2,0.06280320286750793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,4,power_law_1.01,2.1198591232299804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,64,power_law_1.2,0.06635519862174988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,64,power_law_1.2,0.06589440107345582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,4,power_law_1.01,3.407398223876953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,64,power_law_1.2,0.07035520076751708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,64,power_law_1.2,0.07497599720954895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,4,power_law_1.01,6.190476989746093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,64,power_law_1.2,0.07772160172462464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,64,power_law_1.2,0.08165119886398316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,64,power_law_1.2,0.09259520173072815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,64,power_law_1.2,0.10206079483032227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,64,power_law_1.2,0.1260543942451477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,64,power_law_1.2,0.13393280506134034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,64,power_law_1.2,0.17816959619522094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,64,balanced,0.05229333539803823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,64,balanced,0.05654400090376536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,64,balanced,0.05836800237496694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,64,balanced,0.06659199794133504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,64,balanced,0.070592001080513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,64,balanced,0.06575466692447662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,64,balanced,0.06850666801134746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,64,balanced,0.07522133489449818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,64,balanced,0.07635200023651123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,64,balanced,0.07529599964618683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,64,balanced,0.0771679977575938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,64,balanced,0.07823466757933299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,64,balanced,0.07840533554553986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,64,balanced,0.07985066870848338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,64,balanced,0.08072000245253245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,64,balanced,0.08426133791605632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,64,balanced,0.09072533249855042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,64,balanced,0.10714133580525716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,64,balanced,0.11860799789428711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,64,power_law_1.2,0.2063744068145752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,64,balanced,0.13607999682426453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,64,balanced,0.15787733594576517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,64,balanced,0.2141973376274109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,64,balanced,0.21846934159596762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,64,balanced,0.28592000404993695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,64,balanced,0.33605865637461346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,64,balanced,0.4678560098012288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,64,balanced,0.5816959937413534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,2,64,balanced,0.6995733578999838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,2,64,balanced,1.0598986943562825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,64,power_law_1.2,0.2660223960876465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,2,64,balanced,2.2395359675089517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,64,power_law_1.2,0.3651456117630005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,64,power_law_1.2,0.5439231872558594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,64,power_law_1.2,0.6215424060821533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,64,power_law_1.2,0.7869887828826905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,64,power_law_1.2,1.1495231628417968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,4,power_law_1.01,0.03529599905014038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,4,power_law_1.01,0.04110080003738403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,4,power_law_1.01,0.03640959858894348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,4,power_law_1.01,0.04431360065937042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,4,power_law_1.01,0.055289602279663085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,4,power_law_1.01,0.06629760265350342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,64,power_law_1.2,2.5860544204711915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,4,power_law_1.01,0.0711296021938324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,4,power_law_1.01,0.07386239767074584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,4,power_law_1.01,0.07525759935379028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,4,power_law_1.01,0.07628160119056701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,4,power_law_1.01,0.0759552001953125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,4,power_law_1.01,0.07741439938545228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,4,power_law_1.01,0.0785535991191864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,4,power_law_1.01,0.08099200129508972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,4,power_law_1.01,0.08624640107154846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,4,power_law_1.01,0.08915200233459472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,4,power_law_1.01,0.09528319835662842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,4,power_law_1.01,0.10959999561309815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,4,power_law_1.01,0.11164799928665162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,4,power_law_1.01,0.12967679500579835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,4,power_law_1.01,0.13766399621963502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,4,power_law_1.01,0.1704767942428589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,4,power_law_1.01,0.1898751974105835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,4,power_law_1.01,0.23470079898834229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,4,power_law_1.01,0.27689599990844727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,4,power_law_1.01,0.3899967908859253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,1,balanced,0.05144000053405762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,1,balanced,0.0643093337615331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,1,balanced,0.0825439989566803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,4,power_law_1.01,0.5042687892913819
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,1,balanced,0.11463466286659241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,1,balanced,0.17540266116460165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,1,balanced,0.20655999581019083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,1,balanced,0.21258666117986044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,1,balanced,0.21267199516296387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,1,balanced,0.2139093279838562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,1,balanced,0.21843733390172324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,1,balanced,0.22379199663798013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,1,4,power_law_1.01,0.6094399929046631
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,1,balanced,0.23016534248987833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,1,balanced,0.23037334283192953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,1,balanced,0.240064005057017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,1,balanced,0.24755734205245972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,1,balanced,0.2564319968223572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,1,balanced,0.27637332677841187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,1,balanced,0.3132586677869161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,1,balanced,0.3365279833475749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,1,balanced,0.44915199279785156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,1,4,power_law_1.01,0.9032896041870118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,1,balanced,0.48185598850250244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,1,balanced,0.6981333096822103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,1,balanced,0.794426679611206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,1,balanced,1.2635306517283122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,1,balanced,1.440890630086263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,1,balanced,2.177589257558187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,1,4,power_law_1.01,1.5941760063171386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,1,balanced,2.7257652282714844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,1,balanced,3.2491785685221353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,1,balanced,5.155669212341309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,1,balanced,9.99451764424642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,64,power_law_1.01,0.049318400025367734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,64,power_law_1.01,0.030649599432945252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,4,2,power_law_1.2,0.05541120171546936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,4,2,power_law_1.2,0.05400320291519165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,64,power_law_1.01,0.02999039888381958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,4,2,power_law_1.2,0.05501440167427063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,64,power_law_1.01,0.029023998975753786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,4,2,power_law_1.2,0.07471359968185425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,64,power_law_1.01,0.02968960106372833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,4,2,power_law_1.2,0.07797759771347046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,64,power_law_1.01,0.030368000268936157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,4,2,power_law_1.2,0.07907840013504028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,4,2,power_law_1.2,0.08310400247573853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,64,power_law_1.01,0.03080959916114807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,4,2,power_law_1.2,0.08094720244407654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,64,power_law_1.01,0.031225600838661195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,4,2,power_law_1.2,0.08290560245513916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,4,2,power_law_1.2,0.08455039858818054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,64,power_law_1.01,0.030803200602531434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,4,2,power_law_1.2,0.08778240084648133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,64,power_law_1.01,0.031513598561286923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,4,2,power_law_1.2,0.08857600092887878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,4,2,power_law_1.2,0.08895360231399536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,64,power_law_1.01,0.03160960078239441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,4,2,power_law_1.2,0.09843199849128723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,64,power_law_1.01,0.03240320086479187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,4,2,power_law_1.2,0.1094655990600586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,4,2,power_law_1.2,0.11166080236434936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,64,power_law_1.01,0.03346560001373291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,4,2,power_law_1.2,0.1278656005859375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,64,power_law_1.01,0.03424000144004822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,4,2,power_law_1.2,0.1589311957359314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,4,2,power_law_1.2,0.16193920373916626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,64,power_law_1.01,0.037676799297332766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,4,2,power_law_1.2,0.20548479557037352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,64,power_law_1.01,0.038764798641204835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,4,2,power_law_1.2,0.26019198894500734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,64,power_law_1.01,0.0392767995595932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,4,2,power_law_1.2,0.3552383899688721
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,4,2,power_law_1.2,0.4226240158081055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,64,power_law_1.01,0.04312959909439087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,4,2,power_law_1.2,0.5786367893218994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,64,power_law_1.01,0.045542401075363156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,4,2,power_law_1.2,0.670739221572876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,64,power_law_1.01,0.05086719989776611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,4,2,power_law_1.2,1.028927993774414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,64,power_law_1.01,0.055340802669525145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,4,2,power_law_1.2,1.296268844604492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,64,power_law_1.01,0.06928640007972717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,4,2,power_law_1.2,1.808799934387207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,64,power_law_1.01,0.089055997133255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,64,power_law_1.01,0.11407999992370606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,4,2,power_law_1.2,2.5438528060913086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,64,power_law_1.01,0.14549119472503663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,64,power_law_1.01,0.20286080837249756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,4,2,power_law_1.2,6.0708160400390625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,64,power_law_1.01,0.24785280227661133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,1,64,power_law_1.01,0.2887808084487915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,1,64,power_law_1.01,0.41478400230407714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,1,64,power_law_1.01,0.8214400291442872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,8,32,balanced,0.0524479995171229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,8,32,balanced,0.04863999783992767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,8,32,balanced,0.04834666848182678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,8,32,balanced,0.048058668772379555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,8,32,balanced,0.05198400219281515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,8,32,balanced,0.0514933317899704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,8,32,balanced,0.05203733344872793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,8,32,balanced,0.05399466554323832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,8,32,balanced,0.054101333022117615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,8,32,balanced,0.05385066568851471
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,8,32,balanced,0.05589866638183594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,8,32,balanced,0.05622933308283488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,8,32,balanced,0.0543093333641688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,8,32,balanced,0.05823466678460439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,8,32,balanced,0.06029866635799408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,8,32,balanced,0.060677334666252136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,8,32,balanced,0.06463466584682465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,8,32,balanced,0.06664533416430156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,8,32,balanced,0.07220800220966339
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,8,32,balanced,0.08488000432650249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,8,32,balanced,0.08903466661771138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,8,32,balanced,0.11575999855995178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,8,32,balanced,0.13803733388582864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,8,32,balanced,0.17517866690953574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,8,32,balanced,0.20397865772247314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,8,32,balanced,0.28753066062927246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,8,32,balanced,0.3487519820531209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,8,32,balanced,0.41380266348520917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,8,32,balanced,0.62336532274882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,8,32,balanced,1.2034080028533936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,16,balanced,0.05231999854246775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,16,balanced,0.05842666824658712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,16,balanced,0.07640533149242401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,16,balanced,0.1011306643486023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,16,balanced,0.156442662080129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,16,balanced,0.18465600411097208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,16,balanced,0.18473066886266074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,16,balanced,0.18531199296315512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,16,balanced,0.18619734048843384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,16,balanced,0.18363199631373087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,16,balanced,0.1861120065053304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,16,balanced,0.1855413317680359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,16,balanced,0.1848106582959493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,16,balanced,0.18556799491246542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,16,balanced,0.19135467211405435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,16,balanced,0.19088532527287802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,16,balanced,0.19243200620015463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,16,balanced,0.21767999728520712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,16,balanced,0.20681599775950113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,16,balanced,0.2405280073483785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,16,balanced,0.2316960096359253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,16,balanced,0.3040906588236491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,16,balanced,0.2893226742744446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,16,balanced,0.4316106637318929
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,16,balanced,0.4301919937133789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,16,balanced,0.6848800182342529
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,16,balanced,0.7060693105061849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,16,balanced,0.7898453076680502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,16,balanced,1.2350239753723145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,16,balanced,2.3022400538126626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,2,2,balanced,0.05877333382765452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,2,2,balanced,0.08097066481908162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,2,2,balanced,0.1218346655368805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,2,2,balanced,0.12396267056465149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,2,2,balanced,0.12425067027409871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,2,2,balanced,0.12387200196584065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,2,2,balanced,0.12552533547083536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,2,2,balanced,0.12516799569129944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,2,2,balanced,0.12572800119717917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,2,2,balanced,0.12600533167521158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,2,2,balanced,0.13149333000183105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,2,2,balanced,0.12785599629084268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,2,2,balanced,0.12971199552218118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,2,2,balanced,0.1381119986375173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,2,2,balanced,0.15154133240381876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,2,2,balanced,0.15357333421707153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,2,2,balanced,0.1574613352616628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,2,2,balanced,0.18875734011332193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,2,2,balanced,0.19615467389424643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,2,2,balanced,0.2409493327140808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,2,2,balanced,0.2730240027109782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,2,2,balanced,0.3734240134557088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,2,2,balanced,0.4660160144170125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,2,2,balanced,0.6502079963684082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,2,2,balanced,0.831658681233724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,2,2,balanced,1.2234079837799072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,2,2,balanced,1.6079200108846028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,2,2,balanced,1.999637285868327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,2,2,balanced,3.141146659851074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,2,2,balanced,6.15115229288737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,32,power_law_1.01,0.057792001962661745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,32,power_law_1.01,0.052160000801086424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,32,power_law_1.01,0.04975999891757965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,32,power_law_1.01,0.04953599870204926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,32,power_law_1.01,0.05057280063629151
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,32,power_law_1.01,0.05203840136528015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,32,power_law_1.01,0.05310080051422119
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,32,power_law_1.01,0.054016000032424925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,32,power_law_1.01,0.05370879769325256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,32,power_law_1.01,0.054771202802658084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,32,power_law_1.01,0.055315202474594115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,32,power_law_1.01,0.05699840188026428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,32,power_law_1.01,0.0573248028755188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,32,power_law_1.01,0.05856639742851257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,32,power_law_1.01,0.06450560092926025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,32,power_law_1.01,0.06572160124778748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,32,power_law_1.01,0.07022719979286193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,32,power_law_1.01,0.08065279722213745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,32,power_law_1.01,0.08803840279579163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,32,power_law_1.01,0.10686719417572021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,32,power_law_1.01,0.11594239473342896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,32,power_law_1.01,0.15508480072021485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,32,power_law_1.01,0.18097280263900756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,32,power_law_1.01,0.25019519329071044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,32,power_law_1.01,0.30364160537719725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,32,power_law_1.01,0.43704957962036134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,32,power_law_1.01,0.5584320068359375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,32,power_law_1.01,0.6605567932128906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,32,power_law_1.01,1.1460351943969727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,32,power_law_1.01,2.1362367630004884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,2,power_law_1.2,0.04544639885425568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,2,power_law_1.2,0.04954879879951477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,2,power_law_1.2,0.05273600220680237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,2,power_law_1.2,0.059001600742340087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,2,power_law_1.2,0.06583679914474487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,2,power_law_1.2,0.09125120043754578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,2,power_law_1.2,0.0958400011062622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,2,power_law_1.2,0.09642239809036254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,2,power_law_1.2,0.10204160213470459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,2,power_law_1.2,0.1049280047416687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,2,power_law_1.2,0.1089087963104248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,2,power_law_1.2,0.11458560228347778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,2,power_law_1.2,0.11581439971923828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,2,power_law_1.2,0.12244479656219483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,2,power_law_1.2,0.1317055940628052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,2,power_law_1.2,0.1380288004875183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,2,power_law_1.2,0.14954240322113038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,2,power_law_1.2,0.17475839853286743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,2,power_law_1.2,0.19612159729003906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,2,power_law_1.2,0.2486272096633911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,2,power_law_1.2,0.31794559955596924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,2,power_law_1.2,0.4315904140472412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,2,power_law_1.2,0.5328896045684814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,2,power_law_1.2,0.7380032062530517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,2,power_law_1.2,0.9494720458984375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,2,power_law_1.2,1.355827236175537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,2,power_law_1.2,1.7095487594604493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,2,power_law_1.2,2.0550079345703125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,2,power_law_1.2,3.242700958251953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,2,power_law_1.2,6.675961303710937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,2,balanced,0.10107733805974324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,2,balanced,0.2412373423576355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,2,balanced,0.25597333908081055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,2,balanced,0.46856534481048584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,2,balanced,0.8850826422373453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,2,balanced,1.0992000102996826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,2,balanced,1.101685365041097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,2,balanced,1.1013867060343425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,2,balanced,1.1072906653086345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,2,balanced,1.107632001241048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,2,balanced,1.1087946891784668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,2,balanced,1.1162880261739094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,2,balanced,1.1162772973378499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,2,balanced,1.1269973119099934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,2,balanced,1.1413386662801106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,2,balanced,1.1478933493296306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,2,balanced,1.1572319666544597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,2,balanced,1.235317309697469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,2,balanced,1.2067306836446126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,2,balanced,1.5311039288838704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,2,balanced,1.3195573488871257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,2,balanced,2.0293919245402017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,2,balanced,1.5578719774882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,2,balanced,2.5363465944925943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,2,balanced,2.308677355448405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,2,balanced,3.1858507792154946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,2,balanced,4.392592112223308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,2,balanced,4.836949348449707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,2,balanced,7.470879872639974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,2,balanced,13.827754974365234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,2,power_law_1.01,0.06025599837303162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,2,power_law_1.01,0.06620799899101257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,2,power_law_1.01,0.07022079825401306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,2,power_law_1.01,0.08135679960250855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,2,power_law_1.01,0.10111360549926758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,2,power_law_1.01,0.12037760019302368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,4,power_law_1.2,0.056403201818466184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,2,power_law_1.01,0.14755840301513673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,4,power_law_1.2,0.06227200031280518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,2,power_law_1.01,0.1551103949546814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,4,power_law_1.2,0.05633280277252197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,2,power_law_1.01,0.15375360250473022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,4,power_law_1.2,0.06283519864082336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,2,power_law_1.01,0.15973119735717772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,4,power_law_1.2,0.06943359971046448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,2,power_law_1.01,0.16619520187377929
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,4,power_law_1.2,0.07488639950752259
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,2,power_law_1.01,0.16597119569778443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,4,power_law_1.2,0.08807039856910706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,2,power_law_1.01,0.17221759557723998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,4,power_law_1.2,0.09160320162773132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,2,power_law_1.01,0.1794111967086792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,2,power_law_1.01,0.1862272024154663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,4,power_law_1.2,0.09358720183372497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,2,power_law_1.01,0.1958400011062622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,4,power_law_1.2,0.09743360280990601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,2,power_law_1.01,0.20523519515991212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,4,power_law_1.2,0.09947519898414611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,2,power_law_1.01,0.22417919635772704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,4,power_law_1.2,0.10136959552764893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,4,power_law_1.2,0.1050495982170105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,2,power_law_1.01,0.2503040075302124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,4,power_law_1.2,0.10920959711074829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,2,power_law_1.01,0.2962752103805542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,4,power_law_1.2,0.12168960571289063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,2,power_law_1.01,0.35376639366149903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,4,power_law_1.2,0.12736639976501465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,2,power_law_1.01,0.4557375907897949
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,4,power_law_1.2,0.1280959963798523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,2,power_law_1.01,0.5497024059295654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,4,power_law_1.2,0.14384640455245973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,2,power_law_1.01,0.7635200023651123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,4,power_law_1.2,0.16657279729843139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,4,power_law_1.2,0.19304319620132446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,2,power_law_1.01,0.95217924118042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,4,power_law_1.2,0.23738880157470704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,2,power_law_1.01,1.3239616394042968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,4,power_law_1.2,0.29011199474334715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,4,power_law_1.2,0.36978559494018554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,2,power_law_1.01,1.812460708618164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,4,power_law_1.2,0.4959616184234619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,2,power_law_1.01,2.202796745300293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,4,power_law_1.2,0.6661312103271484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,4,power_law_1.2,0.9696255683898926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,2,power_law_1.01,3.2950592041015625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,4,power_law_1.2,1.151193618774414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,4,power_law_1.2,1.530726432800293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,2,power_law_1.01,6.518605041503906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,32,1,balanced,0.029450667401154835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,32,1,balanced,0.03336533407370249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,32,1,balanced,0.03570133447647095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,32,1,balanced,0.03585600107908249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,32,1,balanced,0.03356266766786575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,32,1,balanced,0.04188266893227895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,32,1,balanced,0.04165866722663244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,32,1,balanced,0.036415999134381614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,32,1,balanced,0.038133333126703896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,32,1,balanced,0.03651199986537298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,4,power_law_1.2,2.3708927154541017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,32,1,balanced,0.03973866750796636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,32,1,balanced,0.04062933226426443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,32,1,balanced,0.04163199911514918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,32,1,balanced,0.043247997760772705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,32,1,balanced,0.053823997577031456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,32,1,balanced,0.05604266623655955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,32,1,balanced,0.061834668119748436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,32,1,balanced,0.07499200105667114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,32,1,balanced,0.07738666733105977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,32,1,balanced,0.09893332918485005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,32,1,balanced,0.10734933614730835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,32,1,balanced,0.14698132872581482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,32,1,balanced,0.16028799613316855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,32,1,balanced,0.223797341187795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,32,1,balanced,0.28510934114456177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,32,1,balanced,0.4007413387298584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,32,1,balanced,0.5052853425343832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,32,1,balanced,0.627455989519755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,32,1,balanced,0.9600160121917725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,32,1,balanced,1.847647984822591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,4,power_law_1.2,4.469350433349609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,2,balanced,0.05021866659323374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,2,balanced,0.05286400020122528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,2,balanced,0.055546666185061135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,2,balanced,0.0637066662311554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,2,balanced,0.0936853289604187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,2,balanced,0.13205867012341818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,2,balanced,0.16723199685414633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,2,balanced,0.17082132895787558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,2,balanced,0.17100266615549722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,2,balanced,0.17261866728464761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,2,balanced,0.17475199699401855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,2,balanced,0.18058133125305176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,2,balanced,0.18086934089660645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,2,balanced,0.18742932875951132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,2,balanced,0.19603200753529867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,2,balanced,0.19814399878184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,2,balanced,0.21690134207407633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,2,balanced,0.24064000447591147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,2,balanced,0.26255999008814496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,2,balanced,0.3123679955800374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,2,balanced,0.3518079916636149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,2,balanced,0.44969598452250165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,2,balanced,0.5633440017700195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,2,balanced,0.748032013575236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,2,balanced,0.9991093476613363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,2,balanced,1.4191093444824219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,2,balanced,1.830992062886556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,2,balanced,2.2847253481547036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,32,power_law_1.2,0.15921920537948608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,2,16,balanced,0.023285334308942158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,2,16,balanced,0.023503998915354412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,2,16,balanced,0.024101334313551586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,2,balanced,3.492762565612793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,2,16,balanced,0.02359466751416524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,2,16,balanced,0.027429332335789997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,2,16,balanced,0.027290667096773785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,2,16,balanced,0.027295999228954315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,2,16,balanced,0.031557333966096245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,2,16,balanced,0.03081600119670232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,2,16,balanced,0.03129599988460541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,2,16,balanced,0.031557333966096245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,2,16,balanced,0.03143466760714849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,32,power_law_1.2,0.14304640293121337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,2,16,balanced,0.03352533280849457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,2,16,balanced,0.033546666304270424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,2,16,balanced,0.03551466763019562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,2,16,balanced,0.037674665451049805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,2,16,balanced,0.037402667105197906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,2,16,balanced,0.037962667644023895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,16,balanced,0.03950933367013931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,16,balanced,0.04197333256403605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,16,balanced,0.04590400060017904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,16,balanced,0.0499839981396993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,16,balanced,0.05715199808279673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,16,balanced,0.07243200143178304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,16,balanced,0.08876799543698628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,16,balanced,0.10896533727645874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,2,balanced,6.7995147705078125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,16,balanced,0.13006933530171713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,16,balanced,0.15486933787663779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,32,power_law_1.2,0.07199360132217407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,16,balanced,0.21148266394933066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,16,balanced,0.3813013235727946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,32,power_law_1.2,0.08577920198440551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,32,power_law_1.2,0.11330560445785523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,32,power_law_1.2,0.11749119758605957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,32,power_law_1.2,0.12663040161132813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,32,power_law_1.2,0.13251199722290039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,32,power_law_1.2,0.1352255940437317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,32,power_law_1.2,0.1360383987426758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,32,power_law_1.2,0.14568959474563598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,32,power_law_1.2,0.14407680034637452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,32,power_law_1.2,0.15008000135421753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,32,power_law_1.2,0.1540992021560669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,32,power_law_1.2,0.1692031979560852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,32,power_law_1.2,0.18113919496536254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,4,4,balanced,0.04166933397452036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,4,4,balanced,0.039594667653242745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,4,4,balanced,0.04364266494909922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,4,4,balanced,0.04308799902598063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,4,4,balanced,0.04372799893220266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,4,4,balanced,0.043605332573254905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,1,balanced,0.09913067022959392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,4,4,balanced,0.0446720023949941
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,4,4,balanced,0.04577599962552389
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,1,balanced,0.18841065963109335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,4,4,balanced,0.043621331453323364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,4,4,balanced,0.04363733530044556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,1,balanced,0.246288001537323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,4,4,balanced,0.04580800235271454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,4,4,balanced,0.0459199994802475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,1,balanced,0.44120534261067706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,4,4,balanced,0.04597333570321401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,4,4,balanced,0.04996799925963084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,4,4,balanced,0.06006933252016703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,32,power_law_1.2,0.17638399600982665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,4,4,balanced,0.0605973352988561
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,1,balanced,0.8331039746602377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,4,4,balanced,0.06634666522343953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,4,4,balanced,0.07689066727956136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,4,4,balanced,0.08165866633256276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,4,4,balanced,0.0969546635945638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,4,4,balanced,0.10419199864069621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,1,balanced,1.6179893811543782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,4,4,balanced,0.12974933783213297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,4,4,balanced,0.14685866236686707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,4,4,balanced,0.18981866041819254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,4,4,balanced,0.24197334051132202
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,1,balanced,2.4054133097330728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,4,4,balanced,0.329312006632487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,32,power_law_1.2,0.20533759593963624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,4,4,balanced,0.41225067774454754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,4,4,balanced,0.5017760197321574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,1,balanced,2.411189397176107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,4,4,balanced,0.7627306779225668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,4,4,balanced,1.4700907071431477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,1,balanced,2.418890635172526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,32,power_law_1.2,0.1974079966545105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,1,balanced,2.4207679430643716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,1,balanced,2.4285546938578286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,32,power_law_1.2,0.2390592098236084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,1,balanced,2.44378662109375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,1,balanced,2.4538826942443848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,1,balanced,2.454469362894694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,32,power_law_1.2,0.24147839546203614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,1,balanced,2.4631519317626953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,1,balanced,2.475813388824463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,32,power_law_1.2,0.32004480361938475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,1,balanced,2.501525402069092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,1,balanced,2.556719938913981
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,32,power_law_1.2,0.375654411315918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,1,balanced,2.604431947072347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,1,balanced,2.7159945170084634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,1,balanced,2.807162602742513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,32,power_law_1.2,0.5143936157226563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,1,balanced,3.024085362752279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,1,balanced,3.173839886983236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,32,power_law_1.2,0.6620800018310546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,1,balanced,3.577695846557617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,1,balanced,4.033520062764485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,32,power_law_1.2,0.9030336380004883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,1,balanced,4.648325284322103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,1,balanced,6.0513121287028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,32,power_law_1.2,1.3513792037963868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,1,balanced,7.546933492024739
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,32,balanced,0.0441599984963735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,32,balanced,0.04409599800904592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,32,balanced,0.04436799883842468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,32,balanced,0.04608533283074697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,32,balanced,0.04976533353328705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,32,balanced,0.05231466889381409
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,32,power_law_1.2,1.5113280296325684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,32,balanced,0.05266666909058889
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,32,balanced,0.05439466734727224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,32,balanced,0.05431999762852987
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,32,balanced,0.05425600210825602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,32,balanced,0.05551999807357788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,32,balanced,0.057445332407951355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,32,balanced,0.05909333129723867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,32,balanced,0.06001066664854685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,32,balanced,0.06457066535949707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,32,balanced,0.06739733119805653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,32,balanced,0.0707893321911494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,32,balanced,0.08099199831485748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,32,balanced,0.09522666533788045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,32,balanced,0.1197706659634908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,32,balanced,0.14220266540845236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,32,balanced,0.17883733908335367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,32,balanced,0.21546665827433267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,1,balanced,10.53925895690918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,32,balanced,0.2922293345133464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,32,balanced,0.36558401584625244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,32,balanced,0.5238666534423828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,32,balanced,0.669871966044108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,32,balanced,0.825055996576945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,32,power_law_1.2,2.0984895706176756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,32,balanced,1.2666026751200359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,32,balanced,2.486629327138265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,1,balanced,19.94557825724284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,32,power_law_1.2,4.224249649047851
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,32,1,power_law_1.2,0.03255679905414581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,32,1,power_law_1.2,0.03432320058345795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,32,1,power_law_1.2,0.038796800374984744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,32,1,power_law_1.2,0.04319359958171844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,32,1,power_law_1.2,0.05300480127334595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,128,power_law_1.01,0.033855998516082765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,32,1,power_law_1.2,0.07470719814300537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,128,power_law_1.01,0.03361279964447021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,32,1,power_law_1.2,0.07801600098609925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,8,power_law_1.01,0.10220160484313964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,128,power_law_1.01,0.03369599878787995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,32,1,power_law_1.2,0.07996799945831298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,128,power_law_1.01,0.033504000306129454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,8,power_law_1.01,0.14986239671707152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,32,1,power_law_1.2,0.08261759877204895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,128,power_law_1.01,0.03418880105018616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,8,power_law_1.01,0.07648640275001525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,32,1,power_law_1.2,0.08321920037269592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,128,power_law_1.01,0.0353408008813858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,8,power_law_1.01,0.09618560075759888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,32,1,power_law_1.2,0.08789759874343872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,128,power_law_1.01,0.03521279990673065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,8,power_law_1.01,0.11605119705200195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,128,power_law_1.01,0.03609600067138672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,32,1,power_law_1.2,0.09025279879570007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,128,power_law_1.01,0.035174399614334106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,32,1,power_law_1.2,0.0948415994644165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,8,power_law_1.01,0.15519360303878785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,128,power_law_1.01,0.035545599460601804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,32,1,power_law_1.2,0.1026304006576538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,8,power_law_1.01,0.165721595287323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,128,power_law_1.01,0.03612799942493439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,32,1,power_law_1.2,0.11014399528503419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,8,power_law_1.01,0.15872639417648315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,128,power_law_1.01,0.03616639971733093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,32,1,power_law_1.2,0.1146623969078064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,8,power_law_1.01,0.1605504035949707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,128,power_law_1.01,0.036985599994659425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,32,1,power_law_1.2,0.12170239686965942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,128,power_law_1.01,0.03780480027198792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,8,power_law_1.01,0.1673408031463623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,32,1,power_law_1.2,0.14421759843826293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,128,power_law_1.01,0.039673599600791934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,32,1,power_law_1.2,0.16946560144424438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,8,power_law_1.01,0.16748160123825073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,128,power_law_1.01,0.039961600303649904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,32,1,power_law_1.2,0.20870399475097656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,8,power_law_1.01,0.1683135986328125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,128,power_law_1.01,0.04120959937572479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,32,1,power_law_1.2,0.2564671993255615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,8,power_law_1.01,0.17738879919052125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,128,power_law_1.01,0.04373759925365448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,32,1,power_law_1.2,0.35527040958404543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,8,power_law_1.01,0.18561919927597045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,128,power_law_1.01,0.04498560130596161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,128,power_law_1.01,0.04829440116882324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,32,1,power_law_1.2,0.4150720119476318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,8,power_law_1.01,0.18842240571975707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,128,power_law_1.01,0.053529602289199826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,8,power_law_1.01,0.20157439708709718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,32,1,power_law_1.2,0.5859776020050049
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,128,power_law_1.01,0.06225280165672302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,8,power_law_1.01,0.20466558933258056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,32,1,power_law_1.2,0.7433472156524659
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,128,power_law_1.01,0.07259520292282104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,8,power_law_1.01,0.2340480089187622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,128,power_law_1.01,0.08965759873390197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,32,1,power_law_1.2,1.0621376037597656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,8,power_law_1.01,0.2407360076904297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,128,power_law_1.01,0.10957440137863159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,8,power_law_1.01,0.30208001136779783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,32,1,power_law_1.2,1.392959976196289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,128,power_law_1.01,0.15960320234298705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,8,power_law_1.01,0.2951616048812866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,8,power_law_1.01,0.3697792053222656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,128,power_law_1.01,0.2053056001663208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,32,1,power_law_1.2,1.732454490661621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,8,power_law_1.01,0.418393611907959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,128,power_law_1.01,0.23440639972686766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,8,power_law_1.01,0.49802241325378416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,32,1,power_law_1.2,2.7113855361938475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,8,power_law_1.01,0.6623936176300049
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,128,power_law_1.01,0.3873280048370361
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,8,power_law_1.01,0.9602239608764649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,8,power_law_1.01,1.066579246520996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,128,power_law_1.01,0.7262976169586182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,32,1,power_law_1.2,5.246726226806641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,8,power_law_1.01,1.4871423721313477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,8,power_law_1.01,1.9560575485229492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,8,power_law_1.01,4.250105667114258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,32,balanced,0.045312002301216125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,32,balanced,0.04394666850566864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,2,power_law_1.2,0.04085119962692261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,32,balanced,0.043807998299598694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,32,balanced,0.04599999884764353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,32,balanced,0.049312000473340355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,32,balanced,0.07375466823577881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,32,balanced,0.0745119998852412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,32,balanced,0.07492800056934357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,32,balanced,0.07719466586907704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,32,balanced,0.07666133344173431
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,32,balanced,0.07689066727956136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,2,power_law_1.2,0.03829759955406189
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,32,balanced,0.0798826664686203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,32,balanced,0.08058133224646251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,32,balanced,0.08131733536720276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,32,balanced,0.0892693301041921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,32,balanced,0.08945066730181377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,32,balanced,0.09291733304659526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,32,balanced,0.10336533188819885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,2,power_law_1.2,0.040031999349594116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,32,balanced,0.10891733566919963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,32,balanced,0.12667199969291687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,32,balanced,0.1433013379573822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,32,balanced,0.17568000157674155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,32,balanced,0.2097919980684916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,32,balanced,0.2768639922142029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,2,power_law_1.2,0.043647998571395875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,32,balanced,0.3405066728591919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,32,balanced,0.47543466091156006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,32,balanced,0.6095199982325236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,2,power_law_1.2,0.04572800099849701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,32,balanced,0.7449759642283121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,32,balanced,1.1216426690419514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,2,power_law_1.2,0.057894402742385866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,32,balanced,2.186746597290039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,2,power_law_1.2,0.060140800476074216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,2,power_law_1.2,0.059987199306488034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,2,power_law_1.2,0.0615231990814209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,2,power_law_1.2,0.06310399770736694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,2,power_law_1.2,0.06410239934921265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,16,1,power_law_1.2,0.039417600631713866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,16,1,power_law_1.2,0.05093119740486145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,16,1,power_law_1.2,0.05690240263938904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,2,power_law_1.2,0.0692799985408783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,16,1,power_law_1.2,0.07242239713668823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,16,1,power_law_1.2,0.07166079878807068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,2,power_law_1.2,0.07239680290222168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,16,1,power_law_1.2,0.07411199808120728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,16,1,power_law_1.2,0.07612800002098083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,16,1,power_law_1.2,0.07701119780540466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,2,power_law_1.2,0.07141119837760926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,16,1,power_law_1.2,0.07877119779586791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,16,1,power_law_1.2,0.08008319735527039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,16,1,power_law_1.2,0.08006399869918823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,2,power_law_1.2,0.07763199806213379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,16,1,power_law_1.2,0.08266879916191101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,16,1,power_law_1.2,0.08523520231246948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,16,1,power_law_1.2,0.09181439876556396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,2,power_law_1.2,0.08232319951057435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,16,1,power_law_1.2,0.10185600519180298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,16,1,power_law_1.2,0.10269440412521362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,2,power_law_1.2,0.08886399865150452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,16,1,power_law_1.2,0.10705920457839965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,16,1,power_law_1.2,0.133951997756958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,16,1,power_law_1.2,0.13463040590286254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,2,power_law_1.2,0.10464639663696289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,16,1,power_law_1.2,0.16821759939193726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,16,1,power_law_1.2,0.1916607975959778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,2,power_law_1.2,0.12314239740371705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,16,1,power_law_1.2,0.2517695903778076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,16,1,power_law_1.2,0.29005439281463624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,2,power_law_1.2,0.14156800508499146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,16,1,power_law_1.2,0.3989824056625366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,16,1,power_law_1.2,0.5127552032470704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,2,power_law_1.2,0.17764480113983155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,16,1,power_law_1.2,0.7115456104278565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,2,power_law_1.2,0.25662078857421877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,16,1,power_law_1.2,0.9126720428466797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,16,1,power_law_1.2,1.1338111877441406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,2,power_law_1.2,0.2837503910064697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,16,1,power_law_1.2,1.7564607620239259
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,2,power_law_1.2,0.4279808044433594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,2,power_law_1.2,0.5429120063781738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,16,1,power_law_1.2,3.4304576873779298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,2,power_law_1.2,0.8510720252990722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,2,power_law_1.2,1.0268223762512207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,2,power_law_1.2,1.2161791801452637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,2,power_law_1.2,2.197907257080078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,2,power_law_1.2,4.335731124877929
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,32,1,power_law_1.01,0.05161600112915039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,32,1,power_law_1.01,0.055078399181365964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,32,1,power_law_1.01,0.058483201265335086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,32,1,power_law_1.01,0.0725055992603302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,32,1,power_law_1.01,0.10806399583816528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,32,1,power_law_1.01,0.12974079847335815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,4,balanced,0.04808000226815542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,4,balanced,0.05046933392683665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,4,balanced,0.07398400207360585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,4,balanced,0.09774933258692424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,4,balanced,0.15043200055758157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,4,balanced,0.24785067637761435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,4,balanced,0.3449759880701701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,4,balanced,0.3452746470769246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,4,balanced,0.34883201122283936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,4,balanced,0.3492906490961711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,4,balanced,0.35258134206136066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,4,balanced,0.3590826590855916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,32,1,power_law_1.01,0.14554879665374756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,4,balanced,0.36179200808207196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,4,balanced,0.3659573396046956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,4,balanced,0.3705759843190511
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,4,balanced,0.3731946547826131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,4,balanced,0.38632531960805255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,4,balanced,0.40516801675160724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,4,balanced,0.4210826555887858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,4,balanced,0.45629334449768066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,4,balanced,0.486191987991333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,4,balanced,0.5676746765772501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,4,balanced,0.6342026789983114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,32,1,power_law_1.01,0.17017600536346436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,4,balanced,0.7733120123545328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,4,balanced,0.942255973815918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,4,balanced,1.2456639607747395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,4,balanced,1.5739572842915852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,32,1,power_law_1.01,0.17831679582595825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,4,balanced,1.9661547342936199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,4,balanced,2.878357251485189
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,32,1,power_law_1.01,0.1859455943107605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,4,balanced,5.5655517578125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,32,1,power_law_1.01,0.19256319999694824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,32,1,power_law_1.01,0.2070080041885376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,32,1,power_law_1.01,0.21839361190795897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,32,1,power_law_1.01,0.23276801109313966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,32,1,power_law_1.01,0.2563136100769043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,32,1,power_law_1.01,0.2758656024932861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,32,1,power_law_1.01,0.30538880825042725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,32,1,power_law_1.01,0.3809407949447632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,2,8,balanced,0.05762133498986562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,32,1,power_law_1.01,0.3689280033111572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,2,8,balanced,0.06972266733646393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,2,8,balanced,0.057520002126693726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,2,8,balanced,0.05798399945100149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,2,8,balanced,0.05712533493836721
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,2,8,balanced,0.05845866600672404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,2,8,balanced,0.0587360014518102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,2,8,balanced,0.05977599819501241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,2,8,balanced,0.058746665716171265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,2,8,balanced,0.06007466713587443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,2,8,balanced,0.06382933259010315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,2,8,balanced,0.0628053347269694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,2,8,balanced,0.06206933160622915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,2,8,balanced,0.06859200199445088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,2,8,balanced,0.07236266632874806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,2,8,balanced,0.07449600100517273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,2,8,balanced,0.0779306689898173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,2,8,balanced,0.0885813335577647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,2,8,balanced,0.09497066338857015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,2,8,balanced,0.11941867073376973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,2,8,balanced,0.12974400321642557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,2,8,balanced,0.16381866733233133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,32,1,power_law_1.01,0.46292481422424314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,2,8,balanced,0.2000853419303894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,2,8,balanced,0.26864532629648846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,2,8,balanced,0.3343786795934041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,2,8,balanced,0.47569068272908527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,2,8,balanced,0.617738684018453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,2,8,balanced,0.7519893646240234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,2,8,balanced,1.1699306964874268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,32,1,power_law_1.01,0.5439616203308105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,2,8,balanced,2.2744372685750327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,32,1,power_law_1.01,0.7292736053466797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,32,1,power_law_1.01,0.9068351745605469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,32,1,power_law_1.01,1.236832046508789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,32,1,power_law_1.01,1.5641152381896972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,32,1,power_law_1.01,2.2408575057983398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,32,1,power_law_1.01,2.886323165893555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,1,power_law_1.01,0.11105279922485352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,32,1,power_law_1.01,3.5532352447509767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,1,power_law_1.01,0.1429695963859558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,1,power_law_1.01,0.20101120471954345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,32,1,power_law_1.01,5.696960067749023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,1,power_law_1.01,0.24693760871887208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,1,power_law_1.01,0.41359357833862304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,1,power_law_1.01,0.598745584487915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,1,power_law_1.01,0.7181503772735596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,32,1,power_law_1.01,11.076761627197266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,1,power_law_1.01,0.9516736030578613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,1,power_law_1.01,0.9869183540344239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,1,power_law_1.01,1.037887954711914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,1,power_law_1.01,1.1030143737792968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,1,power_law_1.01,1.1432703971862792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,1,power_law_1.01,1.1857600212097168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,1,power_law_1.01,1.2183679580688476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,1,power_law_1.01,1.2558208465576173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,1,power_law_1.01,1.2848383903503418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,1,power_law_1.01,1.3307456016540526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,16,power_law_1.01,0.0661952018737793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,1,power_law_1.01,1.4056768417358398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,16,power_law_1.01,0.07711359858512878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,1,power_law_1.01,1.509004783630371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,16,power_law_1.01,0.053260797262191774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,1,power_law_1.01,1.6631423950195312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,16,power_law_1.01,0.057894402742385866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,1,power_law_1.01,1.783942413330078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,16,power_law_1.01,0.062003201246261595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,1,power_law_1.01,2.1453119277954102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,16,power_law_1.01,0.07220479846000671
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,1,power_law_1.01,2.269113540649414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,16,power_law_1.01,0.07812479734420777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,1,power_law_1.01,2.8266687393188477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,16,power_law_1.01,0.07653759717941284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,1,power_law_1.01,3.4610366821289062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,16,power_law_1.01,0.07620480060577392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,16,power_law_1.01,0.07832959890365601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,1,power_law_1.01,4.580473709106445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,16,power_law_1.01,0.07899519801139832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,16,power_law_1.01,0.08224639892578126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,1,power_law_1.01,5.623878479003906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,16,power_law_1.01,0.08494719862937927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,16,power_law_1.01,0.08688639998435974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,1,power_law_1.01,6.8087104797363285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,16,power_law_1.01,0.09333119988441467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,16,power_law_1.01,0.09618560075759888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,1,power_law_1.01,9.763763427734375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,16,power_law_1.01,0.10191999673843384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,16,power_law_1.01,0.11866240501403809
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,16,power_law_1.01,0.12498559951782226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,16,power_law_1.01,0.15173759460449218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,1,power_law_1.01,18.082733154296875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,16,power_law_1.01,0.19178240299224852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,16,power_law_1.01,0.23756160736083984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,16,power_law_1.01,0.2639807939529419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,16,power_law_1.01,0.3449408054351807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,16,power_law_1.01,0.4148543834686279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,16,power_law_1.01,0.5033792018890381
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,8,power_law_1.2,0.03504000008106232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,8,power_law_1.2,0.02834559977054596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,8,power_law_1.2,0.028352001309394838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,8,power_law_1.2,0.029023998975753786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,16,power_law_1.01,0.6816703796386718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,8,power_law_1.2,0.029497599601745604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,8,power_law_1.2,0.03342080116271973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,8,power_law_1.2,0.033606401085853575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,8,power_law_1.2,0.034067198634147644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,8,power_law_1.2,0.034220799803733826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,8,power_law_1.2,0.034841600060462954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,8,power_law_1.2,0.037478399276733396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,8,16,power_law_1.01,0.850592041015625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,8,power_law_1.2,0.0384768009185791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,8,power_law_1.2,0.04182400107383728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,8,power_law_1.2,0.04046719968318939
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,8,power_law_1.2,0.04436480104923248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,8,power_law_1.2,0.04572800099849701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,8,power_law_1.2,0.04958080053329468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,8,power_law_1.2,0.053907197713851926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,8,power_law_1.2,0.06058880090713501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,8,16,power_law_1.01,1.35665283203125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,8,power_law_1.2,0.07440000176429748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,8,power_law_1.2,0.0814848005771637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,8,power_law_1.2,0.12161279916763305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,8,power_law_1.2,0.15063680410385133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,8,power_law_1.2,0.22213759422302246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,8,power_law_1.2,0.2726720094680786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,8,power_law_1.2,0.3874943971633911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,8,power_law_1.2,0.5068799972534179
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,8,power_law_1.2,0.6704448223114013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,8,16,power_law_1.01,2.7591232299804687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,8,power_law_1.2,1.0783616065979005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,8,power_law_1.2,2.3183616638183593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,4,power_law_1.2,0.09511039853096008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,4,power_law_1.2,0.16917760372161866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,4,power_law_1.2,0.08551040291786194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,32,balanced,0.052005335688591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,4,power_law_1.2,0.12363519668579101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,32,balanced,0.04321600000063578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,32,balanced,0.043712000052134194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,32,balanced,0.04390400151411692
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,32,balanced,0.04574933151404063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,32,balanced,0.047781333327293396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,32,balanced,0.04817600051561991
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,32,balanced,0.048101335763931274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,32,balanced,0.05022400120894114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,32,balanced,0.04809600114822388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,32,balanced,0.048197334011395775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,32,balanced,0.049866666396458946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,32,balanced,0.04994133114814758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,32,balanced,0.04995200037956238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,32,balanced,0.053914666175842285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,32,balanced,0.05620799958705902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,32,balanced,0.0561653325955073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,32,balanced,0.06310933331648509
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,32,balanced,0.06422399977842967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,4,power_law_1.2,0.1613055944442749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,32,balanced,0.07259733478228252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,32,balanced,0.08079466720422109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,32,balanced,0.10152533650398254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,32,balanced,0.12174933155377705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,32,balanced,0.15676800409952799
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,32,balanced,0.1702453295389811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,32,balanced,0.23730132977167764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,32,balanced,0.2836746573448181
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,32,balanced,0.3285653392473857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,32,balanced,0.4946826696395874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,4,power_law_1.2,0.20500481128692627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,32,balanced,0.9364159901936849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,4,power_law_1.2,0.30138239860534666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,4,power_law_1.2,0.33101439476013184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,4,power_law_1.2,0.3390655994415283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,4,power_law_1.2,0.35905919075012205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,4,power_law_1.2,0.36913280487060546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,4,power_law_1.2,0.3711359977722168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,4,power_law_1.2,0.39346559047698976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,4,power_law_1.2,0.4077951908111572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,4,power_law_1.2,0.4188864231109619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,4,power_law_1.2,0.42663040161132815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,4,power_law_1.2,0.4553855895996094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,4,power_law_1.2,0.5146880149841309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,4,power_law_1.2,0.517900800704956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,4,power_law_1.2,0.5801599979400635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,4,power_law_1.2,0.5824512004852295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,4,power_law_1.2,0.7071680068969727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,4,power_law_1.2,0.8159935951232911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,4,power_law_1.2,1.0323200225830078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,4,power_law_1.2,1.1706560134887696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,4,power_law_1.2,1.5907520294189452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,4,power_law_1.2,2.010150337219238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,4,power_law_1.2,2.38799991607666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,4,power_law_1.2,3.665024185180664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,4,power_law_1.2,6.420838165283203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,16,power_law_1.2,0.055308800935745236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,16,power_law_1.2,0.05267840027809143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,16,power_law_1.2,0.04833920001983642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,16,power_law_1.2,0.050323200225830075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,16,power_law_1.2,0.05307520031929016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,16,power_law_1.2,0.058195197582244874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,16,power_law_1.2,0.06039040088653565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,16,power_law_1.2,0.062745600938797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,16,power_law_1.2,0.06547840237617493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,16,power_law_1.2,0.06698880195617676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,16,power_law_1.2,0.06893439888954163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,16,power_law_1.2,0.07095040082931518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,16,power_law_1.2,0.07446399927139283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,16,power_law_1.2,0.07477759718894958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,4,power_law_1.01,0.11923199892044067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,16,power_law_1.2,0.08635519742965699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,16,power_law_1.2,0.08802559971809387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,4,power_law_1.01,0.24412798881530762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,16,power_law_1.2,0.08910080194473266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,16,power_law_1.2,0.10448640584945679
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,4,power_law_1.01,0.10124160051345825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,16,power_law_1.2,0.11989120244979859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,16,power_law_1.2,0.133951997756958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,4,power_law_1.01,0.16371840238571167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,16,power_law_1.2,0.17211519479751586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,16,power_law_1.2,0.22894721031188964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,4,power_law_1.01,0.22077438831329346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,16,power_law_1.2,0.273305606842041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,4,power_law_1.01,0.3368256092071533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,16,power_law_1.2,0.36674559116363525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,4,power_law_1.01,0.3622591972351074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,16,power_law_1.2,0.4812032222747803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,16,power_law_1.2,0.7084415912628174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,4,power_law_1.01,0.38168959617614745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,16,power_law_1.2,0.9440192222595215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,4,power_law_1.01,0.3866944074630737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,16,power_law_1.2,1.1390975952148437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,4,power_law_1.01,0.3926912069320679
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,4,power_law_1.01,0.4046656131744385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,16,power_law_1.2,1.8658111572265625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,4,power_law_1.01,0.4188223838806152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,4,power_law_1.01,0.42501120567321776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,16,power_law_1.2,4.422848129272461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,4,power_law_1.01,0.4331647872924805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,4,power_law_1.01,0.4462719917297363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,4,power_law_1.01,0.45983362197875977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,4,power_law_1.01,0.4999104022979736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,4,power_law_1.01,0.58122878074646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,4,power_law_1.01,0.5581888198852539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,4,power_law_1.01,0.6432064056396485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,32,power_law_1.01,0.044614401459693906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,32,power_law_1.01,0.04046080112457275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,4,power_law_1.01,0.7039936065673829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,32,power_law_1.01,0.03692800104618073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,4,power_law_1.01,0.8315008163452149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,32,power_law_1.01,0.037273600697517395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,32,power_law_1.01,0.036294400691986084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,4,power_law_1.01,0.9030591964721679
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,32,power_law_1.01,0.0394239991903305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,32,power_law_1.01,0.039110401272773744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,32,power_law_1.01,0.037990400195121767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,4,power_law_1.01,1.1859392166137694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,32,power_law_1.01,0.03840000033378601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,32,power_law_1.01,0.03935999870300293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,32,power_law_1.01,0.03932799994945526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,4,power_law_1.01,1.4105664253234864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,32,power_law_1.01,0.04008319973945618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,32,power_law_1.01,0.041465601325035094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,32,power_law_1.01,0.04317440092563629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,4,power_law_1.01,1.949113655090332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,32,power_law_1.01,0.04915199875831604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,32,power_law_1.01,0.048332801461219786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,32,power_law_1.01,0.054527997970581055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,32,power_law_1.01,0.060115200281143186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,4,power_law_1.01,2.2400640487670898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,32,power_law_1.01,0.06817280054092408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,32,power_law_1.01,0.07771520018577575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,32,power_law_1.01,0.08314239978790283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,32,power_law_1.01,0.11006079912185669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,32,power_law_1.01,0.1369279980659485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,1,4,power_law_1.01,2.6370431900024416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,32,power_law_1.01,0.1840000033378601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,32,power_law_1.01,0.22651519775390624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,32,power_law_1.01,0.29845120906829836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,32,power_law_1.01,0.35082240104675294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,1,4,power_law_1.01,4.054803085327149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,32,power_law_1.01,0.4424704074859619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,32,power_law_1.01,0.6976191997528076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,32,power_law_1.01,1.42926082611084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,1,4,power_law_1.01,7.788909149169922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,4,8,power_law_1.01,0.038790398836135866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,4,8,power_law_1.01,0.04119040071964264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,4,8,power_law_1.01,0.0379967987537384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,4,8,power_law_1.01,0.040601599216461184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,4,8,power_law_1.01,0.03954559862613678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,4,8,power_law_1.01,0.041075199842453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,4,8,power_law_1.01,0.040992000699043275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,4,8,power_law_1.01,0.04069119989871979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,4,8,power_law_1.01,0.04213759899139404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,8,power_law_1.01,0.04552960097789764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,4,8,power_law_1.01,0.04215039908885956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,4,8,power_law_1.01,0.043161600828170776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,4,8,power_law_1.01,0.04347519874572754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,8,power_law_1.01,0.05314559936523437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,4,8,power_law_1.01,0.04948480129241943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,4,8,power_law_1.01,0.050809597969055174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,4,8,power_law_1.01,0.06108160018920898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,8,power_law_1.01,0.049209600687026976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,4,8,power_law_1.01,0.061414402723312375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,4,8,power_law_1.01,0.06800640225410462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,4,8,power_law_1.01,0.07891200184822082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,8,power_law_1.01,0.05011839866638183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,4,8,power_law_1.01,0.09363840222358703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,4,8,power_law_1.01,0.11387519836425782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,8,power_law_1.01,0.05500800013542175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,4,8,power_law_1.01,0.1278656005859375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,4,8,power_law_1.01,0.15741440057754516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,4,8,power_law_1.01,0.1917248010635376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,8,power_law_1.01,0.06275839805603027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,4,8,power_law_1.01,0.2374272108078003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,4,8,power_law_1.01,0.35187840461730957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,8,power_law_1.01,0.0707647979259491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,4,8,power_law_1.01,0.5031871795654297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,4,8,power_law_1.01,0.6561024188995361
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,8,power_law_1.01,0.0747327983379364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,4,8,power_law_1.01,0.8094143867492676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,8,power_law_1.01,0.07594879865646362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,4,8,power_law_1.01,1.2345600128173828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,8,power_law_1.01,0.07727360129356384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,4,8,power_law_1.01,2.4020288467407225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,8,power_law_1.01,0.07919999957084656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,8,power_law_1.01,0.08377599716186523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,8,power_law_1.01,0.08439679741859436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,8,power_law_1.01,0.09098880290985108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,8,power_law_1.01,0.09889919757843017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,8,power_law_1.01,0.10325119495391846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,8,power_law_1.01,0.10607359409332276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,8,power_law_1.01,0.1269503951072693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,8,power_law_1.01,0.13708159923553467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,8,power_law_1.01,0.1630336046218872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,8,power_law_1.01,0.20634880065917968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,8,power_law_1.01,0.24951679706573487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,8,power_law_1.01,0.3190079927444458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,1,64,power_law_1.01,0.027827200293540955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,8,power_law_1.01,0.41224961280822753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,1,64,power_law_1.01,0.02686080038547516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,8,power_law_1.01,0.5262784004211426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,1,64,power_law_1.01,0.02696320116519928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,8,power_law_1.01,0.7958144187927246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,1,64,power_law_1.01,0.029625600576400755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,1,64,power_law_1.01,0.02776319980621338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,8,power_law_1.01,0.9619008064270019
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,1,64,power_law_1.01,0.029465600848197937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,1,64,power_law_1.01,0.030675199627876282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,8,power_law_1.01,1.212723159790039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,1,64,power_law_1.01,0.031097599864006044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,1,64,power_law_1.01,0.032492798566818235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,8,power_law_1.01,1.9313343048095704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,1,64,power_law_1.01,0.03383040130138397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,1,64,power_law_1.01,0.03423359990119934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,1,64,power_law_1.01,0.0353408008813858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,1,64,power_law_1.01,0.03501439988613129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,8,power_law_1.01,3.643590545654297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,1,64,power_law_1.01,0.032876798510551454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,1,64,power_law_1.01,0.03549439907073974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,1,64,power_law_1.01,0.03603839874267578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,1,64,power_law_1.01,0.037990400195121767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,1,64,power_law_1.01,0.0400191992521286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,64,power_law_1.01,0.04688000082969666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,64,power_law_1.01,0.04843519926071167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,64,power_law_1.01,0.051148802042007446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,64,power_law_1.01,0.0599295973777771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,64,power_law_1.01,0.07090560197830201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,64,power_law_1.01,0.08815360069274902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,64,power_law_1.01,0.10200320482254029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,64,power_law_1.01,0.16599680185317994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,64,power_law_1.01,0.19909759759902954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,64,power_law_1.01,0.23851521015167237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,64,power_law_1.01,0.3579456090927124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,32,power_law_1.2,0.07189760208129883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,64,power_law_1.01,0.5974143981933594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,32,power_law_1.2,0.0692736029624939
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,32,power_law_1.2,0.0578495979309082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,32,power_law_1.2,0.059654402732849124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,32,power_law_1.2,0.061484801769256595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,32,power_law_1.2,0.0686016023159027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,8,power_law_1.01,0.03356800079345703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,8,power_law_1.01,0.03949440121650696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,8,power_law_1.01,0.02858879864215851
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,32,power_law_1.2,0.06855040192604064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,8,power_law_1.01,0.031142398715019226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,8,power_law_1.01,0.03185279965400696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,8,power_law_1.01,0.03511680066585541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,32,power_law_1.2,0.07077119946479797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,8,power_law_1.01,0.03682560026645661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,16,balanced,0.05671999851862589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,16,balanced,0.04749333361784617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,16,balanced,0.04570133487383524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,16,balanced,0.04900266726811727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,8,power_law_1.01,0.03466239869594574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,16,balanced,0.05394133428732554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,16,balanced,0.07513066629568736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,16,balanced,0.09165333708127339
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,16,balanced,0.09218133489290874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,16,balanced,0.0960106650988261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,8,power_law_1.01,0.03526400029659271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,16,balanced,0.0939626693725586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,16,balanced,0.09307199716567993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,32,power_law_1.2,0.07153279781341552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,16,balanced,0.09624532858530681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,16,balanced,0.09622400005658467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,16,balanced,0.09611733754475911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,8,power_law_1.01,0.03505280017852783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,16,balanced,0.10372266173362732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,16,balanced,0.10313600301742554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,16,balanced,0.1088053286075592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,16,balanced,0.11528000235557556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,16,balanced,0.12221333384513855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,8,power_law_1.01,0.03619199991226196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,16,balanced,0.1374506652355194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,16,balanced,0.14917332927385965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,16,balanced,0.17776532967885336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,16,balanced,0.2109760046005249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,8,power_law_1.01,0.03729279935359955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,16,balanced,0.2621919910113017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,16,balanced,0.3248800039291382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,32,power_law_1.2,0.07189120054244995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,16,balanced,0.4384426673253377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,8,power_law_1.01,0.03961600065231323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,16,balanced,0.5508426825205485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,16,balanced,0.6978027025858561
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,8,power_law_1.01,0.04082559943199158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,16,balanced,1.0097706317901611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,8,power_law_1.01,0.04493440091609955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,16,balanced,1.95143461227417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,32,power_law_1.2,0.07121919989585876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,8,power_law_1.01,0.048467200994491574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,8,power_law_1.01,0.050400000810623166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,8,power_law_1.01,0.05761920213699341
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,32,power_law_1.2,0.07491840124130249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,8,power_law_1.01,0.055929601192474365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,8,power_law_1.01,0.06258559823036194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,8,power_law_1.01,0.07238399982452393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,32,power_law_1.2,0.07714560031890869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,8,power_law_1.01,0.08535040020942689
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,8,power_law_1.01,0.09535999894142151
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,32,power_law_1.2,0.08396160006523132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,8,power_law_1.01,0.12868479490280152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,8,power_law_1.01,0.15719679594039918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,8,power_law_1.01,0.21932799816131593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,32,power_law_1.2,0.08866559863090515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,8,power_law_1.01,0.24292480945587158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,4,8,power_law_1.01,0.30652799606323244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,32,power_law_1.2,0.0913856029510498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,4,8,power_law_1.01,0.47424001693725587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,32,power_law_1.2,0.09741439819335937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,4,8,power_law_1.01,0.8750656127929688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,2,8,power_law_1.01,0.08218240141868591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,32,power_law_1.2,0.11495039463043213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,2,8,power_law_1.01,0.0840063989162445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,2,8,power_law_1.01,0.08010879755020142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,2,8,power_law_1.01,0.0804032027721405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,2,8,power_law_1.01,0.08010240197181702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,32,power_law_1.2,0.1316799998283386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,2,8,power_law_1.01,0.08459519743919372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,2,8,power_law_1.01,0.09091839790344239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,2,8,power_law_1.01,0.08938239812850952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,2,8,power_law_1.01,0.09493759870529175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,2,8,power_law_1.01,0.09774079918861389
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,32,power_law_1.2,0.16109440326690674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,2,8,power_law_1.01,0.09351680278778077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,2,8,power_law_1.01,0.09711999893188476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,2,8,power_law_1.01,0.09826560020446777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,2,8,power_law_1.01,0.12729599475860595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,2,8,power_law_1.01,0.14995839595794677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,2,8,power_law_1.01,0.16292480230331421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,32,power_law_1.2,0.19086719751358033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,2,8,power_law_1.01,0.13336960077285767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,2,8,power_law_1.01,0.18083839416503905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,2,8,power_law_1.01,0.23340160846710206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,2,8,power_law_1.01,0.2818687915802002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,2,8,power_law_1.01,0.3292992115020752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,32,power_law_1.2,0.24542078971862794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,2,8,power_law_1.01,0.4964735984802246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,2,8,power_law_1.01,0.6298880100250244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,2,8,power_law_1.01,1.0326272010803224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,2,8,power_law_1.01,1.2961088180541993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,32,power_law_1.2,0.268723201751709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,2,8,power_law_1.01,1.9148160934448242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,2,8,power_law_1.01,2.355801582336426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,32,power_law_1.2,0.3795072078704834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,2,8,power_law_1.01,3.1139135360717773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,8,power_law_1.01,0.04659200012683869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,8,power_law_1.01,0.03849599957466125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,8,power_law_1.01,0.04287360012531281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,2,8,power_law_1.01,4.6802112579345705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,8,power_law_1.01,0.04630399942398071
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,32,power_law_1.2,0.5071743965148926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,8,power_law_1.01,0.05073919892311096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,8,power_law_1.01,0.06044160127639771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,8,power_law_1.01,0.06531839966773986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,8,power_law_1.01,0.06408320069313049
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,2,8,power_law_1.01,9.299641418457032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,8,power_law_1.01,0.06448000073432922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,32,power_law_1.2,0.6565760135650635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,8,power_law_1.01,0.06728320121765137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,8,power_law_1.01,0.06902400255203248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,8,power_law_1.01,0.06955519914627076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,8,power_law_1.01,0.07094399929046631
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,8,power_law_1.01,0.07148159742355346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,8,power_law_1.01,0.08090879917144775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,32,power_law_1.2,0.7668799877166748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,8,power_law_1.01,0.08238720297813415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,8,power_law_1.01,0.08512639999389648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,8,power_law_1.01,0.09400320053100586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,8,power_law_1.01,0.11459200382232666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,8,power_law_1.01,0.13303040266036986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,8,power_law_1.01,0.14574719667434693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,2,32,power_law_1.2,1.0542783737182617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,8,power_law_1.01,0.18091520071029663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,8,power_law_1.01,0.21514880657196045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,8,power_law_1.01,0.2832319974899292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,8,power_law_1.01,0.33597440719604493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,8,power_law_1.01,0.4668992042541504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,2,32,power_law_1.2,1.7415807723999024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,8,power_law_1.01,0.6448768138885498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,8,power_law_1.2,0.052121597528457644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,8,power_law_1.2,0.04056960046291351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,8,power_law_1.01,0.8450112342834473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,8,power_law_1.2,0.03852800130844116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,8,power_law_1.01,1.2239551544189453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,8,power_law_1.2,0.04614399969577789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,8,power_law_1.2,0.050400000810623166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,8,power_law_1.01,2.5017791748046876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,8,power_law_1.2,0.07125120162963867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,8,power_law_1.2,0.0710528016090393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,2,32,power_law_1.2,4.088256072998047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,8,power_law_1.2,0.06753280162811279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,8,power_law_1.2,0.07166079878807068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,8,power_law_1.2,0.073580801486969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,8,power_law_1.2,0.07363839745521546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,8,power_law_1.2,0.0738048017024994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,8,power_law_1.2,0.07444480061531067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,8,power_law_1.2,0.08195199966430664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,8,power_law_1.2,0.0931007981300354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,8,power_law_1.2,0.09717119932174682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,8,power_law_1.2,0.09488639831542969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,8,power_law_1.2,0.10770560503005981
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,8,power_law_1.2,0.12083840370178223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,8,power_law_1.2,0.14771840572357178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,8,power_law_1.2,0.12961280345916748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,8,power_law_1.2,0.169977605342865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,8,power_law_1.2,0.1681920051574707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,1,1,power_law_1.01,0.21191039085388183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,1,1,power_law_1.01,0.28688640594482423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,1,1,power_law_1.01,0.4303999900817871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,128,balanced,0.03842133283615112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,128,balanced,0.04044266790151596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,128,balanced,0.04012800008058548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,1,1,power_law_1.01,0.7073984146118164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,128,balanced,0.03828266759713491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,128,balanced,0.03957333415746689
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,128,balanced,0.04036800066630045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,8,power_law_1.2,0.2512320041656494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,128,balanced,0.042463997999827065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,128,balanced,0.04173333446184794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,1,1,power_law_1.01,0.7371456146240234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,128,balanced,0.04127999891837438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,128,balanced,0.04247466723124186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,128,balanced,0.04215466479460398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,128,balanced,0.04187199970086416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,128,balanced,0.04247466723124186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,128,balanced,0.04215999941031138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,1,1,power_law_1.01,0.7208640098571777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,128,balanced,0.048325334986050926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,128,balanced,0.046426668763160706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,128,balanced,0.04654933512210846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,128,balanced,0.04801600178082784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,128,balanced,0.05146666864554087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,1,1,power_law_1.01,0.7552896022796631
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,128,balanced,0.05269333223501841
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,128,balanced,0.05461333195368449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,128,balanced,0.062447999914487205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,8,power_law_1.2,0.3041215896606445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,128,balanced,0.06483733157316844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,128,balanced,0.08078399797280629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,1,1,power_law_1.01,0.7523712158203125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,128,balanced,0.09634666641553243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,128,balanced,0.12085333466529846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,128,balanced,0.14632532993952432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,2688,1856,6,128,1,128,balanced,0.17173866430918375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,1,1,power_law_1.01,0.7573056221008301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,2688,1856,6,128,1,128,balanced,0.24815466006596884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,2688,1856,6,128,1,128,balanced,0.45003732045491535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,1,1,power_law_1.01,0.788102388381958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,1,1,power_law_1.01,0.8148863792419434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,8,power_law_1.2,0.39574398994445803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,1,1,power_law_1.01,0.8345472335815429
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,1,1,power_law_1.01,0.8815168380737305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,1,1,power_law_1.01,0.8515135765075683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,8,power_law_1.2,0.49032320976257326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,1,1,power_law_1.01,0.910854434967041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,1,1,power_law_1.01,0.9163519859313964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,1,1,power_law_1.01,0.9991168022155762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,1,8,power_law_1.2,0.5783616065979004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,1,1,power_law_1.01,1.1055616378784179
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,1,1,power_law_1.01,1.2742143630981446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,1,1,power_law_1.01,1.5730560302734375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,1,1,power_law_1.01,1.7159551620483398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,1,8,power_law_1.2,1.0132160186767578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,1,1,power_law_1.01,2.2413183212280274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,1,1,power_law_1.01,2.8173376083374024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,1,8,power_law_1.2,1.6831039428710937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,1,1,power_law_1.01,4.002316665649414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,1,1,power_law_1.01,5.177734375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,1,1,power_law_1.01,7.6893760681152346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,64,power_law_1.2,0.030553600192070006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,1,1,power_law_1.01,9.984646606445313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,64,power_law_1.2,0.03107840120792389
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,64,power_law_1.2,0.029798400402069092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,64,power_law_1.2,0.030752000212669373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,64,power_law_1.2,0.030771198868751525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,64,power_law_1.2,0.031839999556541446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,1,1,power_law_1.01,12.559616088867188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,64,power_law_1.2,0.03192319869995117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,64,power_law_1.2,0.032671999931335446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,64,power_law_1.2,0.03282560110092163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,64,power_law_1.2,0.032646399736404416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,64,power_law_1.2,0.033913600444793704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,64,power_law_1.2,0.03568640053272247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,64,power_law_1.2,0.03585920035839081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,64,power_law_1.2,0.036595198512077334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,4,power_law_1.01,0.05120000243186951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,1,1,power_law_1.01,19.77015075683594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,64,power_law_1.2,0.04110080003738403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,64,power_law_1.2,0.04240640103816986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,4,power_law_1.01,0.07281280159950257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,64,power_law_1.2,0.04395520091056824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,64,power_law_1.2,0.048403200507164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,64,power_law_1.2,0.0532800018787384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,64,power_law_1.2,0.0632960021495819
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,4,power_law_1.01,0.0597055971622467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,64,power_law_1.2,0.07378559708595275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,64,power_law_1.2,0.10313600301742554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,64,power_law_1.2,0.15652480125427246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,4,power_law_1.01,0.06476160287857055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,64,power_law_1.2,0.18956799507141114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,64,power_law_1.2,0.24688000679016114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,4,power_law_1.01,0.08894720077514648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,64,power_law_1.2,0.3997951984405518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,64,power_law_1.2,0.4817535877227783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,1,1,power_law_1.01,38.82012023925781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,4,power_law_1.01,0.09887999892234803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,64,power_law_1.2,0.6130623817443848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,4,power_law_1.01,0.12435840368270874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,64,power_law_1.2,0.9910400390625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,4,power_law_1.01,0.12629120349884032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,4,power_law_1.01,0.13528319597244262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,64,power_law_1.2,2.2356224060058594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,4,power_law_1.01,0.1325376033782959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,4,power_law_1.01,0.14480639696121217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,4,power_law_1.01,0.14897279739379882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,4,power_law_1.01,0.15205760002136232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,4,power_law_1.01,0.1615615963935852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,4,power_law_1.01,0.1702015995979309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,1,power_law_1.01,0.05431680083274841
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,1,power_law_1.01,0.07156479954719544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,4,power_law_1.01,0.17397119998931884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,1,power_law_1.01,0.09537280201911927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,1,power_law_1.01,0.14734079837799072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,4,power_law_1.01,0.17982720136642455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,1,power_law_1.01,0.18220160007476807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,1,power_law_1.01,0.30495998859405515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,4,power_law_1.01,0.1932031989097595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,1,power_law_1.01,0.32228479385375974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,1,power_law_1.01,0.3303231954574585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,4,power_law_1.01,0.20971519947052003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,1,power_law_1.01,0.35023999214172363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,1,power_law_1.01,0.35886080265045167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,4,power_law_1.01,0.2514879941940308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,1,power_law_1.01,0.3603327989578247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,1,power_law_1.01,0.37763841152191163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,1,power_law_1.01,0.3841856002807617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,4,power_law_1.01,0.284006404876709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,1,power_law_1.01,0.39214720726013186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,1,power_law_1.01,0.4133887767791748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,4,power_law_1.01,0.34588799476623533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,1,power_law_1.01,0.4263552188873291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,1,power_law_1.01,0.43398399353027345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,4,power_law_1.01,0.4206399917602539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,1,power_law_1.01,0.4782080173492432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,1,power_law_1.01,0.5074687957763672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,4,power_law_1.01,0.5203199863433838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,1,power_law_1.01,0.5962560176849365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,1,power_law_1.01,0.6551231861114502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,4,power_law_1.01,0.6730751991271973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,1,power_law_1.01,0.816921615600586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,1,power_law_1.01,0.9697600364685058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,4,power_law_1.01,0.8899647712707519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,1,power_law_1.01,1.2935744285583497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,1,power_law_1.01,1.629747200012207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,4,power_law_1.01,1.170796775817871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,1,power_law_1.01,2.2556991577148438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,4,power_law_1.01,1.5332223892211914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,1,power_law_1.01,2.899033546447754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,1,power_law_1.01,3.5338817596435548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,4,power_law_1.01,2.2084928512573243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,1,power_law_1.01,5.424249649047852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,4,power_law_1.01,4.1250560760498045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,1,power_law_1.01,10.39596176147461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,128,power_law_1.01,0.03930880129337311
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,128,power_law_1.01,0.03873279988765717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,128,power_law_1.01,0.03835520148277283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,2,8,balanced,0.04625066618124644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,2,8,balanced,0.05145066479841868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,2,8,balanced,0.04655466477076212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,2,8,balanced,0.04613866905371348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,2,8,balanced,0.04606399933497111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,2,8,balanced,0.046207999189694725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,2,8,balanced,0.048357332746187844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,128,power_law_1.01,0.04079360067844391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,2,8,balanced,0.04743466774622599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,2,8,balanced,0.04602666695912679
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,2,8,balanced,0.04781866570313772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,2,8,balanced,0.04790933430194855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,2,8,balanced,0.049882665276527405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,2,8,balanced,0.04850666721661886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,2,8,balanced,0.05221866567929586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,2,8,balanced,0.0624533345301946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,2,8,balanced,0.06227200229962667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,2,8,balanced,0.06410133341948192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,2,8,balanced,0.07175466914971669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,2,8,balanced,0.0809440016746521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,2,8,balanced,0.09321066737174988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,2,8,balanced,0.0969599982102712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,2,8,balanced,0.11771733562151591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,2,8,balanced,0.13859200477600098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,2,8,balanced,0.1749066710472107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,2,8,balanced,0.2148053248723348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,2,8,balanced,0.29847466945648193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,128,power_law_1.01,0.039929598569869995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,2,8,balanced,0.3909653425216675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,2,8,balanced,0.46302934487660724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,2,8,balanced,0.7013440132141113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,2,8,balanced,1.3355679512023926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,128,power_law_1.01,0.041228801012039185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,128,power_law_1.01,0.04007680118083954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,128,power_law_1.01,0.0406464010477066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,128,power_law_1.01,0.04056960046291351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,128,power_law_1.01,0.04022400081157684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,128,power_law_1.01,0.04049279987812042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,128,power_law_1.01,0.0406143993139267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,128,power_law_1.01,0.03989120125770569
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,128,power_law_1.01,0.040627199411392215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,128,power_law_1.01,0.043347200751304625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,128,power_law_1.01,0.04480000138282776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,128,power_law_1.01,0.04429439902305603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,128,power_law_1.01,0.045440000295639035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,128,power_law_1.01,0.04630399942398071
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,2,balanced,0.06080000102519989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,2,balanced,0.06454400221506755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,2,balanced,0.07099733253320058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,2,balanced,0.08541333675384521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,128,power_law_1.01,0.052025598287582395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,2,balanced,0.12256532907485962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,2,balanced,0.1834133267402649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,2,balanced,0.18573866287867227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,2,balanced,0.1814346710840861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,2,balanced,0.18379199504852295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,2,balanced,0.1861813267072042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,2,balanced,0.18555732568105063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,2,balanced,0.18853867053985596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,2,balanced,0.1901386578877767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,2,balanced,0.19350934028625488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,128,power_law_1.01,0.056627202033996585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,2,balanced,0.20233599344889322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,2,balanced,0.20474666357040405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,2,balanced,0.2130133310953776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,2,balanced,0.23348800341288248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,2,balanced,0.25303467114766437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,2,balanced,0.2928053339322408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,2,balanced,0.32612266143163043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,2,balanced,0.41838932037353516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,128,power_law_1.01,0.07098879814147949
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,2,balanced,0.49741868178049725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,2,balanced,0.7107199827829996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,2,balanced,0.8926080067952474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,2,balanced,1.2782613436381023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,128,power_law_1.01,0.07758079767227173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,2,balanced,1.6471947034200032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,2,balanced,2.062298615773519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,128,power_law_1.01,0.10402560234069824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,2,balanced,3.1285759607950845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,128,power_law_1.01,0.12458879947662353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,2,balanced,6.025818506876628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,128,power_law_1.01,0.16417920589447021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,128,power_law_1.01,0.2336127996444702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,1,128,power_law_1.01,0.25347840785980225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,1,128,power_law_1.01,0.43073282241821287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,1,128,power_law_1.01,0.7309760093688965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,1,power_law_1.01,0.072953599691391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,1,power_law_1.01,0.09805439710617066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,1,power_law_1.01,0.14830080270767212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,1,power_law_1.01,0.17512960433959962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,1,balanced,0.07439466814200084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,1,balanced,0.12018133203188579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,1,balanced,0.14979733030001322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,1,balanced,0.24975999196370444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,1,balanced,0.44206400712331134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,1,balanced,0.8354026476542155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,1,balanced,0.8402079741160074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,1,balanced,0.8453653653462728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,1,power_law_1.01,0.3236479997634888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,1,balanced,0.8506186803181967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,1,balanced,0.8536266485850016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,1,balanced,0.8552800019582113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,1,balanced,0.8655680020650228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,1,balanced,0.8731359640757242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,1,balanced,0.8788533210754395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,1,balanced,0.8950986862182617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,1,power_law_1.01,0.47388157844543455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,1,balanced,0.9094399611155192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,1,balanced,0.9297706286112467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,1,balanced,0.9802560011545817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,1,balanced,1.0119893550872803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,1,balanced,1.0969279607137044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,1,power_law_1.01,0.5722623825073242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,1,balanced,1.1817973454793294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,1,balanced,1.333253304163615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,1,balanced,1.5214613278706868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,1,balanced,1.9337813059488933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,1,power_law_1.01,0.752723217010498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,1,balanced,2.2341492970784507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,1,balanced,3.196533203125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,1,power_law_1.01,0.7817599773406982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,1,balanced,3.940959930419922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,1,balanced,4.9710133870442705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,1,power_law_1.01,0.8340096473693848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,1,balanced,7.314138412475586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,1,power_law_1.01,0.8800512313842773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,1,power_law_1.01,0.9088959693908691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,1,balanced,14.091669718424479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,1,power_law_1.01,0.9496000289916993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,1,power_law_1.01,0.977177619934082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,1,power_law_1.01,0.9978943824768066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,1,power_law_1.01,1.0500032424926757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,1,power_law_1.01,1.1031871795654298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,1,power_law_1.01,1.1938176155090332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,1,power_law_1.01,1.281510353088379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,1,power_law_1.01,1.4482175827026367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,4,balanced,0.055306668082873024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,4,balanced,0.05774933099746704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,4,balanced,0.05972800155480703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,4,balanced,0.07035199801127116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,4,balanced,0.09935999910036723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,4,balanced,0.1409173309803009
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,1,power_law_1.01,1.4500415802001954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,4,balanced,0.14150399963061014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,4,balanced,0.14201066891352335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,4,balanced,0.14353600144386292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,4,balanced,0.1432906687259674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,4,balanced,0.14563199877738953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,4,balanced,0.1483626663684845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,4,balanced,0.14938132961591086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,4,balanced,0.15032000343004862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,4,balanced,0.15987199544906616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,4,balanced,0.16108266512552896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,4,balanced,0.16234133640925089
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,4,balanced,0.17755732933680216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,1,power_law_1.01,1.656550407409668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,4,balanced,0.1893226703008016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,4,balanced,0.21422400077184042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,4,balanced,0.2383093237876892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,4,balanced,0.2922453284263611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,4,balanced,0.33872532844543457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,4,balanced,0.4760426680246989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,4,balanced,0.5780320167541504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,4,balanced,0.822330633799235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,4,balanced,1.055509328842163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,1,power_law_1.01,1.9639551162719726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,4,balanced,1.3074346383412678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,4,balanced,1.975397268931071
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,4,balanced,3.837402661641439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,1,power_law_1.01,2.4681215286254883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,1,power_law_1.01,2.7069375991821287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,1,power_law_1.01,3.4822975158691407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,1,power_law_1.01,4.454451370239258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,1,power_law_1.01,5.380908966064453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,8,balanced,0.04790933430194855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,8,balanced,0.05235200126965841
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,8,balanced,0.07273600002129872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,8,balanced,0.10326932867368062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,8,balanced,0.15562132994333902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,8,balanced,0.18175999323527017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,8,balanced,0.18226132790247598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,8,balanced,0.18218666315078735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,8,balanced,0.18439465761184692
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,8,balanced,0.1848319967587789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,8,balanced,0.18640534083048502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,8,balanced,0.18769067525863647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,8,balanced,0.18836265802383423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,8,balanced,0.18895467122395834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,8,balanced,0.19438934326171875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,8,balanced,0.19606399536132812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,8,balanced,0.19684267044067383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,1,power_law_1.01,7.766214752197266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,8,balanced,0.2157920002937317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,8,balanced,0.21261332432428995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,8,balanced,0.24607467651367188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,8,balanced,0.23940799633661905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,8,balanced,0.3137226700782776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,8,balanced,0.30188266436258954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,8,balanced,0.41658135255177814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,8,balanced,0.4561226765314738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,8,balanced,0.6818719704945883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,8,balanced,0.7616213162740072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,8,balanced,0.8705759843190511
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,8,balanced,1.3497440020243328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,8,balanced,2.5088747342427573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,1,power_law_1.01,13.665055847167968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,32,1,power_law_1.2,0.030527999997138976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,32,1,power_law_1.2,0.03056640028953552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,32,1,power_law_1.2,0.031033599376678468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,32,1,power_law_1.2,0.03601279854774475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,32,1,power_law_1.2,0.036575999855995175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,32,1,power_law_1.2,0.038764798641204835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,32,1,power_law_1.2,0.03858560025691986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,32,1,power_law_1.2,0.037913599610328676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,32,1,power_law_1.2,0.038771200180053714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,32,1,power_law_1.2,0.03927040100097656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,32,1,power_law_1.2,0.03938559889793396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,8,1,power_law_1.2,0.06467199921607972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,32,1,power_law_1.2,0.04188799858093262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,32,1,power_law_1.2,0.042694398760795595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,8,1,power_law_1.2,0.07091839909553528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,32,1,power_law_1.2,0.04774399995803833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,32,1,power_law_1.2,0.05554559826850891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,32,1,power_law_1.2,0.058950400352478026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,8,1,power_law_1.2,0.08306559920310974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,32,1,power_law_1.2,0.057574397325515746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,32,1,power_law_1.2,0.07109760046005249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,8,1,power_law_1.2,0.09819520115852357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,32,1,power_law_1.2,0.07904639840126038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,32,1,power_law_1.2,0.10188159942626954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,32,1,power_law_1.2,0.11652480363845825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,8,1,power_law_1.2,0.11568640470504761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,32,1,power_law_1.2,0.15734399557113649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,32,1,power_law_1.2,0.1860352039337158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,8,1,power_law_1.2,0.17672319412231446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,32,1,power_law_1.2,0.2624255895614624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,8,1,power_law_1.2,0.18536959886550902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,32,1,power_law_1.2,0.338483190536499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,32,1,power_law_1.2,0.4910272121429443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,8,1,power_law_1.2,0.19367040395736695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,32,1,power_law_1.2,0.6366399765014649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,8,1,power_law_1.2,0.19806079864501952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,32,1,power_law_1.2,0.7886079788208008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,8,1,power_law_1.2,0.2007551908493042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,32,1,power_law_1.2,1.2670207977294923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,8,1,power_law_1.2,0.20609281063079835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,8,1,power_law_1.2,0.21768319606781006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,32,1,power_law_1.2,2.4828607559204103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,8,1,power_law_1.2,0.21921279430389404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,8,1,power_law_1.2,0.22821760177612305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,8,1,power_law_1.2,0.2415168046951294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,8,1,power_law_1.2,0.24924800395965577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,8,1,power_law_1.2,0.2726720094680786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,8,1,power_law_1.2,0.32050559520721433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,8,1,power_law_1.2,0.35653760433197024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,8,1,power_law_1.2,0.44718079566955565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,8,1,power_law_1.2,0.5213568210601807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,8,1,power_law_1.2,0.6922688007354736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,8,1,power_law_1.2,0.8482368469238282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,8,1,power_law_1.2,1.1742464065551759
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,8,1,power_law_1.2,1.5893759727478027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,8,1,power_law_1.2,2.2138559341430666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,8,1,power_law_1.2,2.782566452026367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,16,power_law_1.01,0.145196795463562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,8,1,power_law_1.2,3.428793716430664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,16,power_law_1.01,0.24248960018157958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,8,1,power_law_1.2,5.374777603149414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,16,power_law_1.01,0.11826560497283936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,16,power_law_1.01,0.11043200492858887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,16,power_law_1.01,0.13590400218963622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,8,1,power_law_1.2,10.455974578857422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,16,power_law_1.01,0.1642367959022522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,16,power_law_1.01,0.19519360065460206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,16,power_law_1.01,0.2157952070236206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,16,power_law_1.01,0.2173759937286377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,16,power_law_1.01,0.22521600723266602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,16,power_law_1.01,0.22028160095214844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,8,8,power_law_1.01,0.03345920145511627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,8,8,power_law_1.01,0.03528960049152374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,8,8,power_law_1.01,0.03299199938774109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,8,8,power_law_1.01,0.03289600014686585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,8,8,power_law_1.01,0.03331199884414673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,8,8,power_law_1.01,0.033932799100875856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,16,power_law_1.01,0.22841598987579345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,8,8,power_law_1.01,0.034355199337005614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,8,8,power_law_1.01,0.03461759984493255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,8,8,power_law_1.01,0.03521920144557953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,8,8,power_law_1.01,0.03488639891147614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,8,8,power_law_1.01,0.035743999481201175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,8,8,power_law_1.01,0.036748799681663516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,8,8,power_law_1.01,0.036723199486732486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,16,power_law_1.01,0.23233919143676757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,8,8,power_law_1.01,0.0384768009185791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,8,8,power_law_1.01,0.04379520118236542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,8,8,power_law_1.01,0.04663679897785187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,8,8,power_law_1.01,0.049184000492095946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,8,8,power_law_1.01,0.05551360249519348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,8,8,power_law_1.01,0.05907840132713318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,4,balanced,0.04825599988301595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,16,power_law_1.01,0.2255039930343628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,4,balanced,0.06074133515357971
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,8,8,power_law_1.01,0.07184000015258789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,4,balanced,0.07062933345635732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,4,balanced,0.09857599933942159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,4,balanced,0.1483466625213623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,8,8,power_law_1.01,0.07920640110969543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,4,balanced,0.2437546650568644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,4,balanced,0.24396266539891562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,8,8,power_law_1.01,0.10104320049285889
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,4,balanced,0.24550400177637735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,4,balanced,0.24819199244181314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,4,balanced,0.24759467442830405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,8,8,power_law_1.01,0.1270143985748291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,4,balanced,0.24810133377710977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,4,balanced,0.25199999411900836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,4,balanced,0.25384533405303955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,4,balanced,0.25439467032750446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,8,8,power_law_1.01,0.169542396068573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,4,balanced,0.2656426628430684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,4,balanced,0.2676266630490621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,16,power_law_1.01,0.23918719291687013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,4,balanced,0.27140265703201294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,8,8,power_law_1.01,0.21269121170043945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,4,balanced,0.28537599245707196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,4,balanced,0.2978773315747579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,4,balanced,0.32706133524576825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,8,8,power_law_1.01,0.3050240039825439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,4,balanced,0.3575733502705892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,4,balanced,0.3999093373616536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,4,balanced,0.44919999440511066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,8,8,power_law_1.01,0.3700736045837402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,4,balanced,0.5744586785634359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,4,balanced,0.6555253267288208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,4,balanced,0.8896533648173014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,8,8,power_law_1.01,0.44199042320251464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,16,power_law_1.01,0.24110078811645508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,4,balanced,1.096175988515218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,4,balanced,1.383600076039632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,8,8,power_law_1.01,0.6502079963684082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,4,balanced,2.0086453755696616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,8,8,power_law_1.01,1.2448448181152343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,16,power_law_1.01,0.26318719387054446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,4,balanced,3.8852532704671225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,16,power_law_1.01,0.28775041103363036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,16,power_law_1.01,0.28069119453430175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,16,power_law_1.01,0.3444607973098755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,16,power_law_1.01,0.33980159759521483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,16,power_law_1.01,0.4191296100616455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,16,power_law_1.01,0.45065598487854003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,16,power_law_1.01,0.6051328182220459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,16,power_law_1.01,0.6776256084442138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,16,power_law_1.01,0.9270272254943848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,2,balanced,0.03156266609827677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,2,balanced,0.03346133232116699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,2,balanced,0.04387199878692627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,2,balanced,0.06679999828338623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,2,balanced,0.1076639990011851
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,2,balanced,0.12608533104260763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,2,balanced,0.1258026659488678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,2,balanced,0.1267093320687612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,2,balanced,0.1279146671295166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,2,balanced,0.12711466352144876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,2,balanced,0.12843733032544455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,2,balanced,0.127920001745224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,2,balanced,0.12921067078908285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,2,balanced,0.13030399878819784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,2,balanced,0.13564266761144003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,2,balanced,0.13619732856750488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,2,balanced,0.1397333343823751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,2,balanced,0.17154133319854736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,2,balanced,0.15026666720708212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,2,balanced,0.18599466482798258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,2,balanced,0.17381866772969565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,2,balanced,0.25838400920232135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,2,balanced,0.23966934283574423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,2,balanced,0.3143893281618754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,16,power_law_1.01,1.1694592475891112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,2,balanced,0.3556053241093953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,2,balanced,0.49746131896972656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,2,balanced,0.5803413391113281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,128,power_law_1.2,0.0616703987121582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,2,2,balanced,0.7121813297271729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,2,2,balanced,1.0734506448109944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,2,2,balanced,2.0406773885091147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,16,power_law_1.01,1.4829183578491212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,128,power_law_1.2,0.06410880088806152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,128,power_law_1.2,0.0588096022605896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,16,power_law_1.01,2.169913673400879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,128,power_law_1.2,0.05913599729537964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,128,power_law_1.2,0.06083199977874756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,128,power_law_1.2,0.061536002159118655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,16,power_law_1.01,4.147071838378906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,128,power_law_1.2,0.06427519917488098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,128,power_law_1.2,0.06464639902114869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,128,power_law_1.2,0.06516479849815368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,8,8,balanced,0.04214400053024292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,8,8,balanced,0.04398400088151296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,8,8,balanced,0.03549333413441976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,8,8,balanced,0.035045333206653595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,8,8,balanced,0.0354720006386439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,8,8,balanced,0.03565333286921183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,8,8,balanced,0.03540800015131632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,8,8,balanced,0.03753600021203359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,8,8,balanced,0.035589332381884255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,8,8,balanced,0.03540800015131632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,8,8,balanced,0.03745600084463755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,8,8,balanced,0.03745600084463755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,8,8,balanced,0.035536001125971474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,8,8,balanced,0.03965333352486292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,8,8,balanced,0.04378133515516917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,8,8,balanced,0.04582933088143667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,8,8,balanced,0.05020800232887268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,8,8,balanced,0.05861866474151611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,8,8,balanced,0.05407999952634176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,8,8,balanced,0.07175466914971669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,8,8,balanced,0.0745066652695338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,8,8,balanced,0.09274666508038838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,128,power_law_1.2,0.06465280055999756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,8,8,balanced,0.10147200028101604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,8,8,balanced,0.13084800044695535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,8,8,balanced,0.15405333042144775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,8,8,balanced,0.2021226684252421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,8,8,balanced,0.2597813407580058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,8,8,balanced,0.31666133801142377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,8,8,balanced,0.4354613224665324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,8,8,balanced,0.8134826819101969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,128,power_law_1.2,0.06556159853935242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,128,power_law_1.2,0.06743680238723755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,128,power_law_1.2,0.06823679804801941
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,128,power_law_1.2,0.06849920153617858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,128,power_law_1.2,0.07670400142669678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,1,balanced,0.07355733215808868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,1,balanced,0.12398933370908101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,1,balanced,0.1492853363355001
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,1,balanced,0.24466667572657266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,1,balanced,0.44300798575083417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,1,balanced,0.829535961151123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,1,balanced,0.8326773643493652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,128,power_law_1.2,0.07668480277061462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,1,balanced,0.836085319519043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,1,balanced,0.8384160200754801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,1,balanced,0.84224534034729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,1,balanced,0.8489653269449869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,1,balanced,0.8590719699859619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,1,balanced,0.862506628036499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,128,power_law_1.2,0.08286719918251037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,1,balanced,0.8693013191223145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,1,balanced,0.8855840365091959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,1,balanced,0.8930293718973795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,1,balanced,0.9137333234151205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,1,balanced,0.956378698348999
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,1,balanced,0.9921920299530029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,128,power_law_1.2,0.09473919868469238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,1,balanced,1.0706186294555664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,1,balanced,1.131925344467163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,1,balanced,1.259989341100057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,1,balanced,1.4020373026529949
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,1,balanced,1.880949338277181
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,128,power_law_1.2,0.11078399419784546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,8,8,balanced,0.03598399957021078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,8,8,balanced,0.03779733429352442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,1,balanced,2.0532213846842446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,8,8,balanced,0.036144000788529716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,8,8,balanced,0.03432533393303553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,8,8,balanced,0.0359199990828832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,8,8,balanced,0.03619199991226196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,8,8,balanced,0.035887998839219414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,8,8,balanced,0.03626666714747747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,8,8,balanced,0.03580799947182337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,8,8,balanced,0.03717333326737086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,8,8,balanced,0.03626666714747747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,8,8,balanced,0.03793599953254064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,1,balanced,2.881279945373535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,8,8,balanced,0.03798400113979975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,8,8,balanced,0.03945599993069967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,8,8,balanced,0.042362665136655174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,8,8,balanced,0.04357333481311798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,8,8,balanced,0.04381333291530609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,128,power_law_1.2,0.13716479539871215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,8,8,balanced,0.04394133388996124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,8,8,balanced,0.04809066653251648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,8,8,balanced,0.05228800078233083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,8,8,balanced,0.05594133337338766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,8,8,balanced,0.06864533325036366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,8,8,balanced,0.0800853321949641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,1,balanced,3.548704147338867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,8,8,balanced,0.10399466753005981
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,8,8,balanced,0.11927466591199239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,8,8,balanced,0.15523733695348105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,8,8,balanced,0.19089599450429282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,8,8,balanced,0.2242400050163269
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,8,8,balanced,0.33237866560618085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,8,8,balanced,0.6089653174082438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,1,balanced,4.475920041402181
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,128,power_law_1.2,0.15822720527648926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,1,balanced,6.521055857340495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,128,power_law_1.2,0.2153088092803955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,128,power_law_1.2,0.25553278923034667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,1,balanced,12.433519999186197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,128,power_law_1.2,0.35556480884552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,128,power_law_1.2,0.466099214553833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,128,power_law_1.2,0.6593728065490723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,128,power_law_1.2,0.8847552299499511
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,128,balanced,0.058864002426465355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,128,balanced,0.05691733459631602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,128,balanced,0.058506667613983154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,128,balanced,0.05889600018660227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,128,balanced,0.05896000067392985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,128,balanced,0.06222933530807495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,128,balanced,0.06301333506902058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,128,balanced,0.062047998110453285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,128,balanced,0.06259733438491821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,128,balanced,0.06470933556556702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,128,balanced,0.062277331948280334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,128,balanced,0.06431999802589417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,128,power_law_1.2,1.1195008277893066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,128,balanced,0.0645653357108434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,128,balanced,0.06638399759928386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,128,balanced,0.07051733136177063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,128,balanced,0.07121066749095917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,128,balanced,0.07448000212510426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,128,balanced,0.08081600069999695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,128,balanced,0.0867680013179779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,128,balanced,0.09749333063761394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,128,balanced,0.10937600334485371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,128,balanced,0.14010666807492575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,128,balanced,0.16273599863052368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,128,balanced,0.20984532435735068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,128,balanced,0.25570134321848553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,128,balanced,0.3497653404871623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,128,balanced,0.4400533437728882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,128,balanced,0.5402986605962118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,128,balanced,0.8220319747924805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,128,balanced,1.600661277770996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,128,power_law_1.2,1.7781631469726562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,1,8,balanced,0.06405866642793019
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,1,8,balanced,0.06700266897678375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,1,8,balanced,0.05784533421198527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,1,8,balanced,0.057562669118245445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,1,8,balanced,0.05680533250172933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,1,8,balanced,0.05844266712665558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,1,8,balanced,0.05754666527112325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,1,8,balanced,0.05795733133951823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,1,8,balanced,0.0589279979467392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,1,8,balanced,0.05719466507434845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,1,8,balanced,0.059706668059031166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,1,8,balanced,0.06061866879463196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,1,8,balanced,0.06217599908510844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,1,8,balanced,0.06173333525657654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,1,8,balanced,0.07097599903742473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,1,8,balanced,0.07334400216738383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,1,8,balanced,0.07528000076611836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,1,8,balanced,0.07974933087825775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,1,8,balanced,0.08264000217119853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,1,8,balanced,0.1035040020942688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,1,8,balanced,0.10362666845321655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,1,8,balanced,0.13303466637929282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,1,8,balanced,0.15587733189264932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,1,8,balanced,0.20372267564137778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,1,8,balanced,0.2481173276901245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,1,8,balanced,0.34379732608795166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,1,8,balanced,0.4352320035298665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,1,8,balanced,0.5486720005671183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,1,8,balanced,0.806442658106486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,1,8,balanced,1.5486772855122883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,128,power_law_1.2,3.7465728759765624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,2,power_law_1.01,0.06355199813842774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,2,power_law_1.01,0.10440959930419921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,2,power_law_1.01,0.0871936023235321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,2,power_law_1.01,0.10380799770355224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,4,64,balanced,0.03589333345492681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,4,64,balanced,0.03369066615899404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,4,64,balanced,0.03400533397992452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,4,64,balanced,0.0336053321758906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,4,64,balanced,0.03580799947182337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,4,64,balanced,0.03402133285999298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,4,64,balanced,0.03570666660865148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,4,64,balanced,0.036101333796978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,4,64,balanced,0.035429333647092186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,2,power_law_1.01,0.1535104036331177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,4,64,balanced,0.03622400015592575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,4,64,balanced,0.03615466753641764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,4,64,balanced,0.037920000652472176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,4,64,balanced,0.03704000016053518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,4,64,balanced,0.036490666369597115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,4,64,balanced,0.04054400076468786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,4,64,balanced,0.041365332901477814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,1,power_law_1.01,0.0543615996837616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,4,64,balanced,0.04154666761557261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,4,64,balanced,0.03997866561015447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,4,64,balanced,0.042277331153551735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,4,64,balanced,0.04418133199214935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,4,64,balanced,0.048623998959859215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,2,power_law_1.01,0.1773375988006592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,4,64,balanced,0.05035733183224996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,4,64,balanced,0.05215999980767568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,1,power_law_1.01,0.08429440259933471
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,4,64,balanced,0.06262399752934773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,4,64,balanced,0.07106666763623555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,4,64,balanced,0.0890933374563853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,4,64,balanced,0.11160533626874287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,4,64,balanced,0.1295840044816335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,2,power_law_1.01,0.23585920333862304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,1,power_law_1.01,0.09244160056114196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,4,64,balanced,0.17338667313257852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,4,64,balanced,0.2893066604932149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,1,power_law_1.01,0.13946239948272704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,2,power_law_1.01,0.2512063980102539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,1,power_law_1.01,0.18865280151367186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,1,power_law_1.01,0.3050175905227661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,2,power_law_1.01,0.2525439977645874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,1,power_law_1.01,0.3272063970565796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,2,power_law_1.01,0.25766398906707766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,1,power_law_1.01,0.33610239028930666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,1,power_law_1.01,0.34383358955383303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,2,power_law_1.01,0.28032000064849855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,1,power_law_1.01,0.3557248115539551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,1,power_law_1.01,0.3587392091751099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,2,power_law_1.01,0.28586881160736083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,1,power_law_1.01,0.3700671911239624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,2,power_law_1.01,0.2905855894088745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,1,power_law_1.01,0.3719104051589966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,1,power_law_1.01,0.38414080142974855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,2,power_law_1.01,0.296012806892395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,1,power_law_1.01,0.39502720832824706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,2,power_law_1.01,0.3142848014831543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,1,power_law_1.01,0.4083392143249512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,1,power_law_1.01,0.41239681243896487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,2,power_law_1.01,0.32309119701385497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,1,power_law_1.01,0.45925121307373046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,2,power_law_1.01,0.33841280937194823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,1,power_law_1.01,0.4759488105773926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,1,power_law_1.01,0.5450047969818115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,2,power_law_1.01,0.383571195602417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,1,power_law_1.01,0.5949312210083008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,2,power_law_1.01,0.4207871913909912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,1,power_law_1.01,0.7181759834289551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,32,1,balanced,0.039173332353432976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,32,1,balanced,0.039093332986036934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,32,1,balanced,0.039818666875362396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,32,1,balanced,0.03938133269548416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,32,1,balanced,0.03930133332808813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,2,power_law_1.01,0.4951295852661133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,32,1,balanced,0.04156800111134847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,32,1,balanced,0.04155733436346054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,32,1,balanced,0.040906667709350586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,1,power_law_1.01,0.8524352073669433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,32,1,balanced,0.04289066791534424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,32,1,balanced,0.04298133154710134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,32,1,balanced,0.043791999419530235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,32,1,balanced,0.0436106671889623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,32,1,balanced,0.045509333411852516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,32,1,balanced,0.04515199859937032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,32,1,balanced,0.05259733398755392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,32,1,balanced,0.053082664807637535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,2,power_law_1.01,0.536620807647705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,32,1,balanced,0.05941333373387655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,32,1,balanced,0.06481599807739258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,1,power_law_1.01,1.118841552734375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,32,1,balanced,0.07156800230344136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,32,1,balanced,0.09423999985059102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,32,1,balanced,0.10569066802660625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,32,1,balanced,0.14065066973368326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,32,1,balanced,0.1661066710948944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,32,1,balanced,0.23180800676345825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,32,1,balanced,0.293338676293691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,32,1,balanced,0.4200479984283447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,2,power_law_1.01,0.6655680179595947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,1,power_law_1.01,1.3904319763183595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,32,1,balanced,0.5484373172124227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,32,1,balanced,0.6824639638264974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,32,1,balanced,1.0496373176574707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,32,1,balanced,2.04697593053182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,2,power_law_1.01,0.7509503841400147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,1,power_law_1.01,1.8986112594604492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,2,power_law_1.01,1.0214079856872558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,1,power_law_1.01,2.417344093322754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,2,power_law_1.01,1.2534527778625488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,1,power_law_1.01,2.958470344543457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,2,power_law_1.01,1.7137792587280274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,1,power_law_1.01,4.454367828369141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,2,power_law_1.01,2.1468927383422853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,2,power_law_1.01,2.589107131958008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,1,power_law_1.01,8.481593322753906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,2,power_law_1.01,4.026367950439453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,4,power_law_1.01,0.10577280521392822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,4,power_law_1.01,0.12383359670639038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,2,power_law_1.01,7.675872039794922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,4,power_law_1.01,0.0951807975769043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,4,power_law_1.01,0.1287616014480591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,4,power_law_1.01,0.16868480443954467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,4,power_law_1.01,0.24677760601043702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,4,power_law_1.01,0.26520960330963134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,4,power_law_1.01,0.2658623933792114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,4,power_law_1.01,0.2803136110305786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,4,balanced,0.05619200070699056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,4,balanced,0.07342933118343353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,4,balanced,0.09750399986902873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,4,balanced,0.14852799971898398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,4,balanced,0.2419040004412333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,4,balanced,0.44004801909128827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,4,balanced,0.6355733474095663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,4,power_law_1.01,0.2888063907623291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,4,balanced,0.6353280146916708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,4,balanced,0.6374613444010416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,4,balanced,0.6376906633377075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,4,balanced,0.6382826566696167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,4,balanced,0.6425600051879883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,4,balanced,0.643392006556193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,4,power_law_1.01,0.295142388343811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,4,balanced,0.6469599803288778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,4,balanced,0.6567626794179281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,4,balanced,0.6614293257395426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,4,balanced,0.6666026512781779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,4,balanced,0.6823573112487793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,4,balanced,0.6980266571044922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,4,power_law_1.01,0.293887996673584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,4,balanced,0.7292160193125407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,4,balanced,0.759440024693807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,4,balanced,0.8099466959635416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,4,balanced,0.873738686243693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,4,balanced,0.9726239840189616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,4,power_law_1.01,0.29456000328063964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,4,balanced,1.1275733311971028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,4,balanced,1.323189338048299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,4,balanced,1.7079253196716309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,4,power_law_1.01,0.3090559959411621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,4,balanced,2.2267093658447266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,4,power_law_1.01,0.32727680206298826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,4,balanced,2.915130615234375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,4,power_law_1.01,0.3292543888092041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,4,balanced,5.494698842366536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,4,power_law_1.01,0.06995840072631836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,4,power_law_1.01,0.3313728094100952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,4,power_law_1.01,0.09702399969100953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,4,power_law_1.01,0.38359038829803466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,4,power_law_1.01,0.0700543999671936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,4,power_law_1.01,0.40071678161621094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,4,power_law_1.01,0.09102079868316651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,4,power_law_1.01,0.4746240139007568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,4,power_law_1.01,0.11955840587615967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,4,power_law_1.01,0.5367424011230468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,4,power_law_1.01,0.144486403465271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,4,power_law_1.01,0.662559986114502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,4,power_law_1.01,0.18849920034408568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,4,power_law_1.01,0.6357312202453613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,4,power_law_1.01,0.19679360389709472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,4,power_law_1.01,0.7522047996520996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,4,power_law_1.01,0.20545918941497804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,4,power_law_1.01,0.9519743919372559
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,4,power_law_1.01,0.2068160057067871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,4,power_law_1.01,0.1981760025024414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,4,power_law_1.01,1.3210240364074708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,4,power_law_1.01,0.22415359020233155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,2,4,balanced,0.06643733382225037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,2,4,balanced,0.08498133222262065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,2,4,balanced,0.08226666847864787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,2,4,balanced,0.0812906672557195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,2,4,balanced,0.08334400256474812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,2,4,balanced,0.08224000036716461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,2,4,balanced,0.0823520024617513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,2,4,balanced,0.0842026670773824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,4,power_law_1.01,0.23043200969696045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,2,4,balanced,0.08683733145395915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,2,4,balanced,0.0857973297437032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,2,4,balanced,0.08912000060081482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,4,power_law_1.01,1.567296028137207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,2,4,balanced,0.08788266777992249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,2,4,balanced,0.0874826709429423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,2,4,balanced,0.09411733349164327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,2,4,balanced,0.10052800178527832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,2,4,balanced,0.10073600212732951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,2,4,balanced,0.10734933614730835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,2,4,balanced,0.1167680025100708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,2,4,balanced,0.12609600027402243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,2,4,balanced,0.1546293298403422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,2,4,balanced,0.15892266233762106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,2,4,balanced,0.21737066904703775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,2,4,balanced,0.24459733565648398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,4,power_law_1.01,0.23956480026245117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,2,4,balanced,0.3412906726201375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,2,4,balanced,0.43243201573689777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,2,4,balanced,0.6370293299357096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,2,4,balanced,0.7928906281789144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,2,4,balanced,0.9902613162994385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,4,power_law_1.01,0.25347199440002444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,2,4,balanced,1.54857603708903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,4,power_law_1.01,1.958291244506836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,2,4,balanced,3.0060532887776694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,4,power_law_1.01,0.25851519107818605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,4,power_law_1.01,0.26652801036834717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,4,power_law_1.01,3.28853759765625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,4,power_law_1.01,0.3051071882247925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,4,power_law_1.01,0.3039423942565918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,4,power_law_1.01,0.3751552104949951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,4,power_law_1.01,0.4028800010681152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,4,power_law_1.01,5.6751361846923825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,4,power_law_1.01,0.48819842338562014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,4,power_law_1.01,0.6093376159667969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,4,power_law_1.01,0.7664703845977783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,4,power_law_1.01,0.9203583717346191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,4,power_law_1.01,1.3060288429260254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,16,power_law_1.2,0.048230400681495665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,16,power_law_1.2,0.046035200357437134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,4,power_law_1.01,1.6162431716918946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,16,power_law_1.2,0.043673598766326906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,16,power_law_1.2,0.04314880073070526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,16,power_law_1.2,0.044275200366973876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,4,power_law_1.01,1.942393684387207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,16,power_law_1.2,0.04755840003490448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,16,power_law_1.2,0.04809600114822388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,16,power_law_1.2,0.04958719909191132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,4,power_law_1.01,3.0244863510131834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,16,power_law_1.2,0.05011839866638183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,16,4,power_law_1.2,0.04717440009117126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,16,power_law_1.2,0.05057280063629151
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,16,4,power_law_1.2,0.03808639943599701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,16,4,power_law_1.2,0.03695360124111176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,16,power_law_1.2,0.0537280023097992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,16,4,power_law_1.2,0.03583360016345978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,16,power_law_1.2,0.05438719987869263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,16,4,power_law_1.2,0.035955199599266054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,16,4,power_law_1.2,0.03691520094871521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,16,power_law_1.2,0.05543680191040039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,16,4,power_law_1.2,0.0381632000207901
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,16,4,power_law_1.2,0.038489601016044615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,16,power_law_1.2,0.057599997520446776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,16,4,power_law_1.2,0.042105600237846375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,16,4,power_law_1.2,0.04359680116176605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,4,power_law_1.01,6.187686538696289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,16,power_law_1.2,0.06320000290870667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,16,4,power_law_1.2,0.04352000057697296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,16,4,power_law_1.2,0.046086400747299194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,16,power_law_1.2,0.0648639976978302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,16,4,power_law_1.2,0.04738560020923614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,16,4,power_law_1.2,0.05322880148887634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,16,power_law_1.2,0.06932479739189149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,16,4,power_law_1.2,0.06069759726524353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,16,4,power_law_1.2,0.0662015974521637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,16,power_law_1.2,0.08170239925384522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,16,4,power_law_1.2,0.06702719926834107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,16,4,power_law_1.2,0.08024960160255432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,16,4,power_law_1.2,0.09325439929962158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,16,power_law_1.2,0.09063680171966552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,16,4,power_law_1.2,0.10909440517425537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,16,4,power_law_1.2,0.13585920333862306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,16,power_law_1.2,0.12310400009155273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,16,4,power_law_1.2,0.15971839427947998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,16,4,power_law_1.2,0.18798719644546508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,16,power_law_1.2,0.1502527952194214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,16,4,power_law_1.2,0.2942975997924805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,16,4,power_law_1.2,0.3770751953125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,16,power_law_1.2,0.19025280475616455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,16,4,power_law_1.2,0.5152128219604493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,16,power_law_1.2,0.2430527925491333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,16,4,power_law_1.2,0.732915210723877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,16,4,power_law_1.2,0.9233216285705567
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,16,power_law_1.2,0.32880001068115233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,16,4,power_law_1.2,1.4926591873168946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,16,power_law_1.2,0.40761599540710447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,16,4,power_law_1.2,2.679769515991211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,16,power_law_1.2,0.6584127902984619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,16,power_law_1.2,0.7900991916656495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,16,power_law_1.2,1.020032024383545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,2,power_law_1.01,0.046700799465179445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,2,power_law_1.01,0.05891839861869812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,16,power_law_1.2,1.6874879837036132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,2,power_law_1.01,0.06035199761390686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,2,power_law_1.01,0.08188160061836243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,2,power_law_1.01,0.09478399753570557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,2,power_law_1.01,0.13215359449386596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,2,power_law_1.01,0.1385151982307434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,2,power_law_1.01,0.13938560485839843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,16,power_law_1.2,3.9714496612548826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,2,power_law_1.01,0.14592000246047973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,2,power_law_1.01,0.1464192032814026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,2,power_law_1.01,0.15077760219573974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,2,power_law_1.01,0.16083840131759644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,2,power_law_1.01,0.1583232045173645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,2,power_law_1.01,0.16532479524612426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,2,power_law_1.01,0.17227519750595094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,2,power_law_1.01,0.18197760581970215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,2,power_law_1.01,0.19479680061340332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,2,power_law_1.01,0.22037758827209472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,2,power_law_1.01,0.25106561183929443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,2,power_law_1.01,0.29104640483856203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,4,4,power_law_1.2,0.030265599489212036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,2,power_law_1.01,0.32769920825958254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,2,power_law_1.01,0.4398399829864502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,4,4,power_law_1.2,0.026745599508285523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,2,power_law_1.01,0.5029183864593506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,4,4,power_law_1.2,0.026553601026535034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,2,power_law_1.01,0.7086207866668701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,4,4,power_law_1.2,0.026067200303077697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,2,power_law_1.01,0.8521663665771484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,4,4,power_law_1.2,0.026604801416397095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,2,power_law_1.01,1.1925248146057128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,2,balanced,0.050842667619387306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,2,balanced,0.056549335519472756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,2,balanced,0.07502933343251546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,2,balanced,0.10331733028093974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,2,balanced,0.15736533204714456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,2,balanced,0.2531733314196269
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,4,4,power_law_1.2,0.030924800038337707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,2,balanced,0.25499733289082843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,2,balanced,0.2579626639684041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,2,balanced,0.260858674844106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,2,balanced,0.2635306715965271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,2,balanced,0.26551467180252075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,2,power_law_1.01,1.5233087539672852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,2,balanced,0.27055466175079346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,2,balanced,0.27477333943049115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,2,balanced,0.28270934025446576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,2,balanced,0.2892106572786967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,2,balanced,0.29526400566101074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,2,balanced,0.3054453333218892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,4,4,power_law_1.2,0.03182080090045929
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,2,balanced,0.3309866587320964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,2,balanced,0.3508373498916626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,2,balanced,0.39737598101298016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,2,balanced,0.4468746582667033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,2,balanced,0.5683146715164185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,2,balanced,0.659877339998881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,4,4,power_law_1.2,0.03440000116825104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,2,balanced,0.8944160143534342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,2,power_law_1.01,1.9463487625122071
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,2,balanced,1.0632533232371013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,2,balanced,1.5259466171264648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,4,4,power_law_1.2,0.03544319868087768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,2,balanced,1.9135039647420247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,2,balanced,2.3755946159362793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,4,4,power_law_1.2,0.035462400317192076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,2,balanced,3.59224001566569
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,2,power_law_1.01,2.999430465698242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,4,4,power_law_1.2,0.036499199271202085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,2,balanced,6.930837631225586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,4,4,power_law_1.2,0.038406398892402646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,4,4,power_law_1.2,0.038464000821113585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,4,4,power_law_1.2,0.04584319889545441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,2,power_law_1.01,5.645209503173828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,4,4,power_law_1.2,0.051769602298736575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,4,4,power_law_1.2,0.056569600105285646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,4,4,power_law_1.2,0.05399680137634277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,4,4,power_law_1.2,0.0580735981464386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,4,power_law_1.2,0.07381119728088378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,4,power_law_1.2,0.0851967990398407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,4,power_law_1.2,0.09564160108566284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,4,power_law_1.2,0.12300800085067749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,4,power_law_1.2,0.13430399894714357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,4,power_law_1.2,0.188864004611969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,4,power_law_1.2,0.23520638942718505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,4,power_law_1.2,0.3670592069625854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,4,2,balanced,0.04468800127506256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,4,2,balanced,0.08086933195590973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,4,2,balanced,0.08213333288828532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,4,2,balanced,0.08193066716194153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,4,power_law_1.2,0.5315392017364502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,4,2,balanced,0.08282133440176646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,4,2,balanced,0.08252266546090443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,4,2,balanced,0.08355200290679932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,4,2,balanced,0.08583466211954753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,4,2,balanced,0.083146666487058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,4,2,balanced,0.08431466420491536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,4,2,balanced,0.08679466446240743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,4,2,balanced,0.08905067046483357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,4,2,balanced,0.08612266182899475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,4,2,balanced,0.08733333150545756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,4,2,balanced,0.10172266761461894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,4,2,balanced,0.10597866773605347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,4,2,balanced,0.10136000315348308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,4,2,balanced,0.12036800384521484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,4,2,balanced,0.12225066622098286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,4,2,balanced,0.14662399888038635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,4,2,balanced,0.16098666191101074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,4,2,balanced,0.21244800090789795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,4,2,balanced,0.2538986603418986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,4,2,balanced,0.34807467460632324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,4,2,balanced,0.4340213139851888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,4,power_law_1.2,0.6703807830810546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,4,2,balanced,0.6325333515803019
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,4,2,balanced,0.8171306451161703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,4,2,balanced,1.0049866835276287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,4,2,balanced,1.5685653686523438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,4,2,balanced,3.057114601135254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,32,power_law_1.2,0.08521599769592285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,4,power_law_1.2,1.0007807731628418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,32,power_law_1.2,0.05912320017814636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,32,power_law_1.2,0.05692160129547119
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,8,power_law_1.01,0.09560959935188293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,32,power_law_1.2,0.05633919835090637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,4,power_law_1.2,1.7184192657470703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,32,power_law_1.2,0.05766400098800659
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,8,power_law_1.01,0.1450943946838379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,32,power_law_1.2,0.0692351996898651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,32,power_law_1.2,0.07404800057411194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,8,power_law_1.01,0.23948159217834472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,32,power_law_1.2,0.07345920205116271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,32,power_law_1.2,0.07343360185623168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,8,power_law_1.01,0.10273280143737792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,32,power_law_1.2,0.07356160283088684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,32,power_law_1.2,0.07533439993858337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,8,power_law_1.01,0.13751039505004883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,32,power_law_1.2,0.07542399764060974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,8,power_law_1.01,0.1742527961730957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,32,power_law_1.2,0.07750399708747864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,32,power_law_1.2,0.08081279993057251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,8,power_law_1.01,0.2039936065673828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,32,power_law_1.2,0.0852288007736206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,32,power_law_1.2,0.08963840007781983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,8,power_law_1.01,0.2470463991165161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,32,power_law_1.2,0.09774079918861389
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,32,power_law_1.2,0.10350719690322877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,8,power_law_1.01,0.24613759517669678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,32,power_law_1.2,0.11609599590301514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,32,power_law_1.2,0.142848002910614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,16,power_law_1.01,0.0833728015422821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,8,power_law_1.01,0.26258559226989747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,32,power_law_1.2,0.16072959899902345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,32,power_law_1.2,0.20692479610443115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,8,power_law_1.01,0.2790208101272583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,16,power_law_1.01,0.12954239845275878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,32,power_law_1.2,0.24793601036071777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,8,power_law_1.01,0.2755392074584961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,32,power_law_1.2,0.3697920083999634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,16,power_law_1.01,0.14218239784240722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,32,power_law_1.2,0.4557951927185059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,8,power_law_1.01,0.2863615989685059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,16,power_law_1.01,0.073990398645401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,32,power_law_1.2,0.6480447769165039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,8,power_law_1.2,0.08369920253753663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,8,power_law_1.01,0.2900991916656494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,16,power_law_1.01,0.08302720189094544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,32,power_law_1.2,0.8572287559509277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,8,power_law_1.01,0.2946367979049683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,16,power_law_1.01,0.10252799987792968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,32,power_law_1.2,1.0342399597167968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,8,power_law_1.01,0.29324800968170167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,16,power_law_1.01,0.10308480262756348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,8,power_law_1.2,0.07568640112876893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,8,power_law_1.01,0.32999680042266843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,32,power_law_1.2,1.5398655891418458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,16,power_law_1.01,0.12353919744491577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,8,power_law_1.01,0.34011518955230713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,16,power_law_1.01,0.1196671962738037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,8,power_law_1.2,0.06907520294189454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,8,power_law_1.01,0.372435188293457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,16,power_law_1.01,0.1180799961090088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,32,power_law_1.2,4.094675064086914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,8,power_law_1.01,0.4406144142150879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,16,power_law_1.01,0.12778240442276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,8,power_law_1.2,0.08170239925384522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,8,power_law_1.01,0.4381375789642334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,16,power_law_1.01,0.12892160415649415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,8,power_law_1.01,0.5083775997161866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,16,power_law_1.01,0.13705600500106813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,8,power_law_1.2,0.09132800102233887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,8,power_law_1.01,0.5969024181365967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,16,power_law_1.01,0.1350399971008301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,8,power_law_1.01,0.7474944114685058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,16,power_law_1.01,0.14032000303268433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,8,power_law_1.2,0.11610879898071289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,16,power_law_1.01,0.1435263991355896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,8,power_law_1.01,0.810035228729248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,16,power_law_1.01,0.15393279790878295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,8,power_law_1.2,0.12174079418182374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,8,power_law_1.01,1.0441856384277344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,16,power_law_1.01,0.16670080423355102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,8,power_law_1.2,0.12520320415496827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,16,power_law_1.01,0.18158719539642335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,8,power_law_1.01,1.3611647605895996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,32,power_law_1.01,0.039129599928855896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,32,power_law_1.01,0.03620480000972748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,16,power_law_1.01,0.20661120414733886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,32,power_law_1.01,0.03445119857788086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,8,power_law_1.01,1.7218816757202149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,8,power_law_1.2,0.12579200267791749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,32,power_law_1.01,0.034067198634147644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,16,power_law_1.01,0.23469440937042235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,32,power_law_1.01,0.0348800003528595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,32,power_law_1.01,0.03615359961986542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,32,power_law_1.01,0.036396801471710205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,16,power_law_1.01,0.2910592079162598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,32,power_law_1.01,0.036313599348068236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,8,power_law_1.2,0.12810879945755005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,8,power_law_1.01,2.45230712890625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,32,power_law_1.01,0.03640959858894348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,16,power_law_1.01,0.32551040649414065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,32,power_law_1.01,0.03700479865074158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,32,power_law_1.01,0.03729279935359955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,32,power_law_1.01,0.03844479918479919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,8,power_law_1.2,0.13293440341949464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,16,power_law_1.01,0.423635196685791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,32,power_law_1.01,0.03943040072917938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,32,power_law_1.01,0.04081279933452606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,32,power_law_1.01,0.04631040096282959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,16,power_law_1.01,0.5390399932861328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,8,power_law_1.01,4.35035514831543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,32,power_law_1.01,0.04631040096282959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,8,power_law_1.2,0.14111360311508178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,32,power_law_1.01,0.04815999865531921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,32,power_law_1.01,0.052902400493621826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,16,power_law_1.01,0.7214848041534424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,32,power_law_1.01,0.05857920050621033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,32,power_law_1.01,0.06698240041732788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,8,power_law_1.2,0.14191360473632814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,32,power_law_1.01,0.07372159957885742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,16,power_law_1.01,0.9312128067016602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,32,power_law_1.01,0.09508479833602905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,32,power_law_1.01,0.1089792013168335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,8,power_law_1.2,0.14803839921951295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,32,power_law_1.01,0.14878079891204835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,16,power_law_1.01,1.1179967880249024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,32,power_law_1.01,0.20054399967193604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,32,power_law_1.01,0.2751039981842041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,8,power_law_1.2,0.15550719499588012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,32,power_law_1.01,0.34595839977264403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,16,power_law_1.01,1.7232128143310548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,32,power_law_1.01,0.4418367862701416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,8,power_law_1.2,0.16399359703063965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,32,power_law_1.01,0.5908351898193359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,16,power_law_1.01,3.461075210571289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,8,power_law_1.2,0.17304960489273072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,2,power_law_1.2,0.06839039921760559
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,32,power_law_1.01,1.2609919548034667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,2,power_law_1.2,0.11735039949417114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,2,power_law_1.2,0.08241279721260071
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,8,power_law_1.2,0.19786239862442018
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,2,power_law_1.2,0.12010879516601562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,2,power_law_1.2,0.16128640174865722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,8,power_law_1.2,0.22839679718017578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,2,power_law_1.2,0.20914559364318847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,2,power_law_1.2,0.30346879959106443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,8,power_law_1.2,0.28687999248504636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,2,power_law_1.2,0.3215487957000732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,2,power_law_1.2,0.34318718910217283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,2,power_law_1.2,0.34778239727020266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,8,power_law_1.2,0.32583041191101075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,2,power_law_1.2,0.36847360134124757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,2,power_law_1.2,0.3722879886627197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,8,power_law_1.2,0.42887039184570314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,2,power_law_1.2,0.3794559955596924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,4,power_law_1.01,0.05545600056648255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,2,power_law_1.2,0.41001601219177247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,8,1,power_law_1.2,0.03485440015792847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,8,1,power_law_1.2,0.03653759956359863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,4,power_law_1.01,0.062412798404693604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,8,1,power_law_1.2,0.03656960129737854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,2,power_law_1.2,0.44077439308166505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,8,1,power_law_1.2,0.04444800019264221
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,8,1,power_law_1.2,0.04528000056743622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,4,power_law_1.01,0.055878400802612305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,8,1,power_law_1.2,0.04609920084476471
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,8,1,power_law_1.2,0.047167998552322385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,8,power_law_1.2,0.5023615837097168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,8,1,power_law_1.2,0.04783360064029694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,2,power_law_1.2,0.4357439994812012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,8,1,power_law_1.2,0.04894720017910004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,4,power_law_1.01,0.06641280055046081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,8,1,power_law_1.2,0.04973439872264862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,8,1,power_law_1.2,0.05443199872970581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,8,1,power_law_1.2,0.054579198360443115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,4,power_law_1.01,0.07936000227928161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,8,1,power_law_1.2,0.05805439949035644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,8,1,power_law_1.2,0.06198400259017944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,2,power_law_1.2,0.4701568126678467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,8,1,power_law_1.2,0.07100160121917724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,8,1,power_law_1.2,0.07232000231742859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,4,power_law_1.01,0.10094720125198364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,8,1,power_law_1.2,0.07434239983558655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,8,1,power_law_1.2,0.08446080088615418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,8,1,power_law_1.2,0.0998207986354828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,2,power_law_1.2,0.5100031852722168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,4,power_law_1.01,0.10747519731521607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,8,1,power_law_1.2,0.11826560497283936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,8,1,power_law_1.2,0.13802239894866944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,8,power_law_1.2,0.6392127990722656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,8,1,power_law_1.2,0.17832319736480712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,4,power_law_1.01,0.10922240018844605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,8,1,power_law_1.2,0.21816959381103515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,2,power_law_1.2,0.528172779083252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,8,1,power_law_1.2,0.30271360874176023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,4,power_law_1.01,0.11198079586029053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,8,1,power_law_1.2,0.38071680068969727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,8,1,power_law_1.2,0.5458240032196044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,4,power_law_1.01,0.11415040493011475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,2,power_law_1.2,0.6030399799346924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,8,1,power_law_1.2,0.6957888126373291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,4,power_law_1.01,0.1137984037399292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,8,1,power_law_1.2,0.8634880065917969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,2,power_law_1.2,0.6471104145050048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,4,power_law_1.01,0.11682560443878173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,8,1,power_law_1.2,1.3103232383728027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,8,power_law_1.2,0.8879424095153808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,4,power_law_1.01,0.11692160367965698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,2,power_law_1.2,0.7959296226501464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,4,power_law_1.01,0.12660479545593262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,8,1,power_law_1.2,2.592870330810547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,4,power_law_1.01,0.13308160305023192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,2,power_law_1.2,0.907436752319336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,4,power_law_1.01,0.13696000576019288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,4,power_law_1.01,0.14060159921646118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,2,power_law_1.2,1.1066880226135254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,8,power_law_1.2,1.2513088226318358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,4,power_law_1.01,0.15825920104980468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,4,power_law_1.01,0.172435200214386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,2,power_law_1.2,1.3965567588806151
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,4,power_law_1.01,0.2011967897415161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,4,power_law_1.01,0.2391808032989502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,2,power_law_1.2,1.825324821472168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,4,power_law_1.01,0.29976959228515626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,8,power_law_1.2,1.448025608062744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,4,power_law_1.01,0.37129600048065187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,2,power_law_1.2,2.274163246154785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,4,power_law_1.01,0.5053887844085694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,4,power_law_1.01,0.6009344100952149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,2,power_law_1.2,2.7476287841796876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,4,power_law_1.01,0.8212224006652832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,2,8,power_law_1.2,1.921708869934082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,4,power_law_1.01,1.150220775604248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,2,power_law_1.2,4.018841552734375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,8,balanced,0.04587199787298838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,8,balanced,0.04574933151404063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,8,balanced,0.04785599807898203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,8,balanced,0.05089599887530009
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,8,balanced,0.05514133473237356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,8,balanced,0.08002133170763652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,8,balanced,0.09711466232935588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,8,balanced,0.09944533308347066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,8,balanced,0.09920533498128255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,8,balanced,0.10321600238482158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,8,balanced,0.0990133285522461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,8,balanced,0.10123200217882793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,8,balanced,0.10355732838312785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,4,power_law_1.01,1.2702976226806642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,8,balanced,0.10525866349538167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,8,balanced,0.10931733250617981
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,8,balanced,0.11177600423494975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,8,balanced,0.11958400408426921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,8,balanced,0.12746133406956991
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,8,balanced,0.13857600092887878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,8,balanced,0.15496533115704855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,8,balanced,0.1720106601715088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,8,balanced,0.20755199591318765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,8,balanced,0.24920000632603964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,8,balanced,0.32068800926208496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,8,balanced,0.40696001052856445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,8,balanced,0.5655733346939087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,8,balanced,0.6999039649963379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,8,balanced,0.8849493662516276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,8,balanced,1.309216022491455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,4,power_law_1.01,2.1481088638305663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,8,balanced,2.5258026123046875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,2,8,power_law_1.2,3.1055423736572267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,2,power_law_1.2,8.726739501953125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,4,power_law_1.01,4.301139068603516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,2,8,power_law_1.2,6.718022155761719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,4,2,power_law_1.01,0.047712001204490664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,4,2,power_law_1.01,0.04736000001430511
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,4,2,power_law_1.01,0.05534719824790955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,4,2,power_law_1.01,0.0661184012889862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,4,2,power_law_1.01,0.06421120166778564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,4,2,power_law_1.01,0.06394879817962647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,4,2,power_law_1.01,0.06551679968833923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,4,balanced,0.031632001201311745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,4,balanced,0.03136533250411352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,4,2,power_law_1.01,0.06654719710350036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,4,balanced,0.03334933271010717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,4,balanced,0.05734399954477946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,4,balanced,0.07648533085982005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,4,2,power_law_1.01,0.06835839748382569
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,4,balanced,0.07865599791208903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,4,balanced,0.07934933404127757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,4,balanced,0.0784693310658137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,4,balanced,0.07916266719500224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,4,2,power_law_1.01,0.07017599940299987
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,4,balanced,0.08024533092975616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,4,balanced,0.07876266539096832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,4,balanced,0.08078933258851369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,4,balanced,0.08348266283671062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,4,2,power_law_1.01,0.06977279782295227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,4,balanced,0.08504533767700195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,4,balanced,0.09076799949010213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,4,balanced,0.09205866853396098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,4,balanced,0.09277333815892537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,4,2,power_law_1.01,0.07051519751548767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,4,balanced,0.10469866792360942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,4,balanced,0.10698133707046509
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,4,balanced,0.1223466694355011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,4,2,power_law_1.01,0.0717311978340149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,4,balanced,0.13594133655230203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,4,balanced,0.16614933808644614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,4,2,power_law_1.01,0.07531520128250122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,4,balanced,0.19491199652353922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,4,balanced,0.2521546681722005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,4,2,power_law_1.01,0.08526719808578491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,4,balanced,0.31590932607650757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,2,balanced,0.04598933458328247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,2,balanced,0.05978666742642721
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,4,balanced,0.42560001214345294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,4,2,power_law_1.01,0.08725119829177856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,2,balanced,0.07637866834799449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,2,balanced,0.10499200224876404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,4,balanced,0.5410133202870687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,2,balanced,0.1558080017566681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,2,balanced,0.18770132462183634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,4,2,power_law_1.01,0.09009280204772949
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,2,balanced,0.18662399053573608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,4,balanced,0.6502079963684082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,2,balanced,0.18736533323923746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,2,balanced,0.18831467628479004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,4,2,power_law_1.01,0.10830080509185791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,2,balanced,0.18959466616312662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,4,balanced,0.9884320100148519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,2,balanced,0.1941439906756083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,4,2,power_law_1.01,0.14232319593429565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,2,balanced,0.19443732500076294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,2,balanced,0.19577600558598837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,2,balanced,0.20134933789571127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,4,2,power_law_1.01,0.1539903998374939
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,2,balanced,0.20865599314371744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,4,balanced,1.8931573232014973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,2,balanced,0.21041067441304526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,2,balanced,0.22000000874201456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,4,2,power_law_1.01,0.18220160007476807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,2,balanced,0.2443093260129293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,2,balanced,0.25640533367792767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,2,balanced,0.3094240029652913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,4,2,power_law_1.01,0.23603200912475586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,2,balanced,0.3350293238957723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,2,balanced,0.45286401112874347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,4,2,power_law_1.01,0.31031041145324706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,2,balanced,0.4978933334350586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,2,balanced,0.7347679932912191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,4,2,power_law_1.01,0.4118783950805664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,2,balanced,0.8199626604715983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,4,2,power_law_1.01,0.5285952091217041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,2,balanced,1.2621333599090576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,2,balanced,1.479296048482259
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,4,2,power_law_1.01,0.8550848007202149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,2,balanced,1.7265812555948894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,4,2,power_law_1.01,0.9477952003479004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,2,balanced,2.7635520299275718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,4,2,power_law_1.01,1.1398783683776856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,2,balanced,5.236741383870442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,4,2,power_law_1.01,1.820729637145996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,4,2,power_law_1.01,3.3733440399169923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,32,2,power_law_1.01,0.06275839805603027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,32,2,power_law_1.01,0.06782720088958741
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,32,2,power_law_1.01,0.06662399768829345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,32,2,power_law_1.01,0.07550719976425171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,32,2,power_law_1.01,0.08979840278625488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,32,2,power_law_1.01,0.106278395652771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,4,power_law_1.01,0.03601920008659363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,4,power_law_1.01,0.029580798745155335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,32,2,power_law_1.01,0.12547199726104735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,4,power_law_1.01,0.032953599095344545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,4,power_law_1.01,0.03686400055885315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,4,power_law_1.01,0.037363201379776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,32,2,power_law_1.01,0.13165440559387206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,4,power_law_1.01,0.05141119956970215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,4,power_law_1.01,0.051974397897720334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,32,2,power_law_1.01,0.13093119859695435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,8,balanced,0.03181333343187968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,4,power_law_1.01,0.05617280006408691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,8,balanced,0.03398933261632919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,8,balanced,0.03355200091997782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,8,balanced,0.03603200117746989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,8,balanced,0.05898133416970571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,8,balanced,0.058880001306533813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,8,balanced,0.06121600170930227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,8,balanced,0.06145599981149038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,8,balanced,0.06089599927266439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,4,power_law_1.01,0.05444480180740356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,8,balanced,0.06305066744486491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,8,balanced,0.06124266485373179
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,8,balanced,0.06374399860699971
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,8,balanced,0.06347199777762096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,8,balanced,0.06464533507823944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,8,balanced,0.07281066477298737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,8,balanced,0.07194666564464569
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,8,balanced,0.07547733187675476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,4,power_law_1.01,0.05628799796104431
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,32,2,power_law_1.01,0.1355712056159973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,8,balanced,0.07962666451931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,8,balanced,0.09475732843081157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,8,balanced,0.10590400298436482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,8,balanced,0.1106666624546051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,8,balanced,0.13737066586812338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,8,balanced,0.16900267203648886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,8,balanced,0.2237386703491211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,4,power_law_1.01,0.05632640123367309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,8,balanced,0.26665600140889484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,8,balanced,0.365664005279541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,8,balanced,0.47679467995961505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,8,balanced,0.5796693166097006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,4,power_law_1.01,0.05968639850616455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,32,2,power_law_1.01,0.07028480172157288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,8,balanced,0.8864800135294596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,32,2,power_law_1.01,0.13886079788208008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,8,balanced,1.6915094057718914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,4,power_law_1.01,0.062080001831054686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,4,power_law_1.01,0.06350079774856568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,32,2,power_law_1.01,0.07670400142669678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,32,2,power_law_1.01,0.1451200008392334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,4,power_law_1.01,0.06803200244903565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,4,power_law_1.01,0.07223680019378662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,32,2,power_law_1.01,0.08654720187187195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,4,power_law_1.01,0.07527679800987244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,32,2,power_law_1.01,0.1463744044303894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,4,power_law_1.01,0.08310400247573853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,4,power_law_1.01,0.10392960309982299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,32,2,power_law_1.01,0.08303999900817871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,32,2,power_law_1.01,0.15199999809265136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,4,power_law_1.01,0.12172160148620606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,4,power_law_1.01,0.12431999444961547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,32,2,power_law_1.01,0.10775680541992187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,32,2,power_law_1.01,0.16008960008621215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,4,power_law_1.01,0.17669119834899902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,4,power_law_1.01,0.19274879693984986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,32,2,power_law_1.01,0.1254271984100342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,32,2,power_law_1.01,0.16388479471206666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,4,power_law_1.01,0.2809920072555542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,32,2,power_law_1.01,0.17381759881973266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,32,2,power_law_1.01,0.13694080114364623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,4,power_law_1.01,0.37255680561065674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,4,power_law_1.01,0.49799680709838867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,32,2,power_law_1.01,0.196563196182251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,32,2,power_law_1.01,0.14972800016403198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,4,power_law_1.01,0.644927978515625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,32,2,power_law_1.01,0.22518401145935057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,32,2,power_law_1.01,0.15527039766311646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,4,power_law_1.01,0.7371903896331787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,32,2,power_law_1.01,0.2731712102890015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,32,2,power_law_1.01,0.163264000415802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,4,power_law_1.01,1.2265727996826172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,32,2,power_law_1.01,0.32661759853363037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,32,2,power_law_1.01,0.16915199756622315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,1,power_law_1.2,0.05004159808158874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,32,2,power_law_1.01,0.1751487970352173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,32,2,power_law_1.01,0.42566399574279784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,1,power_law_1.2,0.05750399827957153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,32,2,power_law_1.01,0.1788032054901123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,1,power_law_1.2,0.06684799790382386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,4,power_law_1.01,2.382899284362793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,32,2,power_law_1.01,0.5211967945098877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,1,power_law_1.2,0.09144319891929627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,32,2,power_law_1.01,0.18778879642486573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,32,2,power_law_1.01,0.7555263996124267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,1,power_law_1.2,0.11525119543075561
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,32,2,power_law_1.01,0.19524480104446412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,1,power_law_1.2,0.14780800342559813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,32,2,power_law_1.01,0.9037247657775879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,32,2,power_law_1.01,0.2035072088241577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,1,power_law_1.2,0.21143040657043458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,1,power_law_1.2,0.22700159549713134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,32,2,power_law_1.01,0.20938880443573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,32,2,power_law_1.01,1.3804287910461426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,1,power_law_1.2,0.23359360694885253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,32,2,power_law_1.01,0.2371392011642456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,1,power_law_1.2,0.24214398860931396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,32,2,power_law_1.01,1.7450176239013673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,32,2,power_law_1.01,0.26333439350128174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,1,power_law_1.2,0.25201919078826907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,1,power_law_1.2,0.2686208009719849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,32,2,power_law_1.01,0.31189761161804197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,32,2,power_law_1.01,2.138444709777832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,1,power_law_1.2,0.27393279075622556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,32,2,power_law_1.01,0.36747519969940184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,1,power_law_1.2,0.28840320110321044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,1,power_law_1.2,0.29456000328063964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,32,2,power_law_1.01,0.47317118644714357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,16,balanced,0.03789333254098892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,16,balanced,0.03126933425664902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,16,balanced,0.031301334500312805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,16,balanced,0.03346133232116699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,16,balanced,0.03385066737731298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,16,balanced,0.035642666121323906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,16,balanced,0.03763733307520548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,16,balanced,0.036992001036802925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,16,balanced,0.03774400055408478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,16,balanced,0.035930665830771126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,32,2,power_law_1.01,3.3196990966796873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,16,balanced,0.03745600084463755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,16,balanced,0.03753600021203359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,16,balanced,0.03825599948565165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,16,balanced,0.03944533318281174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,1,power_law_1.2,0.316377592086792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,16,balanced,0.04160533348719279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,16,balanced,0.046682665745417275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,16,balanced,0.04574933151404063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,16,balanced,0.055104002356529236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,16,balanced,0.050069332122802734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,16,balanced,0.06479999919732411
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,16,balanced,0.06663999954859416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,16,balanced,0.08593599994977315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,16,balanced,0.08961600065231323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,16,balanced,0.12317333618799846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,16,balanced,0.1364959975083669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,16,balanced,0.19034665822982788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,32,2,power_law_1.01,0.5871295928955078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,16,balanced,0.2129759987195333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,1,16,balanced,0.25678932666778564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,1,16,balanced,0.37332268555959064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,1,power_law_1.2,0.3420480012893677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,1,16,balanced,0.6936853726704916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,1,power_law_1.2,0.3871615886688232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,32,2,power_law_1.01,0.8248384475708008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,1,power_law_1.2,0.4239359855651855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,32,2,power_law_1.01,1.0128000259399415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,1,power_law_1.2,0.5146111965179443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,32,2,power_law_1.01,6.332102584838867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,1,power_law_1.2,0.5829696178436279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,32,2,power_law_1.01,1.4528256416320802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,1,power_law_1.2,0.7374527931213379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,1,power_law_1.2,0.914367961883545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,32,2,power_law_1.01,1.8880895614624023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,1,power_law_1.2,1.218131160736084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,32,2,power_law_1.01,2.347417640686035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,1,power_law_1.2,1.5105536460876465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,1,power_law_1.2,2.0977407455444337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,32,2,power_law_1.01,3.579795074462891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,1,power_law_1.2,2.7135040283203127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,32,power_law_1.2,0.10122879743576049
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,32,power_law_1.2,0.07579519748687744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,1,power_law_1.2,3.3207935333251952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,8,power_law_1.2,0.040031999349594116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,32,power_law_1.2,0.06371840238571166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,32,2,power_law_1.01,7.012012481689453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,8,power_law_1.2,0.03242239952087402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,1,power_law_1.2,5.117984008789063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,32,power_law_1.2,0.06918399930000305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,8,power_law_1.2,0.033555200695991515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,8,power_law_1.2,0.03676159977912903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,32,power_law_1.2,0.07853440046310425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,8,power_law_1.2,0.03861120045185089
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,32,power_law_1.2,0.09278079867362976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,8,power_law_1.2,0.052019202709198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,8,power_law_1.2,0.05415679812431336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,1,power_law_1.2,9.73012466430664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,32,power_law_1.2,0.09338240027427673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,8,power_law_1.2,0.05556480288505554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,8,power_law_1.2,0.05427200198173523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,32,power_law_1.2,0.09576320052146911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,8,power_law_1.2,0.05578240156173706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,8,power_law_1.2,0.05588480234146118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,32,power_law_1.2,0.09676160216331482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,8,power_law_1.2,0.05856000185012818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,8,power_law_1.2,0.0581055998802185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,32,power_law_1.2,0.09833599925041199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,8,power_law_1.2,0.06235520243644714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,16,power_law_1.01,0.07551360130310059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,8,power_law_1.2,0.06974719762802124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,32,power_law_1.2,0.10133119821548461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,16,power_law_1.01,0.04976640045642853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,8,power_law_1.2,0.07221760153770447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,16,power_law_1.01,0.04938240051269531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,8,power_law_1.2,0.0706816017627716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,32,power_law_1.2,0.10609920024871826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,16,power_law_1.01,0.05246080160140991
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,8,power_law_1.2,0.08234879970550538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,16,power_law_1.01,0.05751039981842041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,8,power_law_1.2,0.08636159896850586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,16,power_law_1.01,0.07122560143470764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,32,power_law_1.2,0.10609920024871826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,16,power_law_1.01,0.07188479900360108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,8,power_law_1.2,0.10306559801101685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,16,power_law_1.01,0.07179520130157471
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,8,power_law_1.2,0.11800320148468017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,32,power_law_1.2,0.11132800579071045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,16,power_law_1.01,0.07445120215415954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,16,power_law_1.01,0.07827200293540955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,8,power_law_1.2,0.15893759727478027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,16,power_law_1.01,0.07751039862632751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,8,power_law_1.2,0.20083200931549072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,16,power_law_1.01,0.08475520014762879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,32,power_law_1.2,0.11438080072402954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,16,power_law_1.01,0.08721280097961426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,8,power_law_1.2,0.22892160415649415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,16,power_law_1.01,0.08103039860725403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,16,balanced,0.031925333042939506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,16,power_law_1.01,0.08919680118560791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,16,balanced,0.029893333713213604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,32,power_law_1.2,0.11677440404891967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,16,balanced,0.031504000226656594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,16,balanced,0.03019733230272929
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,16,balanced,0.032058666149775185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,16,balanced,0.033887999753157295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,16,balanced,0.033887999753157295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,8,power_law_1.2,0.3530303955078125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,16,balanced,0.03362133353948593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,16,balanced,0.03374933451414108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,16,balanced,0.03385066737731298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,16,power_law_1.01,0.08865280151367187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,16,balanced,0.03405333310365677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,16,balanced,0.035674666364987694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,16,balanced,0.03369600077470144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,16,balanced,0.03575466573238373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,16,balanced,0.04004266609748205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,16,balanced,0.03996799886226654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,16,balanced,0.03952533255020777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,16,power_law_1.01,0.09042559862136841
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,16,balanced,0.04181333382924398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,16,balanced,0.044138665000597634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,16,balanced,0.04645866652329763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,16,balanced,0.052042668064435325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,16,balanced,0.0602400004863739
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,8,power_law_1.2,0.45306878089904784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,16,balanced,0.06745600203673045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,16,power_law_1.01,0.10403200387954711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,16,balanced,0.08899733424186707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,32,power_law_1.2,0.13201919794082642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,16,balanced,0.1207413375377655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,16,balanced,0.1622933348019918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,16,balanced,0.20253866910934448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,16,balanced,0.2426933248837789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,16,balanced,0.3566186825434367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,16,power_law_1.01,0.11088000535964966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,16,balanced,0.668783982594808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,8,power_law_1.2,0.6620416164398193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,16,power_law_1.01,0.13284480571746826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,32,power_law_1.2,0.15545599460601806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,16,power_law_1.01,0.13059200048446656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,16,power_law_1.01,0.1717120051383972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,8,power_law_1.2,0.7183231830596923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,16,power_law_1.01,0.19004160165786743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,32,power_law_1.2,0.16380159854888915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,16,power_law_1.01,0.2841536045074463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,8,power_law_1.2,1.1944576263427735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,16,power_law_1.01,0.30612480640411377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,32,power_law_1.2,0.2043776035308838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,16,power_law_1.01,0.4812032222747803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,32,power_law_1.2,0.21644799709320067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,16,power_law_1.01,0.5671103954315185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,8,power_law_1.2,2.872723197937012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,16,power_law_1.01,0.6592383861541748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,32,power_law_1.2,0.29914240837097167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,16,power_law_1.01,0.8867839813232422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,32,power_law_1.2,0.2859711885452271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,16,power_law_1.01,1.9102399826049805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,32,power_law_1.2,0.41590399742126466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,8,power_law_1.2,0.10122239589691162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,8,power_law_1.2,0.08052480220794678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,8,power_law_1.2,0.08238080143928528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,32,power_law_1.2,0.51179518699646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,8,power_law_1.2,0.0942080020904541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,8,power_law_1.2,0.1015936017036438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,8,power_law_1.2,0.14147839546203614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,32,power_law_1.2,0.7001920223236084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,8,power_law_1.2,0.163756799697876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,8,power_law_1.2,0.16713600158691405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,8,power_law_1.2,0.16643199920654297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,8,power_law_1.2,0.17336959838867189
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,8,power_law_1.2,0.18240640163421631
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,16,power_law_1.2,0.03857280015945434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,32,power_law_1.2,0.8607487678527832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,8,power_law_1.2,0.18602240085601807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,16,power_law_1.2,0.03088639974594116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,8,power_law_1.2,0.18613120317459106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,8,power_law_1.2,0.19699840545654296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,16,power_law_1.2,0.03089280128479004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,8,power_law_1.2,0.20355839729309083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,8,power_law_1.2,0.21459200382232665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,16,power_law_1.2,0.030553600192070006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,2,32,power_law_1.2,1.0320128440856933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,32,1,power_law_1.01,0.0315200001001358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,32,1,power_law_1.01,0.031948798894882204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,8,power_law_1.2,0.22653439044952392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,32,1,power_law_1.01,0.032179200649261476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,32,1,power_law_1.01,0.03546879887580871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,16,power_law_1.2,0.031219199299812317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,32,1,power_law_1.01,0.03553279936313629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,32,1,power_law_1.01,0.03701759874820709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,8,power_law_1.2,0.26141440868377686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,32,1,power_law_1.01,0.03880960047245026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,32,1,power_law_1.01,0.04110080003738403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,32,1,power_law_1.01,0.043635201454162595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,16,power_law_1.2,0.03297280073165894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,32,1,power_law_1.01,0.044870400428771974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,8,power_law_1.2,0.2778304100036621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,32,1,power_law_1.01,0.04327679872512817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,32,1,power_law_1.01,0.04952960014343262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,32,1,power_law_1.01,0.05118079781532288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,32,1,power_law_1.01,0.05242879986763001
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,8,power_law_1.2,0.35015039443969725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,16,power_law_1.2,0.03261440098285675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,32,1,power_law_1.01,0.06125440001487732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,32,1,power_law_1.01,0.0654528021812439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,32,1,power_law_1.01,0.07395200133323669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,32,1,power_law_1.01,0.08723840117454529
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,8,power_law_1.2,0.39906558990478513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,32,1,power_law_1.01,0.09995520114898682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,16,power_law_1.2,0.03297280073165894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,32,1,power_law_1.01,0.1310912013053894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,32,1,power_law_1.01,0.14396159648895263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,8,power_law_1.2,0.4827392101287842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,32,1,power_law_1.01,0.2037440061569214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,32,1,power_law_1.01,0.24176640510559083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,2,32,power_law_1.2,1.6595136642456054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,16,power_law_1.2,0.033062401413917544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,32,1,power_law_1.01,0.3441087961196899
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,8,power_law_1.2,0.5011648178100586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,32,1,power_law_1.01,0.4400383949279785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,32,1,power_law_1.01,0.636729621887207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,16,power_law_1.2,0.033555200695991515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,32,1,power_law_1.01,0.8207103729248046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,8,power_law_1.2,0.6962111949920654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,32,1,power_law_1.01,1.0157312393188476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,16,power_law_1.2,0.03528960049152374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,8,power_law_1.2,0.9227264404296875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,32,1,power_law_1.01,1.5859328269958497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,16,power_law_1.2,0.03625600039958954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,8,power_law_1.2,1.2631360054016114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,16,power_law_1.2,0.036211198568344115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,32,1,power_law_1.01,3.1162752151489257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,16,power_law_1.2,0.03865599930286408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,8,power_law_1.2,1.8438207626342773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,16,power_law_1.2,0.041657599806785586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,16,power_law_1.2,0.04276480078697205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,8,power_law_1.2,1.9572416305541993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,16,power_law_1.2,0.043007999658584595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,2,32,power_law_1.2,4.169753646850586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,16,power_law_1.2,0.047142401337623596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,16,power_law_1.2,0.05185920000076294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,8,power_law_1.2,3.6784191131591797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,16,power_law_1.2,0.05999360084533691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,16,power_law_1.2,0.06892160177230836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,16,power_law_1.2,0.08919039964675904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,16,power_law_1.2,0.11869440078735352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,16,power_law_1.2,0.16149760484695436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,8,power_law_1.2,6.891884613037109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,16,power_law_1.2,0.20697600841522218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,16,power_law_1.2,0.3407615900039673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,16,power_law_1.2,0.3913536071777344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,16,power_law_1.2,0.5373119831085205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,32,4,power_law_1.2,0.04865919947624207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,32,4,power_law_1.2,0.04940159916877747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,16,power_law_1.2,0.8584447860717773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,32,4,power_law_1.2,0.049420800805091855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,32,4,power_law_1.2,0.033241599798202515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,32,4,power_law_1.2,0.05051519870758057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,32,4,power_law_1.2,0.029862400889396668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,16,power_law_1.2,2.28100471496582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,32,4,power_law_1.2,0.02865920066833496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,32,4,power_law_1.2,0.055776000022888184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,32,4,power_law_1.2,0.03136000037193298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,32,4,power_law_1.2,0.03330560028553009
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,32,4,power_law_1.2,0.036697599291801455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,32,4,power_law_1.2,0.06540160179138184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,32,4,power_law_1.2,0.03746559917926788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,32,4,power_law_1.2,0.03804160058498383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,32,4,power_law_1.2,0.06975359916687011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,32,4,power_law_1.2,0.03781760036945343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,16,power_law_1.2,0.05514240264892578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,32,4,power_law_1.2,0.03891200125217438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,16,power_law_1.2,0.05263360142707825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,32,4,power_law_1.2,0.03959679901599884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,32,4,power_law_1.2,0.0812287986278534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,16,power_law_1.2,0.04893440008163452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,32,4,power_law_1.2,0.04071039855480194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,32,4,power_law_1.2,0.041280001401901245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,16,power_law_1.2,0.0506879985332489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,32,4,power_law_1.2,0.0813759982585907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,32,4,power_law_1.2,0.04472959935665131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,16,power_law_1.2,0.05141119956970215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,32,4,power_law_1.2,0.04937599897384644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,16,power_law_1.2,0.05498239994049072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,32,4,power_law_1.2,0.051846402883529666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,32,4,power_law_1.2,0.08599039912223816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,32,4,power_law_1.2,0.05931519865989685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,16,power_law_1.2,0.05537279844284058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,32,4,power_law_1.2,0.0677183985710144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,16,power_law_1.2,0.05720319747924805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,32,4,power_law_1.2,0.08889600038528442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,32,4,power_law_1.2,0.07568640112876893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,16,power_law_1.2,0.05633280277252197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,32,4,power_law_1.2,0.09436799883842469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,16,power_law_1.2,0.05719040036201477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,32,4,power_law_1.2,0.09457280039787293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,32,4,power_law_1.2,0.11796480417251587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,16,power_law_1.2,0.058329600095748904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,32,4,power_law_1.2,0.14570879936218262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,16,power_law_1.2,0.05868160128593445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,32,4,power_law_1.2,0.0972927987575531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,32,4,power_law_1.2,0.2098304033279419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,16,power_law_1.2,0.059334397315979004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,32,4,power_law_1.2,0.2850944042205811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,1,4,power_law_1.2,0.2450495958328247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,32,4,power_law_1.2,0.10615040063858032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,16,power_law_1.2,0.0673919975757599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,1,4,power_law_1.2,0.18065919876098632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,32,4,power_law_1.2,0.3741631984710693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,16,power_law_1.2,0.07394559979438782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,1,4,power_law_1.2,0.18408960103988647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,1,4,power_law_1.2,0.21473920345306396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,32,4,power_law_1.2,0.115283203125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,16,power_law_1.2,0.07466880083084107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,32,4,power_law_1.2,0.6053952217102051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,1,4,power_law_1.2,0.21665279865264891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,1,4,power_law_1.2,0.22536959648132324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,16,power_law_1.2,0.08131200075149536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,1,4,power_law_1.2,0.2357759952545166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,32,4,power_law_1.2,0.12104959487915039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,32,4,power_law_1.2,0.7114687919616699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,16,power_law_1.2,0.09394559860229493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,1,4,power_law_1.2,0.23331840038299562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,1,4,power_law_1.2,0.23297278881072997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,16,power_law_1.2,0.10298240184783936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,1,4,power_law_1.2,0.2527424097061157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,32,4,power_law_1.2,0.13616640567779542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,32,4,power_law_1.2,0.8867903709411621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,1,4,power_law_1.2,0.24034559726715088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,16,power_law_1.2,0.12506879568099977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,1,4,power_law_1.2,0.27455360889434816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,1,4,power_law_1.2,0.2810175895690918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,16,power_law_1.2,0.14597760438919066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,1,4,power_law_1.2,0.3205312013626099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,32,4,power_law_1.2,0.16124800443649293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,1,4,power_law_1.2,0.3523008108139038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,32,4,power_law_1.2,1.4222528457641601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,16,power_law_1.2,0.18499200344085692
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,1,4,power_law_1.2,0.35941760540008544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,1,4,power_law_1.2,0.37022080421447756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,32,4,power_law_1.2,0.16707839965820312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,16,power_law_1.2,0.23411200046539307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,1,4,power_law_1.2,0.48040318489074707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,1,4,power_law_1.2,0.562828779220581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,1,4,power_law_1.2,0.7666368007659912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,16,power_law_1.2,0.3269695997238159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,32,4,power_law_1.2,0.20945920944213867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,1,4,power_law_1.2,0.9999551773071289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,16,power_law_1.2,0.3998656034469604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,1,4,power_law_1.2,1.4257535934448242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,32,4,power_law_1.2,3.194592094421387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,32,4,power_law_1.2,0.2516671895980835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,1,4,power_law_1.2,2.20263671875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,16,power_law_1.2,0.6276991844177247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,1,4,power_law_1.2,2.6738815307617188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,32,4,power_law_1.2,0.3351936101913452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,16,power_law_1.2,0.7802688121795655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,1,4,power_law_1.2,4.117593765258789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,32,4,power_law_1.2,0.42099199295043943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,16,power_law_1.2,0.9087231636047364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,1,4,power_law_1.2,4.461318588256836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,32,4,power_law_1.2,0.5640128135681153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,16,power_law_1.2,1.7171136856079101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,1,4,power_law_1.2,5.563199996948242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,32,4,power_law_1.2,0.7391039848327636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,1,4,power_law_1.2,8.038253021240234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,32,4,power_law_1.2,1.119699192047119
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,16,power_law_1.2,3.2776065826416017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,32,4,power_law_1.2,1.4486592292785645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,1,4,power_law_1.2,15.022694396972657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,32,4,power_law_1.2,1.6734207153320313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,32,4,power_law_1.2,2.577868843078613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,1,4,power_law_1.2,30.163284301757812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,32,4,power_law_1.2,5.199974441528321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,128,power_law_1.2,0.059545600414276124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,128,power_law_1.2,0.05603839755058289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,128,power_law_1.2,0.0564736008644104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,128,power_law_1.2,0.05470719933509827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,128,power_law_1.2,0.05725439786911011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,128,power_law_1.2,0.060159999132156375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,128,power_law_1.2,0.06111999750137329
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,128,power_law_1.2,0.058719998598098753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,128,power_law_1.2,0.05987840294837952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,128,power_law_1.2,0.06164479851722717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,1,8,power_law_1.01,0.12380160093307495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,128,power_law_1.2,0.061849600076675414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,1,8,power_law_1.01,0.12947839498519897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,32,1,balanced,0.03387200087308884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,1,8,power_law_1.01,0.1261504054069519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,32,1,balanced,0.034186666210492454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,128,power_law_1.2,0.0616320013999939
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,32,1,balanced,0.03675200045108795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,32,1,balanced,0.038191998998324074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,32,1,balanced,0.03813866774241129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,32,1,balanced,0.03972266614437103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,32,1,balanced,0.03892799963553747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,1,8,power_law_1.01,0.12442879676818848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,32,1,balanced,0.04002666721741358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,32,1,balanced,0.04173333446184794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,32,1,balanced,0.041637333730856575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,32,1,balanced,0.043578664461771645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,32,1,balanced,0.044026667873064675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,32,1,balanced,0.04491200049718221
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,32,1,balanced,0.05012799799442291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,32,1,balanced,0.06518933176994324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,1,8,power_law_1.01,0.12225919961929321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,32,1,balanced,0.06840000053246816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,32,1,balanced,0.07271466652552287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,32,1,balanced,0.09041066964467366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,128,power_law_1.2,0.06255999803543091
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,32,1,balanced,0.1016426682472229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,32,1,balanced,0.13314132889111838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,1,8,power_law_1.01,0.12848000526428222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,32,1,balanced,0.14614933729171753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,32,1,balanced,0.20432533820470175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,32,1,balanced,0.24048000574111938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,32,1,balanced,0.34641067186991376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,1,8,power_law_1.01,0.16622719764709473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,32,1,balanced,0.4437439839045207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,32,1,balanced,0.6397973299026489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,128,power_law_1.2,0.06302080154418946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,32,1,balanced,0.8228426774342855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,1,8,power_law_1.01,0.14720640182495118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,32,1,balanced,1.0217866897583008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,1,8,power_law_1.01,0.16949119567871093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,32,1,balanced,1.5877973238627117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,1,8,power_law_1.01,0.20088961124420165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,128,power_law_1.2,0.06762880086898804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,1,8,power_law_1.01,0.15022079944610595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,32,1,balanced,3.107050577799479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,1,8,power_law_1.01,0.16984319686889648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,1,8,power_law_1.01,0.1717311978340149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,128,power_law_1.2,0.06922879815101624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,1,8,power_law_1.01,0.23045120239257813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,1,8,power_law_1.01,0.30208640098571776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,128,power_law_1.2,0.07653759717941284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,1,8,power_law_1.01,0.31845118999481203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,1,8,power_law_1.01,0.26158719062805175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,1,8,power_law_1.01,0.33137919902801516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,128,power_law_1.2,0.08602240085601806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,1,8,power_law_1.01,0.4032320022583008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,1,8,power_law_1.01,0.5408959865570069
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,128,power_law_1.2,0.09669119715690613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,1,8,power_law_1.01,0.6650368213653565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,1,8,power_law_1.01,0.9232704162597656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,128,power_law_1.2,0.1244928002357483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,1,8,power_law_1.01,1.2207615852355957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,1,8,power_law_1.01,1.8700799942016602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,128,power_law_1.2,0.14785280227661132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,1,8,power_law_1.01,2.322975921630859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,128,power_law_1.2,0.19716479778289794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,1,8,power_law_1.01,3.5961280822753907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,128,power_law_1.2,0.24225280284881592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,1,8,power_law_1.01,4.815193557739258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,128,power_law_1.2,0.3507136106491089
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,128,power_law_1.2,0.44606719017028806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,1,8,power_law_1.01,5.929068756103516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,128,power_law_1.2,0.6323967933654785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,128,power_law_1.01,0.034867200255393985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,128,power_law_1.01,0.03452799916267395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,1,8,power_law_1.01,9.319974517822265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,128,power_law_1.2,0.9207679748535156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,128,power_law_1.01,0.034764799475669864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,128,power_law_1.01,0.034668800234794614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,128,power_law_1.01,0.03482879996299744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,128,power_law_1.01,0.036051198840141296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,128,power_law_1.2,1.0908927917480469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,128,power_law_1.01,0.0359360009431839
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,128,power_law_1.01,0.03562879860401154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,128,power_law_1.01,0.036422398686408994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,128,power_law_1.01,0.036748799681663516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,128,power_law_1.2,1.7589056015014648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,128,power_law_1.01,0.03752320110797882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,1,8,power_law_1.01,19.537298583984374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,128,power_law_1.01,0.03785600066184998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,128,power_law_1.01,0.03802880048751831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,128,power_law_1.01,0.03845759928226471
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,128,power_law_1.01,0.04174720048904419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,128,power_law_1.01,0.04229759871959686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,128,power_law_1.2,3.7853057861328123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,128,power_law_1.01,0.04307200014591217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,128,power_law_1.01,0.04460160136222839
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,128,power_law_1.01,0.04644480049610138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,128,power_law_1.01,0.05248640179634094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,128,power_law_1.01,0.05555199980735779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,128,power_law_1.01,0.06839680075645446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,128,power_law_1.01,0.0819711983203888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,128,power_law_1.01,0.11608320474624634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,128,power_law_1.01,0.1471168041229248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,128,power_law_1.01,0.21279358863830566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,128,power_law_1.01,0.26387200355529783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,1,128,power_law_1.01,0.312608003616333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,1,128,power_law_1.01,0.4806784152984619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,16,power_law_1.2,0.15869439840316774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,1,128,power_law_1.01,1.0245887756347656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,8,balanced,0.052442664901415505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,8,balanced,0.0572320024172465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,16,power_law_1.2,0.10022399425506592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,8,balanced,0.07495999832948048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,8,balanced,0.09752532839775085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,8,balanced,0.12944533427556357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,8,balanced,0.15777066349983215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,8,balanced,0.15619200468063354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,8,balanced,0.15929599603017172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,8,balanced,0.16013333201408386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,8,balanced,0.15853333473205566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,8,balanced,0.1634773313999176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,8,balanced,0.16446399688720703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,8,balanced,0.16503467162450156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,8,balanced,0.16959466536839804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,8,balanced,0.17227200667063394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,8,balanced,0.17669866482416788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,8,balanced,0.184005339940389
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,8,balanced,0.20305599768956503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,16,power_law_1.2,0.0821120023727417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,8,balanced,0.20794665813446045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,8,balanced,0.28679466247558594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,8,balanced,0.2702293395996094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,8,balanced,0.3693759838740031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,8,balanced,0.3636159896850586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,8,balanced,0.5173493226369222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,8,balanced,0.5818986495335897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,8,balanced,0.8481973012288412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,16,power_law_1.2,0.09834880232810975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,8,balanced,0.9753759702046713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,2,8,balanced,1.2309119701385498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,2,8,balanced,1.839786688486735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,16,power_law_1.2,0.12261760234832764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,2,8,balanced,3.6061280568440757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,16,power_law_1.2,0.15819519758224487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,16,power_law_1.2,0.1629631996154785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,16,power_law_1.2,0.17722879648208617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,16,power_law_1.2,0.1734976053237915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,16,power_law_1.2,0.1786687970161438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,16,power_law_1.2,0.17607680559158326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,16,power_law_1.2,0.1861567974090576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,16,power_law_1.2,0.18880000114440917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,8,8,power_law_1.2,0.03386879861354828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,8,8,power_law_1.2,0.034995201230049136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,8,8,power_law_1.2,0.03139840066432953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,8,8,power_law_1.2,0.030476799607276915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,16,power_law_1.2,0.19422719478607178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,8,8,power_law_1.2,0.030015999078750612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,8,8,power_law_1.2,0.031065601110458373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,8,8,power_law_1.2,0.03240320086479187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,8,8,power_law_1.2,0.03299199938774109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,8,8,power_law_1.2,0.035596799850463864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,8,8,power_law_1.2,0.03828479945659637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,16,power_law_1.2,0.20373120307922363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,8,8,power_law_1.2,0.03446399867534637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,8,8,power_law_1.2,0.040729600191116336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,8,8,power_law_1.2,0.04068480134010315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,8,8,power_law_1.2,0.03854719996452331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,8,8,power_law_1.2,0.05015680193901062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,16,power_law_1.2,0.21246719360351562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,8,8,power_law_1.2,0.052223998308181765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,8,8,power_law_1.2,0.05571200251579285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,8,8,power_law_1.2,0.06437119841575623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,8,8,power_law_1.2,0.07055360078811646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,8,8,power_law_1.2,0.08845440149307252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,16,power_law_1.2,0.2310784101486206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,8,8,power_law_1.2,0.10334080457687378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,8,8,power_law_1.2,0.1401792049407959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,8,8,power_law_1.2,0.15745279788970948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,8,8,power_law_1.2,0.21868159770965576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,16,power_law_1.2,0.2776959896087646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,8,8,power_law_1.2,0.2541311979293823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,8,8,power_law_1.2,0.3862015962600708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,2,power_law_1.01,0.0541375994682312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,8,8,power_law_1.2,0.49185919761657715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,16,power_law_1.2,0.2824320077896118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,2,power_law_1.01,0.06476799845695495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,8,8,power_law_1.2,0.6152192115783691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,2,power_law_1.01,0.06756479740142822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,8,8,power_law_1.2,0.8732864379882812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,16,power_law_1.2,0.35177600383758545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,2,power_law_1.01,0.09563519954681396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,2,power_law_1.01,0.11559040546417236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,8,8,power_law_1.2,1.6793472290039062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,16,power_law_1.2,0.36970880031585696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,2,power_law_1.01,0.17173759937286376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,2,power_law_1.01,0.18465919494628907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,16,power_law_1.2,0.49319682121276853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,2,power_law_1.01,0.18993279933929444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,16,power_law_1.2,0.4982592105865479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,2,power_law_1.01,0.1910591959953308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,2,power_law_1.01,0.19340159893035888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,16,power_law_1.2,0.7340288162231445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,2,power_law_1.01,0.1901695966720581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,2,power_law_1.01,0.20280320644378663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,2,power_law_1.01,0.2003711938858032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,16,power_law_1.2,0.9487360000610352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,2,power_law_1.01,0.210316801071167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,2,power_law_1.01,0.22412800788879395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,16,power_law_1.2,1.2386879920959473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,2,power_law_1.01,0.23187201023101806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,2,power_law_1.01,0.22649600505828857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,2,power_law_1.01,0.24492158889770507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,16,power_law_1.2,1.7519487380981444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,2,power_law_1.01,0.2707520008087158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,2,power_law_1.01,0.33747200965881347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,2,power_law_1.01,0.33505280017852784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,16,power_law_1.2,2.077145576477051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,2,power_law_1.01,0.41199359893798826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,2,power_law_1.01,0.48828802108764646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,4,balanced,0.07076266904671986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,2,power_law_1.01,0.6490047931671142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,4,balanced,0.0920799970626831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,4,balanced,0.129530668258667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,4,balanced,0.20308266083399454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,4,balanced,0.33671998977661133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,16,power_law_1.2,3.327008056640625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,4,balanced,0.44972801208496094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,4,balanced,0.4525119860967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,4,balanced,0.4538826545079549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,4,balanced,0.45523734887441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,4,balanced,0.45772266387939453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,2,power_law_1.01,0.7922239780426026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,4,balanced,0.46207467714945477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,4,balanced,0.46505598227183026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,4,balanced,0.4657653172810872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,4,balanced,0.47180267175038654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,4,balanced,0.4795200030008952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,4,balanced,0.48369598388671875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,4,balanced,0.49273065725962323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,4,balanced,0.5979839960734049
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,2,power_law_1.01,1.0569408416748047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,4,balanced,0.5353759924570719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,4,balanced,0.6799413363138834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,4,balanced,0.6245333353678385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,4,balanced,1.0012426376342773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,4,balanced,0.7742239634195963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,4,balanced,1.1780746777852376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,4,balanced,1.1679786841074626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,2,power_law_1.01,1.4211071968078612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,4,balanced,1.7334613800048828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,4,balanced,1.8874880472819011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,16,power_law_1.2,5.7805633544921875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,1,4,balanced,2.3875892957051597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,2,power_law_1.01,1.716454315185547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,1,4,balanced,3.475717226664225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,1,4,balanced,6.783344268798828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,2,power_law_1.01,2.490483283996582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,2,power_law_1.01,5.254828643798828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,128,balanced,0.06424533327420552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,128,balanced,0.06457066535949707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,128,balanced,0.06330133477846782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,128,balanced,0.06482133269309998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,128,balanced,0.0641653339068095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,128,balanced,0.07291733225186665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,128,balanced,0.08405866225560506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,128,balanced,0.08526399731636047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,128,balanced,0.08916266759236653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,128,balanced,0.08522666494051616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,128,balanced,0.09134933352470398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,128,balanced,0.09635200103123982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,128,balanced,0.08752533793449402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,128,balanced,0.09158933162689209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,128,balanced,0.09755733609199524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,128,balanced,0.0995253324508667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,128,balanced,0.10149866342544556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,128,balanced,0.10971732934315999
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,128,balanced,0.11337600151697795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,128,balanced,0.12584533294041952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,128,balanced,0.135754664738973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,128,balanced,0.16270400087038675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,128,balanced,0.1879253387451172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,128,balanced,0.23469332853953043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,128,balanced,0.2883039911588033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,128,balanced,0.3848479986190796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,128,balanced,0.4819626808166504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,128,balanced,0.6052320003509521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,4,power_law_1.2,0.03535360097885132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,128,balanced,0.9002079963684082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,128,balanced,1.7432907422383626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,4,power_law_1.2,0.030720001459121703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,4,power_law_1.2,0.03116160035133362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,4,power_law_1.2,0.03344640135765076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,4,power_law_1.2,0.03526400029659271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,4,power_law_1.2,0.04072319865226746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,4,power_law_1.2,0.04185599982738495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,4,power_law_1.2,0.04186240136623383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,1,power_law_1.01,0.03338240087032318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,4,power_law_1.2,0.04211840033531189
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,1,power_law_1.01,0.04689919948577881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,4,power_law_1.2,0.04386560022830963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,1,power_law_1.01,0.05550720095634461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,4,power_law_1.2,0.04755200147628784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,1,power_law_1.01,0.07852799892425537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,4,power_law_1.2,0.04764159917831421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,1,power_law_1.01,0.1093440055847168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,4,power_law_1.2,0.051641601324081424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,1,power_law_1.01,0.1632192015647888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,4,power_law_1.2,0.05315200090408325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,1,power_law_1.01,0.1750272035598755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,4,power_law_1.2,0.060736000537872314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,1,power_law_1.01,0.18529280424118041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,4,power_law_1.2,0.06657919883728028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,1,power_law_1.01,0.1866368055343628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,4,power_law_1.2,0.06526079773902893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,1,power_law_1.01,0.19218560457229614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,4,power_law_1.2,0.07334399819374085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,1,power_law_1.01,0.19342080354690552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,4,power_law_1.2,0.09264000058174134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,1,power_law_1.01,0.20096640586853026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,4,power_law_1.2,0.10851839780807496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,1,power_law_1.01,0.19875839948654175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,4,power_law_1.2,0.12707200050354003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,1,power_law_1.01,0.21022720336914064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,4,power_law_1.2,0.16098560094833375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,1,power_law_1.01,0.2203200101852417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,4,power_law_1.2,0.2151360034942627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,1,power_law_1.01,0.22389121055603028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,4,power_law_1.2,0.29964799880981446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,1,power_law_1.01,0.23582079410552978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,1,power_law_1.01,0.26552960872650144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,4,power_law_1.2,0.35729279518127444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,16,power_law_1.01,0.04307839870452881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,1,power_law_1.01,0.26752638816833496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,4,power_law_1.2,0.6351679801940918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,1,power_law_1.01,0.31703040599822996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,16,power_law_1.01,0.03785600066184998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,1,power_law_1.01,0.3313215970993042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,4,power_law_1.2,0.6494656085968018
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,1,power_law_1.01,0.3928128004074097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,16,power_law_1.01,0.036595198512077334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,1,power_law_1.01,0.45774078369140625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,4,power_law_1.2,0.858028793334961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,16,power_law_1.01,0.037529599666595456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,1,power_law_1.01,0.5688511848449707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,16,power_law_1.01,0.03743999898433685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,1,power_law_1.01,0.6585216045379638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,4,power_law_1.2,1.5749247550964356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,16,power_law_1.01,0.03980799913406372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,1,power_law_1.01,0.9019071578979492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,16,power_law_1.01,0.03939839899539947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,1,power_law_1.01,1.1397439956665039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,16,power_law_1.01,0.03941119909286499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,2,1,power_law_1.01,1.3709376335144043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,4,power_law_1.2,2.7939712524414064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,16,power_law_1.01,0.03979519903659821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,16,power_law_1.01,0.040057599544525146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,2,1,power_law_1.01,2.0656320571899416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,16,power_law_1.01,0.040915200114250184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,16,power_law_1.01,0.0450111985206604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,16,power_law_1.01,0.043347200751304625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,2,1,power_law_1.01,4.017644882202148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,16,power_law_1.01,0.04565120041370392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,16,power_law_1.01,0.048742398619651794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,16,power_law_1.01,0.05013759732246399
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,16,power_law_1.01,0.05427200198173523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,16,power_law_1.01,0.05854719877243042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,16,power_law_1.01,0.061286401748657224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,16,power_law_1.01,0.0747327983379364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,16,power_law_1.01,0.08555520176887513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,16,power_law_1.01,0.10119680166244507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,16,power_law_1.01,0.12806400060653686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,32,balanced,0.02977599948644638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,16,power_law_1.01,0.18616960048675538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,32,balanced,0.03002133220434189
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,32,balanced,0.029989334444204967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,32,balanced,0.02937600016593933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,32,balanced,0.03146666785081228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,32,balanced,0.03332266708215078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,32,balanced,0.031957333286603294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,32,balanced,0.03257066756486893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,32,balanced,0.03148799886306127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,32,balanced,0.033701332906881966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,32,balanced,0.033557333052158356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,32,balanced,0.033359999457995095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,32,balanced,0.035642666121323906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,32,balanced,0.03426666557788849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,32,balanced,0.03933866570393244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,32,balanced,0.038378665844599404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,32,balanced,0.039359999199708305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,32,balanced,0.03965333352486292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,32,balanced,0.041349334021409355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,32,balanced,0.043749332427978516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,16,power_law_1.01,0.22439041137695312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,32,balanced,0.045226668318112694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,32,balanced,0.051914667089780174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,32,balanced,0.0565226674079895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,32,balanced,0.07774400214354198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,32,balanced,0.103685329357783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,32,balanced,0.14219733079274496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,32,balanced,0.169978658358256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,32,balanced,0.2040533423423767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,32,balanced,0.29101866483688354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,32,balanced,0.5248479843139648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,16,power_law_1.01,0.309113597869873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,16,power_law_1.01,0.40549120903015134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,1,1,balanced,0.09852799773216248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,1,1,balanced,0.19899199406305948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,1,1,balanced,0.24499199787775675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,1,1,balanced,0.24304000536600748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,1,1,balanced,0.24659200509389242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,1,1,balanced,0.24649600187937418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,1,1,balanced,0.24911999702453613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,1,1,balanced,0.2513920068740845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,16,power_law_1.01,0.5636544227600098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,1,1,balanced,0.25119467576344806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,1,1,balanced,0.25285865863164264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,1,1,balanced,0.256058673063914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,1,1,balanced,0.2582719922065735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,1,1,balanced,0.2648746569951375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,1,1,balanced,0.27461334069569904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,1,1,balanced,0.28620266914367676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,1,1,balanced,0.2919093370437622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,1,1,balanced,0.30509332815806073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,1,1,balanced,0.37281068166097003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,1,1,balanced,0.3893440167109172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,1,1,balanced,0.5070079962412516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,1,1,balanced,0.5877600113550822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,1,1,balanced,0.8012746969858805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,1,1,balanced,1.0078986485799153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,16,power_law_1.01,0.7262271881103516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,1,1,balanced,1.4663732846577961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,1,1,balanced,1.886474609375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,1,1,balanced,2.8282346725463867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,1,1,balanced,3.611584027608236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,1,1,balanced,4.470922787984212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,16,power_law_1.01,1.5444095611572266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,1,1,balanced,6.960917154947917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,1,1,balanced,13.542255401611328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,128,balanced,0.037605332831541695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,128,balanced,0.03557866563399633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,128,balanced,0.03775466730197271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,128,balanced,0.03754133234421412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,128,balanced,0.03793066740036011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,128,balanced,0.037530665596326195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,128,balanced,0.03769599894682566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,128,balanced,0.037920000652472176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,128,balanced,0.03775999943415324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,128,balanced,0.0396373321612676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,128,balanced,0.03790933390458425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,128,balanced,0.03966933240493139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,128,balanced,0.03967999915281931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,128,balanced,0.039877332746982574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,128,balanced,0.04370133578777313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,128,balanced,0.043706665436426796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,128,balanced,0.04367466767628988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,128,balanced,0.04375466704368591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,128,balanced,0.04412800073623657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,128,balanced,0.04785599807898203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,128,balanced,0.04806933303674062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,128,balanced,0.052042668064435325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,128,balanced,0.05407999952634176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,128,balanced,0.06295999884605408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,128,balanced,0.06834133466084798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,128,balanced,0.08256533245245616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,128,balanced,0.09914132952690125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,2,128,balanced,0.11556800206502278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,2,128,balanced,0.1562879979610443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,2,128,balanced,0.28067199389139813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,4,power_law_1.2,0.08774399757385254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,4,power_law_1.2,0.10924160480499268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,8,1,balanced,0.033759998778502144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,8,1,balanced,0.037471999724706016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,8,1,balanced,0.05829866727193197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,8,1,balanced,0.05907199780146281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,8,1,balanced,0.059546664357185364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,8,1,balanced,0.060378665725390114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,8,1,balanced,0.061386664708455406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,8,1,balanced,0.0603359987338384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,8,1,balanced,0.061754668752352394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,8,1,balanced,0.06232533355553945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,8,1,balanced,0.06304533282915752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,8,1,balanced,0.0653599997361501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,8,1,balanced,0.06539733211199443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,8,1,balanced,0.07117333511511485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,8,1,balanced,0.08075733482837677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,8,1,balanced,0.08072000245253245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,4,power_law_1.2,0.07994880080223084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,8,1,balanced,0.08514666557312012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,8,1,balanced,0.10227732857068379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,8,1,balanced,0.10621866583824158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,8,1,balanced,0.1334933340549469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,8,1,balanced,0.16208533445994058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,8,1,balanced,0.21296000480651855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,8,1,balanced,0.2513279914855957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,8,1,balanced,0.3555946747461955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,8,1,balanced,0.45211732387542725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,8,1,balanced,0.6470293203989664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,4,power_law_1.2,0.10792959928512573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,8,1,balanced,0.8339786529541016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,8,1,balanced,1.0257919629414876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,8,1,balanced,1.6009227434794109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,4,power_law_1.2,0.12854399681091308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,8,1,balanced,3.146005312601725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,4,power_law_1.2,0.16112639904022216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,4,power_law_1.2,0.21461119651794433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,4,power_law_1.2,0.232422399520874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,4,power_law_1.2,0.23840639591217042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,4,power_law_1.2,0.23588480949401855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,4,power_law_1.2,0.2535423994064331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,4,power_law_1.2,0.2650496006011963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,4,power_law_1.2,0.27215359210968015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,4,power_law_1.2,0.27990400791168213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,4,power_law_1.2,0.2880959987640381
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,4,power_law_1.2,0.2971456050872803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,4,power_law_1.2,0.31288321018218995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,16,8,power_law_1.01,0.037887999415397645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,16,8,power_law_1.01,0.03640959858894348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,16,8,power_law_1.01,0.034585601091384886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,16,8,power_law_1.01,0.034169599413871765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,4,power_law_1.2,0.3450432062149048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,16,8,power_law_1.01,0.034431999921798705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,16,8,power_law_1.01,0.03585279881954193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,16,8,power_law_1.01,0.03603839874267578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,16,8,power_law_1.01,0.03651840090751648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,16,8,power_law_1.01,0.03711999952793121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,4,power_law_1.2,0.3637120008468628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,16,8,power_law_1.01,0.037011200189590455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,16,8,power_law_1.01,0.03739520013332367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,16,8,power_law_1.01,0.0379967987537384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,16,8,power_law_1.01,0.038771200180053714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,16,8,power_law_1.01,0.04044800102710724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,16,8,power_law_1.01,0.04503679871559143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,4,power_law_1.2,0.4339583873748779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,16,8,power_law_1.01,0.04612480103969574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,16,8,power_law_1.01,0.05117440223693848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,16,8,power_law_1.01,0.05943040251731872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,16,8,power_law_1.01,0.06361600160598754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,16,8,power_law_1.01,0.06883199810981751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,4,power_law_1.2,0.4774655818939209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,16,8,power_law_1.01,0.0921280026435852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,16,8,power_law_1.01,0.11779199838638306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,16,8,power_law_1.01,0.14584959745407106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,16,8,power_law_1.01,0.18453119993209838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,4,power_law_1.2,0.5758912086486816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,16,8,power_law_1.01,0.23369600772857665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,16,8,power_law_1.01,0.33984639644622805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,16,8,power_law_1.01,0.4163839817047119
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,4,power_law_1.2,0.6668543815612793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,16,8,power_law_1.01,0.5137919902801513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,16,8,power_law_1.01,0.7704639911651612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,4,power_law_1.2,0.8632384300231933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,16,8,power_law_1.01,1.5393152236938477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,4,power_law_1.2,1.0412991523742676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,4,power_law_1.2,1.5062911987304688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,4,power_law_1.2,2.1258752822875975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,4,power_law_1.2,2.221887969970703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,4,power_law_1.2,3.3577022552490234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,4,power_law_1.2,7.030540466308594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,4,power_law_1.2,0.049369600415229795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,4,power_law_1.2,0.07275519967079162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,4,power_law_1.2,0.07164160013198853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,4,power_law_1.2,0.07288320064544677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,4,power_law_1.2,0.0877568006515503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,4,power_law_1.2,0.10643839836120605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,4,power_law_1.2,0.11962239742279053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,4,power_law_1.2,0.14468480348587037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,4,power_law_1.2,0.14634239673614502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,16,1,balanced,0.03148266673088074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,16,1,balanced,0.03568000098069509
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,16,1,balanced,0.05794133245944977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,16,1,balanced,0.058431997895240784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,16,1,balanced,0.05771199862162272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,16,1,balanced,0.056602666775385536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,16,1,balanced,0.05890666445096334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,16,1,balanced,0.06043200194835663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,16,1,balanced,0.058549334605534874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,16,1,balanced,0.06007466713587443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,16,1,balanced,0.059877331058184304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,16,1,balanced,0.06159999966621399
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,16,1,balanced,0.06205333272616068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,16,1,balanced,0.06588799754778545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,16,1,balanced,0.07467733323574066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,16,1,balanced,0.07654933134714763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,16,1,balanced,0.0784693310658137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,16,1,balanced,0.09282132983207703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,16,1,balanced,0.09481599926948547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,16,1,balanced,0.11708799997965495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,16,1,balanced,0.13889066378275552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,16,1,balanced,0.1797013282775879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,4,power_law_1.2,0.1527616024017334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,16,1,balanced,0.2125493288040161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,16,1,balanced,0.29338665803273517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,16,1,balanced,0.34193066755930585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,16,1,balanced,0.4833279848098755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,16,1,balanced,0.6253866751988729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,16,1,balanced,0.7659253279368082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,16,1,balanced,1.1806399822235107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,4,power_law_1.2,0.16364799737930297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,16,1,balanced,2.3086400032043457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,4,power_law_1.2,0.16942720413208007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,4,power_law_1.2,0.18182400465011597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,4,power_law_1.2,0.18994560241699218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,4,power_law_1.2,0.2009216070175171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,4,power_law_1.2,0.20167040824890137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,4,power_law_1.2,0.2221951961517334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,4,power_law_1.2,0.24785280227661133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,4,power_law_1.2,0.2735680103302002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,4,power_law_1.2,0.31722240447998046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,4,power_law_1.2,0.3709696054458618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,128,power_law_1.2,0.05761280059814453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,4,power_law_1.2,0.4718912124633789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,128,power_law_1.2,0.05863040089607239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,4,power_law_1.2,0.5952064037322998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,128,power_law_1.2,0.049132800102233885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,4,power_law_1.2,0.7708735942840577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,128,power_law_1.2,0.053600001335144046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,4,power_law_1.2,0.927187156677246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,128,power_law_1.2,0.05182719826698303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,4,power_law_1.2,1.3201215744018555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,128,power_law_1.2,0.05653759837150574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,4,power_law_1.2,1.6224895477294923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,128,power_law_1.2,0.057004797458648684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,128,power_law_1.2,0.05541120171546936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,4,power_law_1.2,2.112895965576172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,128,power_law_1.2,0.061388802528381345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,4,power_law_1.2,3.1877248764038084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,128,power_law_1.2,0.060089600086212155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,128,power_law_1.2,0.06621440052986145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,128,power_law_1.2,0.07120640277862549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,4,power_law_1.2,6.195904159545899
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,128,power_law_1.2,0.07360640168190002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,128,power_law_1.2,0.07614079713821412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,128,power_law_1.2,0.08528640270233154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,128,power_law_1.2,0.08861439824104309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,128,power_law_1.2,0.09350399971008301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,128,power_law_1.2,0.1137727975845337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,128,power_law_1.2,0.12042239904403687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,128,balanced,0.04943466683228811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,128,balanced,0.04996799925963084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,128,balanced,0.05040533343950907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,128,balanced,0.05619733532269796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,128,balanced,0.05699199934800466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,128,balanced,0.05653333167235056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,128,balanced,0.05726400017738342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,128,balanced,0.05796800057093302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,128,balanced,0.060090666015942894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,128,balanced,0.06006399790445963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,128,balanced,0.06005866825580597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,128,balanced,0.062074666221936546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,128,balanced,0.062128002444903054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,128,power_law_1.2,0.1525056004524231
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,128,balanced,0.06620266536871593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,128,balanced,0.06685333450635274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,128,balanced,0.0812906672557195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,128,balanced,0.07649066547552745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,128,balanced,0.09698133667310078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,128,balanced,0.09928533434867859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,128,balanced,0.12782933314641318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,128,balanced,0.13794133067131042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,128,balanced,0.18394132455190024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,128,balanced,0.17917867501576742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,128,balanced,0.22998400529225668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,128,balanced,0.26734399795532227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,128,balanced,0.36403199036916095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,128,balanced,0.43829333782196045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,1,128,balanced,0.5293920040130615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,128,power_law_1.2,0.16747519969940186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,1,128,balanced,0.7850399812062582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,1,128,balanced,1.6400426228841145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,128,power_law_1.2,0.2190783977508545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,128,power_law_1.2,0.24558720588684083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,128,power_law_1.2,0.3335616111755371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,128,power_law_1.2,0.38401920795440675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,128,power_law_1.2,0.5870528221130371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,1,power_law_1.01,0.06584960222244263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,1,power_law_1.01,0.07616000175476074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,128,power_law_1.2,0.706112003326416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,1,power_law_1.01,0.09244800209999085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,1,power_law_1.01,0.12028800249099732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,1,power_law_1.01,0.147980797290802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,1,power_law_1.01,0.21395840644836425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,1,power_law_1.01,0.23257598876953126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,1,power_law_1.01,0.24294400215148926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,1,128,power_law_1.2,0.815283203125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,1,power_law_1.01,0.2424448013305664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,1,power_law_1.01,0.25000319480895994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,1,power_law_1.01,0.25915520191192626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,1,power_law_1.01,0.26771841049194334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,1,power_law_1.01,0.2664704084396362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,1,power_law_1.01,0.2832576036453247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,1,power_law_1.01,0.29171841144561766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,1,128,power_law_1.2,1.5646592140197755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,1,power_law_1.01,0.30176639556884766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,1,power_law_1.01,0.3237951993942261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,1,power_law_1.01,0.37119359970092775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,1,power_law_1.01,0.40984320640563965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,1,power_law_1.01,0.5045375823974609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,1,power_law_1.01,0.580403184890747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,1,power_law_1.01,0.7600704193115234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,1,power_law_1.01,0.9270976066589356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,1,power_law_1.01,1.267520046234131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,1,128,power_law_1.2,3.8739585876464844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,1,power_law_1.01,1.7650495529174806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,1,power_law_1.01,2.4069631576538084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,1,power_law_1.01,2.9875904083251954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,1,power_law_1.01,3.6621761322021484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,1,power_law_1.01,5.610291290283203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,4,balanced,0.05463466544946035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,4,balanced,0.056202664971351624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,4,balanced,0.05736533304055532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,4,balanced,0.06440000236034393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,4,balanced,0.07872533301512401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,4,balanced,0.11767466862996419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,4,balanced,0.11751466989517212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,4,balanced,0.11813333630561829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,4,balanced,0.12178132931391399
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,4,balanced,0.12148267030715942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,4,balanced,0.11997866630554199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,4,balanced,0.1222773293654124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,4,balanced,0.12335999806722005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,4,balanced,0.1274186670780182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,4,balanced,0.1376159985860189
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,4,balanced,0.13808000087738037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,4,balanced,0.13794133067131042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,4,balanced,0.15269333124160767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,4,balanced,0.16205333669980368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,4,balanced,0.18648000558217367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,1,power_law_1.01,10.968064117431641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,4,balanced,0.20920532941818237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,4,balanced,0.256111999352773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,4,balanced,0.30714666843414307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,4,balanced,0.4373439947764079
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,4,balanced,0.5423786640167236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,4,balanced,0.783349355061849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,4,balanced,1.0105066299438477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,4,balanced,1.2559680143992107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,4,balanced,1.8908586502075195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,4,balanced,3.6632372538248696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,16,power_law_1.01,0.06769919991493226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,16,power_law_1.01,0.08309119939804077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,16,power_law_1.01,0.050271999835968015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,16,power_law_1.01,0.056543999910354616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,8,power_law_1.2,0.06485120058059693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,16,power_law_1.01,0.05973119735717773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,8,power_law_1.2,0.08456959724426269
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,16,power_law_1.01,0.06664320230484008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,16,power_law_1.01,0.07227519750595093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,8,power_law_1.2,0.05795199871063232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,16,power_law_1.01,0.07461760044097901
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,8,power_law_1.2,0.0661952018737793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,16,power_law_1.01,0.07585279941558838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,16,power_law_1.01,0.07463039755821228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,8,power_law_1.2,0.07747840285301208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,16,power_law_1.01,0.0770687997341156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,8,power_law_1.2,0.08618879914283753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,16,power_law_1.01,0.07637119889259339
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,16,power_law_1.01,0.0777728021144867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,8,power_law_1.2,0.1007040023803711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,16,power_law_1.01,0.08062719702720642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,8,power_law_1.2,0.1034816026687622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,16,power_law_1.01,0.08601599931716919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,16,power_law_1.01,0.08792319893836975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,8,power_law_1.2,0.10617599487304688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,16,power_law_1.01,0.09281920194625855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,8,power_law_1.2,0.11066240072250366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,16,power_law_1.01,0.10309760570526123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,16,power_law_1.01,0.1159168004989624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,8,power_law_1.2,0.10769920349121094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,16,power_law_1.01,0.13483519554138185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,8,power_law_1.2,0.11247999668121338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,16,power_law_1.01,0.14755200147628783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,8,power_law_1.2,0.11596800088882446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,16,power_law_1.01,0.1902143955230713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,8,power_law_1.2,0.11739519834518433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,16,power_law_1.01,0.21198720932006837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,16,power_law_1.01,0.2939136028289795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,8,power_law_1.2,0.12977919578552247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,16,power_law_1.01,0.3403968095779419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,8,power_law_1.2,0.1338047981262207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,8,power_law_1.2,0.14117759466171265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,16,power_law_1.01,0.5267327785491943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,8,power_law_1.2,0.15614720582962036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,16,power_law_1.01,0.6877888202667236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,8,power_law_1.2,0.1740928053855896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,16,power_law_1.01,0.8299648284912109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,8,power_law_1.2,0.21098880767822265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,8,power_law_1.2,0.23891839981079102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,16,power_law_1.01,1.356281566619873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,8,power_law_1.2,0.3025023937225342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,8,power_law_1.2,0.3657151937484741
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,8,power_law_1.2,0.4508096218109131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,16,power_law_1.01,2.8150463104248047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,8,power_law_1.2,0.5901375770568847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,8,power_law_1.2,0.8597311973571777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,8,power_law_1.2,1.1048831939697266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,8,power_law_1.2,1.3086848258972168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,8,power_law_1.2,2.026041603088379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,8,2,balanced,0.04176533222198486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,8,power_law_1.2,4.570828628540039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,8,2,balanced,0.04586666822433472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,8,2,balanced,0.04784533381462097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,8,2,balanced,0.05629866818586985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,8,2,balanced,0.0763626645008723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,8,2,balanced,0.07670933504899342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,8,2,balanced,0.07664533456166585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,8,2,balanced,0.07850133379300435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,8,2,balanced,0.0780266672372818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,8,2,balanced,0.07872533301512401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,8,2,balanced,0.07867200175921123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,8,2,balanced,0.08051200211048126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,8,2,balanced,0.08468799789746602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,8,2,balanced,0.08549333612124126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,8,2,balanced,0.09113599856694539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,8,2,balanced,0.09672000010808308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,8,2,balanced,0.10457066694895427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,8,2,balanced,0.11733333269755046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,8,2,balanced,0.12980266412099203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,8,2,balanced,0.15660267074902853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,8,2,balanced,0.17635732889175415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,8,2,balanced,0.2518346707026164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,8,2,balanced,0.3044640024503072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,8,2,balanced,0.4385013182957967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,8,2,balanced,0.5673706531524658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,8,2,balanced,0.8290399710337321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,8,2,balanced,1.074885368347168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,8,2,balanced,1.34007994333903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,8,2,balanced,2.0727893511454263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,8,2,balanced,4.020789464314778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,1,power_law_1.01,0.04810880124568939
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,1,power_law_1.01,0.048204800486564635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,1,power_law_1.01,0.0525056004524231
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,1,power_law_1.01,0.05976960062980652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,1,power_law_1.01,0.06462720036506653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,1,power_law_1.01,0.08874239921569824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,1,power_law_1.01,0.09155200123786926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,1,power_law_1.01,0.0959231972694397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,1,power_law_1.01,0.09616640210151672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,1,power_law_1.01,0.09882240295410157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,1,power_law_1.01,0.10166399478912354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,1,power_law_1.01,0.10616320371627808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,1,power_law_1.01,0.10701440572738648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,1,power_law_1.01,0.11297279596328735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,1,power_law_1.01,0.12169599533081055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,1,power_law_1.01,0.12752000093460084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,1,power_law_1.01,0.13852159976959227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,1,power_law_1.01,0.16379519701004028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,1,power_law_1.01,0.18668160438537598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,1,power_law_1.01,0.23808639049530028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,1,power_law_1.01,0.2870847940444946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,1,power_law_1.01,0.3935487985610962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,1,power_law_1.01,0.4987199783325195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,1,power_law_1.01,0.7092800140380859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,1,power_law_1.01,0.9135871887207031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,1,power_law_1.01,1.3267135620117188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,1,power_law_1.01,1.73240966796875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,1,power_law_1.01,2.14215030670166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,1,power_law_1.01,3.3305599212646486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,1,power_law_1.01,6.550630187988281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,8,power_law_1.01,0.03713920116424561
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,8,power_law_1.01,0.03739520013332367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,8,power_law_1.01,0.033471998572349546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,8,power_law_1.01,0.03849599957466125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,8,power_law_1.01,0.03874559998512268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,8,power_law_1.01,0.05182719826698303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,8,power_law_1.01,0.053491199016571046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,8,power_law_1.01,0.052908802032470705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,8,power_law_1.01,0.05452160239219665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,8,power_law_1.01,0.054073601961135864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,8,power_law_1.01,0.05626239776611328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,8,power_law_1.01,0.05779839754104614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,8,power_law_1.01,0.05780479907989502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,8,power_law_1.01,0.05950080156326294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,8,power_law_1.01,0.06638720035552978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,8,power_law_1.01,0.07047039866447449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,8,power_law_1.01,0.06851199865341187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,8,power_law_1.01,0.07920640110969543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,8,power_law_1.01,0.08236799836158752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,8,power_law_1.01,0.09746559858322143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,8,power_law_1.01,0.1097983956336975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,8,power_law_1.01,0.13907840251922607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,8,power_law_1.01,0.1658687949180603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,8,power_law_1.01,0.2193471908569336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,8,power_law_1.01,0.2566015958786011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,8,power_law_1.01,0.3886847972869873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,8,power_law_1.01,0.43695359230041503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,8,power_law_1.01,0.5710912227630616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,8,power_law_1.01,0.8978303909301758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,8,power_law_1.01,1.6149887084960937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,2,balanced,0.045968001087506614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,2,balanced,0.04846400022506714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,2,balanced,0.058789332707722984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,2,balanced,0.08181866506735484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,2,balanced,0.11185600360234578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,2,balanced,0.1293706695238749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,2,balanced,0.1312373379866282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,2,balanced,0.13327466448148093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,2,balanced,0.1367573340733846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,2,balanced,0.13615467151006064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,2,balanced,0.13889066378275552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,2,balanced,0.14178133010864258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,2,balanced,0.1458133359750112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,2,balanced,0.149509330590566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,2,balanced,0.15876799821853638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,2,balanced,0.1616106629371643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,2,balanced,0.1699413259824117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,2,balanced,0.19509865840276083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,2,balanced,0.2087786595026652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,2,balanced,0.2718506654103597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,2,balanced,0.29201066493988037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,2,balanced,0.4148000081380208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,2,balanced,0.4870773156483968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,2,balanced,0.7174293200174967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,2,balanced,0.8521173000335693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,2,balanced,1.2647466659545898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,2,balanced,1.60208527247111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,2,balanced,1.8906346956888835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,2,balanced,2.9901758829752603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,2,balanced,5.799509048461914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,8,1,power_law_1.01,0.04090240001678467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,8,1,power_law_1.01,0.056704002618789676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,8,1,power_law_1.01,0.05429760217666626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,8,1,power_law_1.01,0.07623680233955384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,8,1,power_law_1.01,0.07746559977531434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,8,1,power_law_1.01,0.07898240089416504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,8,1,power_law_1.01,0.0810368001461029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,8,1,power_law_1.01,0.08362879753112792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,8,1,power_law_1.01,0.08424959778785705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,8,1,power_law_1.01,0.0855679988861084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,8,1,power_law_1.01,0.08548480272293091
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,8,1,power_law_1.01,0.0858560025691986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,8,1,power_law_1.01,0.0887615978717804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,8,1,power_law_1.01,0.09508479833602905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,8,1,power_law_1.01,0.1030079960823059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,8,1,power_law_1.01,0.1047935962677002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,8,1,power_law_1.01,0.10983040332794189
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,8,1,power_law_1.01,0.1313472032546997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,8,1,power_law_1.01,0.1371072053909302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,8,1,power_law_1.01,0.16398080587387084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,8,1,power_law_1.01,0.19074560403823854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,8,1,power_law_1.01,0.24644479751586915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,8,power_law_1.01,0.056550401449203494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,8,1,power_law_1.01,0.29061119556427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,8,1,power_law_1.01,0.39486720561981203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,8,1,power_law_1.01,0.49837441444396974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,8,1,power_law_1.01,0.7085760116577149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,8,power_law_1.01,0.06919040083885193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,8,1,power_law_1.01,0.8962176322937012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,8,1,power_law_1.01,1.0971455574035645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,8,1,power_law_1.01,1.6872512817382812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,8,power_law_1.01,0.09192960262298584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,8,1,power_law_1.01,3.30909423828125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,8,power_law_1.01,0.06110720038414001
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,8,power_law_1.01,0.0695743978023529
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,8,power_law_1.01,0.0824895977973938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,8,power_law_1.01,0.08962560296058655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,8,power_law_1.01,0.10142079591751099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,8,power_law_1.01,0.09756799936294555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,8,power_law_1.01,0.10124800205230713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,2,1,power_law_1.01,0.06087039709091187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,8,power_law_1.01,0.10770560503005981
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,2,1,power_law_1.01,0.08479999899864196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,2,1,power_law_1.01,0.09349120259284974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,2,1,power_law_1.01,0.13137279748916625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,2,1,power_law_1.01,0.1335487961769104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,2,1,power_law_1.01,0.1335487961769104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,2,1,power_law_1.01,0.13925119638442993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,2,1,power_law_1.01,0.13730560541152953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,2,1,power_law_1.01,0.14302719831466676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,2,1,power_law_1.01,0.14467840194702147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,8,power_law_1.01,0.11260800361633301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,2,1,power_law_1.01,0.14517760276794434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,2,1,power_law_1.01,0.1507200002670288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,2,1,power_law_1.01,0.1536128044128418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,2,1,power_law_1.01,0.16060800552368165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,2,1,power_law_1.01,0.17499519586563111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,2,1,power_law_1.01,0.1802240014076233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,2,1,power_law_1.01,0.19353599548339845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,2,1,power_law_1.01,0.2203007936477661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,8,power_law_1.01,0.10915199518203736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,2,1,power_law_1.01,0.25068159103393556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,2,1,power_law_1.01,0.30805120468139646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,2,1,power_law_1.01,0.3613759994506836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,2,1,power_law_1.01,0.4783936023712158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,2,1,power_law_1.01,0.5938560009002686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,2,1,power_law_1.01,0.8331775665283203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,8,power_law_1.01,0.11349760293960572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,2,1,power_law_1.01,1.0606143951416016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,2,1,power_law_1.01,1.5281023979187012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,2,1,power_law_1.01,1.9871551513671875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,8,power_law_1.01,0.1272063970565796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,2,1,power_law_1.01,2.4619327545166017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,8,power_law_1.01,0.12439680099487305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,2,1,power_law_1.01,3.82281608581543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,8,power_law_1.01,0.13670400381088257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,2,1,power_law_1.01,7.493081665039062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,8,power_law_1.01,0.15355520248413085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,8,power_law_1.01,0.16149760484695436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,8,power_law_1.01,0.1916607975959778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,8,power_law_1.01,0.21472640037536622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,8,power_law_1.01,0.2709759950637817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,8,power_law_1.01,0.32792959213256834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,8,power_law_1.01,0.4352704048156738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,8,16,power_law_1.01,0.053350400924682614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,8,16,power_law_1.01,0.054611200094223024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,8,power_law_1.01,0.5539391994476318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,8,16,power_law_1.01,0.048588800430297854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,8,16,power_law_1.01,0.04961279928684235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,8,power_law_1.01,0.7804800033569336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,8,16,power_law_1.01,0.05125759840011597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,8,16,power_law_1.01,0.053388798236846925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,8,16,power_law_1.01,0.05377280116081238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,8,16,power_law_1.01,0.054745602607727054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,8,power_law_1.01,0.9696640014648438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,8,16,power_law_1.01,0.055180799961090085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,8,16,power_law_1.01,0.05495039820671081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,8,16,power_law_1.01,0.056358402967453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,8,16,power_law_1.01,0.05759360194206238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,8,power_law_1.01,1.1993984222412108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,8,16,power_law_1.01,0.0578495979309082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,8,16,power_law_1.01,0.060185599327087405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,8,16,power_law_1.01,0.06700159907341004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,8,16,power_law_1.01,0.06861439943313599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,8,power_law_1.01,1.7976383209228515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,8,16,power_law_1.01,0.07438719868659974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,8,16,power_law_1.01,0.08496000170707703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,8,16,power_law_1.01,0.08970239758491516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,8,16,power_law_1.01,0.11981439590454102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,8,16,power_law_1.01,0.13464959859848022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,8,16,power_law_1.01,0.1674623966217041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,8,power_law_1.01,3.6159744262695312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,8,16,power_law_1.01,0.21101438999176025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,8,16,power_law_1.01,0.26859519481658933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,8,16,power_law_1.01,0.3237567901611328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,8,16,power_law_1.01,0.4848896026611328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,8,16,power_law_1.01,0.5686207771301269
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,8,16,power_law_1.01,0.7192831993103027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,8,16,power_law_1.01,1.0717503547668457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,8,16,power_law_1.01,2.3202943801879883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,4,power_law_1.2,0.09653760194778442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,4,power_law_1.2,0.1294144034385681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,4,power_law_1.2,0.0918079972267151
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,4,power_law_1.2,0.12001279592514039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,4,power_law_1.2,0.1659327983856201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,4,power_law_1.2,0.21289598941802979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,4,power_law_1.2,0.3273216009140015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,4,power_law_1.01,0.0908735990524292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,4,power_law_1.2,0.3247551918029785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,4,power_law_1.01,0.11753599643707276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,4,power_law_1.2,0.3603008031845093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,4,power_law_1.01,0.17417600154876708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,4,power_law_1.01,0.11139199733734131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,4,power_law_1.2,0.34435200691223145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,4,power_law_1.01,0.15012480020523072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,4,power_law_1.2,0.36240639686584475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,4,power_law_1.01,0.21119999885559082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,4,power_law_1.2,0.390393590927124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,4,power_law_1.2,0.4011263847351074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,4,power_law_1.01,0.23278720378875734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,4,power_law_1.2,0.4141119956970215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,4,power_law_1.01,0.2929088115692139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,4,power_law_1.2,0.4315968036651611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,4,power_law_1.01,0.29460480213165285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,4,power_law_1.2,0.43688321113586426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,4,power_law_1.01,0.3188159942626953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,4,power_law_1.2,0.4720767974853516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,4,power_law_1.01,0.32704639434814453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,4,power_law_1.2,0.5354879856109619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,4,power_law_1.01,0.33033599853515627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,2,power_law_1.01,0.03160960078239441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,2,power_law_1.01,0.03204480111598969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,4,power_law_1.2,0.5432127952575684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,4,power_law_1.01,0.34669439792633056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,2,power_law_1.01,0.03399679958820343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,2,power_law_1.01,0.0403903990983963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,2,power_law_1.01,0.045440000295639035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,4,power_law_1.01,0.35829761028289797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,4,power_law_1.2,0.6140480041503906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,2,power_law_1.01,0.06175360083580017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,2,power_law_1.01,0.06382079720497132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,2,power_law_1.01,0.06613759994506836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,4,power_law_1.01,0.3694080114364624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,4,power_law_1.2,0.6532479763031006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,2,power_law_1.01,0.0691648006439209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,2,power_law_1.01,0.07120640277862549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,4,power_law_1.01,0.37392640113830566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,2,power_law_1.01,0.0723904013633728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,4,power_law_1.2,0.7844223976135254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,2,power_law_1.01,0.07768319845199585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,2,power_law_1.01,0.07971199750900268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,4,power_law_1.01,0.38442239761352537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,2,power_law_1.01,0.08203520178794861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,4,power_law_1.2,0.890278434753418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,2,power_law_1.01,0.0892416000366211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,2,power_law_1.01,0.0948415994644165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,4,power_law_1.01,0.4121535778045654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,2,power_law_1.01,0.09910399913787842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,4,power_law_1.2,1.1510911941528321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,2,power_law_1.01,0.11743359565734864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,4,power_law_1.01,0.4462272167205811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,2,power_law_1.01,0.13523199558258056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,2,power_law_1.01,0.17044479846954347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,4,power_law_1.2,1.2807104110717773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,4,power_law_1.01,0.48815360069274905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,2,power_law_1.01,0.2077183961868286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,2,power_law_1.01,0.2709183931350708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,4,power_law_1.01,0.5568511962890625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,2,power_law_1.01,0.34380159378051756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,4,power_law_1.2,1.7733503341674806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,2,power_law_1.01,0.4538112163543701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,4,power_law_1.01,0.6810880184173584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,2,power_law_1.01,0.5913087844848632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,4,power_law_1.2,2.187820816040039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,2,power_law_1.01,0.899180793762207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,4,power_law_1.01,0.7247039794921875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,2,power_law_1.01,1.1766207695007325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,4,power_law_1.01,0.9089407920837402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,4,power_law_1.2,2.824563217163086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,2,power_law_1.01,1.4108160018920899
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,4,power_law_1.01,1.1320575714111327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,2,power_law_1.01,2.217753601074219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,4,power_law_1.2,4.388172912597656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,4,power_law_1.01,1.5158656120300293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,4,power_law_1.01,1.9407232284545899
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,2,power_law_1.01,4.383084869384765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,4,power_law_1.01,2.2710079193115233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,4,power_law_1.2,8.9731201171875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,4,power_law_1.01,3.5394302368164063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,4,power_law_1.01,6.400096130371094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,2,power_law_1.01,0.08643839955329895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,8,1,balanced,0.05425600210825602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,8,1,balanced,0.05588266750176748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,8,1,balanced,0.0639626681804657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,8,1,balanced,0.084197332461675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,8,1,balanced,0.12390933434168498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,8,1,balanced,0.1271573305130005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,8,1,balanced,0.12709333499272665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,8,1,balanced,0.1292800009250641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,8,1,balanced,0.12956266601880392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,8,1,balanced,0.1290720005830129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,8,1,balanced,0.13300266861915588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,8,1,balanced,0.13759467005729675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,8,1,balanced,0.1413226624329885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,8,1,balanced,0.1439786652723948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,2,power_law_1.01,0.12288639545440674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,8,1,balanced,0.15039466818173727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,8,1,balanced,0.15779733657836914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,8,1,balanced,0.16706132888793945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,8,1,balanced,0.1928159991900126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,8,1,balanced,0.2146186629931132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,8,1,balanced,0.2727573315302531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,8,1,balanced,0.3188213308652242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,8,1,balanced,0.4736800193786621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,8,1,balanced,0.5811466773351034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,8,1,balanced,0.831397294998169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,8,1,balanced,1.0722986857096355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,2,power_law_1.01,0.09996799826622009
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,8,1,balanced,1.5818079312642415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,8,1,balanced,2.1031360626220703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,8,1,balanced,2.6106506983439126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,2,power_law_1.01,0.15731199979782104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,8,1,balanced,4.011407852172852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,2,power_law_1.01,0.2096384048461914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,8,1,balanced,7.8296051025390625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,32,power_law_1.01,0.06373119950294495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,2,power_law_1.01,0.3316351890563965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,32,power_law_1.01,0.03606399893760681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,32,power_law_1.01,0.03501439988613129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,32,power_law_1.01,0.032364800572395325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,2,power_law_1.01,0.36304640769958496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,32,power_law_1.01,0.03214080035686493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,32,power_law_1.01,0.03340800106525421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,32,power_law_1.01,0.03418239951133728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,2,power_law_1.01,0.3731391906738281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,32,power_law_1.01,0.03375360071659088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,32,power_law_1.01,0.03384959995746613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,32,power_law_1.01,0.034227201342582704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,2,power_law_1.01,0.3844480037689209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,32,power_law_1.01,0.0352512001991272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,32,power_law_1.01,0.03676159977912903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,32,power_law_1.01,0.04037759900093078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,2,power_law_1.01,0.39664640426635744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,32,power_law_1.01,0.04260480105876922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,32,power_law_1.01,0.04832000136375427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,32,power_law_1.01,0.05255680084228516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,2,power_law_1.01,0.40726399421691895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,32,power_law_1.01,0.05259519815444946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,32,power_law_1.01,0.05782399773597717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,32,power_law_1.01,0.058815997838973996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,2,power_law_1.01,0.41986560821533203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,32,power_law_1.01,0.06419199705123901
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,32,power_law_1.01,0.0707264006137848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,32,power_law_1.01,0.08170239925384522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,2,power_law_1.01,0.4303872108459473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,32,power_law_1.01,0.09337599873542786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,32,power_law_1.01,0.11161600351333618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,8,power_law_1.01,0.05160319805145264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,2,power_law_1.01,0.43749117851257324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,32,power_law_1.01,0.14777599573135375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,32,power_law_1.01,0.19969919919967652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,2,power_law_1.01,0.4533952236175537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,32,power_law_1.01,0.2937727928161621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,8,power_law_1.01,0.06392959952354431
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,1,32,power_law_1.01,0.3584511995315552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,2,power_law_1.01,0.4739520072937012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,1,32,power_law_1.01,0.5592895984649658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,2,power_law_1.01,0.4919167995452881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,8,power_law_1.01,0.07539839744567871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,1,32,power_law_1.01,1.1477760314941405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,2,power_law_1.01,0.5355711936950683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,8,power_law_1.01,0.052665597200393675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,2,power_law_1.01,0.5813568115234375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,8,power_law_1.01,0.05932160019874573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,2,power_law_1.01,0.6856704235076905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,8,power_law_1.01,0.06730239987373351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,2,power_law_1.01,0.7532288074493408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,8,power_law_1.01,0.07487360239028931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,2,power_law_1.01,0.9481087684631347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,8,power_law_1.01,0.07996159791946411
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,2,power_law_1.01,1.0679167747497558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,8,power_law_1.01,0.08279680013656616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,32,balanced,0.035973332822322845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,32,balanced,0.03382399926582972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,32,balanced,0.03346133232116699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,32,balanced,0.03340800106525421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,32,balanced,0.035461333890755974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,32,balanced,0.03581333408753077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,32,balanced,0.0354666660229365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,32,balanced,0.03585600107908249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,32,balanced,0.03579733272393545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,32,balanced,0.037765334049860634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,32,balanced,0.0355679988861084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,32,balanced,0.03749866783618927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,32,balanced,0.03773866593837738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,2,power_law_1.01,1.3623167991638183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,32,balanced,0.037903999288876854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,32,balanced,0.04178133110205332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,32,balanced,0.04009599983692169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,32,balanced,0.04206933577855428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,32,balanced,0.041893333196640015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,32,balanced,0.04364266494909922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,32,balanced,0.04598933458328247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,32,balanced,0.04867733518282572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,32,balanced,0.05212266743183136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,32,balanced,0.05613866448402405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,32,balanced,0.07192533214886983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,8,power_law_1.01,0.0856768012046814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,32,balanced,0.081386665503184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,32,balanced,0.10560533404350281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,32,balanced,0.13195199767748514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,32,balanced,0.15478932857513428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,32,balanced,0.21054399013519287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,32,balanced,0.377130667368571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,8,power_law_1.01,0.08659200072288513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,2,power_law_1.01,1.6440319061279296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,8,power_law_1.01,0.08913919925689698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,2,power_law_1.01,2.2567680358886717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,8,power_law_1.01,0.0940671980381012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,8,power_law_1.01,0.09797760248184204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,2,power_law_1.01,2.966707229614258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,8,power_law_1.01,0.10679039955139161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,8,power_law_1.01,0.10932480096817017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,32,8,power_law_1.01,0.03590399920940399
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,32,8,power_law_1.01,0.03518719971179962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,32,8,power_law_1.01,0.03258880078792572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,2,2,power_law_1.01,3.519801712036133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,8,power_law_1.01,0.1128383994102478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,32,8,power_law_1.01,0.032595199346542356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,32,8,power_law_1.01,0.03253119885921478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,32,8,power_law_1.01,0.033504000306129454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,32,8,power_law_1.01,0.03407360017299652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,32,8,power_law_1.01,0.03370240032672882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,32,8,power_law_1.01,0.034220799803733826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,32,8,power_law_1.01,0.03430399894714355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,32,8,power_law_1.01,0.03427839875221252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,32,8,power_law_1.01,0.0355648010969162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,32,8,power_law_1.01,0.03576320111751556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,8,power_law_1.01,0.12867200374603271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,32,8,power_law_1.01,0.035769599676132205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,32,8,power_law_1.01,0.040729600191116336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,32,8,power_law_1.01,0.0411327987909317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,32,8,power_law_1.01,0.04380159974098206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,32,8,power_law_1.01,0.048153600096702574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,32,8,power_law_1.01,0.051551997661590576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,32,8,power_law_1.01,0.06422399878501892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,32,8,power_law_1.01,0.067084801197052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,32,8,power_law_1.01,0.08550400137901307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,8,power_law_1.01,0.13647359609603882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,32,8,power_law_1.01,0.10768640041351318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,32,8,power_law_1.01,0.1483456015586853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,32,8,power_law_1.01,0.17559679746627807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,32,8,power_law_1.01,0.2528320074081421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,32,8,power_law_1.01,0.308351993560791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,2,2,power_law_1.01,5.372051239013672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,8,power_law_1.01,0.16272000074386597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,32,8,power_law_1.01,0.387993597984314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,32,8,power_law_1.01,0.6035071849822998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,32,8,power_law_1.01,1.150592041015625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,8,power_law_1.01,0.18442239761352539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,8,power_law_1.01,0.23872001171112062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,8,power_law_1.01,0.28740479946136477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,8,power_law_1.01,0.3927232027053833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,2,2,power_law_1.01,11.154093170166016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,8,power_law_1.01,0.47742719650268556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,8,power_law_1.01,0.6714560031890869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,8,power_law_1.01,0.8623295783996582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,8,power_law_1.01,1.0015551567077636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,8,power_law_1.01,1.611302375793457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,16,balanced,0.05375466744105021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,16,balanced,0.04982399940490723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,16,balanced,0.05282133320967356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,16,balanced,0.0539680023988088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,16,balanced,0.05600533386071523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,16,balanced,0.062352001667022705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,16,balanced,0.06853333115577698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,16,balanced,0.06841066479682922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,16,balanced,0.06894933183987935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,16,balanced,0.06843733290831248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,16,balanced,0.07233599821726482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,16,balanced,0.07249600191911061
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,16,balanced,0.07437866429487865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,16,balanced,0.07573333382606506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,16,balanced,0.08062933385372162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,16,balanced,0.08264000217119853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,16,balanced,0.08673066894213359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,16,balanced,0.09723732868830363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,16,balanced,0.10683733224868774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,16,balanced,0.12876799702644348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,16,balanced,0.14298666516939798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,16,balanced,0.17934399843215942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,16,balanced,0.21283199389775595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,16,balanced,0.2737920085589091
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,16,balanced,0.3478720188140869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,8,power_law_1.01,2.9671808242797852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,16,balanced,0.4712640047073364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,16,balanced,0.6083999872207642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,16,balanced,0.7595040003458658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,16,balanced,1.148144006729126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,16,balanced,2.2402772903442383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,8,power_law_1.01,0.07890560030937195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,2,power_law_1.01,0.048307201266288756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,8,power_law_1.01,0.08663039803504943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,8,power_law_1.01,0.05856639742851257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,2,power_law_1.01,0.05775359869003296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,8,power_law_1.01,0.07676159739494323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,2,power_law_1.01,0.07525119781494141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,8,power_law_1.01,0.08589439988136291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,2,power_law_1.01,0.06850559711456299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,8,power_law_1.01,0.09864320158958435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,2,power_law_1.01,0.09540479779243469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,8,power_law_1.01,0.10620160102844238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,8,power_law_1.01,0.10548479557037353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,2,power_law_1.01,0.12019200325012207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,8,power_law_1.01,0.11196800470352172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,2,power_law_1.01,0.138099205493927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,8,power_law_1.01,0.1087615966796875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,8,power_law_1.01,0.11093759536743164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,2,power_law_1.01,0.1627776026725769
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,8,power_law_1.01,0.11077760457992554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,2,power_law_1.01,0.1726207971572876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,8,power_law_1.01,0.1151360034942627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,8,power_law_1.01,0.12124160528182984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,2,power_law_1.01,0.1784767985343933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,8,power_law_1.01,0.1257151961326599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,2,power_law_1.01,0.1914240002632141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,8,power_law_1.01,0.12531839609146117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,2,power_law_1.01,0.19760639667510987
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,8,power_law_1.01,0.13211519718170167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,8,power_law_1.01,0.1434175968170166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,2,power_law_1.01,0.19854079484939574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,8,power_law_1.01,0.1607424020767212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,2,power_law_1.01,0.2091264009475708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,8,power_law_1.01,0.1966591954231262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,2,power_law_1.01,0.22614400386810302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,8,power_law_1.01,0.2095616102218628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,8,power_law_1.01,0.25730559825897215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,2,power_law_1.01,0.23176960945129393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,8,power_law_1.01,0.3341248035430908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,2,power_law_1.01,0.2555775880813599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,8,power_law_1.01,0.41304960250854494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,2,power_law_1.01,0.2893631935119629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,8,power_law_1.01,0.46988801956176757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,2,power_law_1.01,0.2942336082458496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,8,power_law_1.01,0.6410048007965088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,2,power_law_1.01,0.3483648061752319
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,8,power_law_1.01,0.7938943862915039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,2,power_law_1.01,0.41887998580932617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,8,power_law_1.01,1.026521587371826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,2,power_law_1.01,0.5322815895080566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,2,power_law_1.01,0.6092991828918457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,8,power_law_1.01,1.4382335662841796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,2,power_law_1.01,0.8001919746398926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,2,power_law_1.01,1.0297087669372558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,8,power_law_1.01,2.879596710205078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,2,power_law_1.01,1.4078720092773438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,2,power_law_1.01,1.7721664428710937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,2,power_law_1.01,2.136172866821289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,16,balanced,0.03149333347876867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,16,balanced,0.03306666761636734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,16,balanced,0.03176533430814743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,16,balanced,0.03143466760714849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,16,balanced,0.03537066777547201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,16,balanced,0.03544000039498011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,16,balanced,0.03748800108830134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,16,balanced,0.035717333356539406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,16,balanced,0.03563733398914337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,16,balanced,0.037589333951473236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,16,balanced,0.03749333322048187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,16,balanced,0.03771200031042099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,16,balanced,0.03758399933576584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,16,balanced,0.03966933240493139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,16,balanced,0.04348266621430715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,16,balanced,0.04354133208592733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,16,balanced,0.04557866851488749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,16,balanced,0.047728002071380615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,16,balanced,0.05379199981689453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,16,balanced,0.06204266846179962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,16,balanced,0.07252266506354015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,16,balanced,0.09321066737174988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,16,balanced,0.1130560040473938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,16,balanced,0.1482080022493998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,16,balanced,0.18707199891408285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,16,balanced,0.2510666648546855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,16,balanced,0.3148639996846517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,16,balanced,0.38605864842732746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,16,balanced,0.5794933239618937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,16,balanced,1.1140800317128499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,2,power_law_1.01,3.4254016876220703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,2,power_law_1.01,6.359353637695312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,16,power_law_1.01,0.10474239587783814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,16,power_law_1.01,0.1253056049346924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,16,power_law_1.01,0.07255039811134338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,16,power_law_1.01,0.08436480164527893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,2,balanced,0.06542400022347768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,2,balanced,0.10796266794204712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,2,balanced,0.1220266620318095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,2,balanced,0.18421866496404013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,2,balanced,0.31301333506902057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,2,balanced,0.37778135140736896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,2,balanced,0.37959468364715576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,2,balanced,0.38229866822560626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,2,balanced,0.3850560188293457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,2,balanced,0.3868639866511027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,2,balanced,0.3886880079905192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,16,power_law_1.01,0.0933247983455658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,2,balanced,0.39215465386708576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,2,balanced,0.3959786494572957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,2,balanced,0.4004053274790446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,2,balanced,0.4103999932607015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,2,balanced,0.41203200817108154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,2,balanced,0.4232800006866455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,2,balanced,0.4462133248647054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,2,balanced,0.4692586660385132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,2,balanced,0.5824533303578695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,2,balanced,0.5611733198165894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,2,balanced,0.7745440006256104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,2,balanced,0.7769280274709066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,2,balanced,1.177898645401001
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,2,balanced,1.3148586750030518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,2,balanced,1.9220800399780273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,16,power_law_1.01,0.1097216010093689
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,2,balanced,2.3665706316630044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,2,balanced,2.7633546193440757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,2,balanced,4.534538586934407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,16,power_law_1.01,0.11391999721527099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,2,balanced,8.39959462483724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,16,power_law_1.01,0.12122880220413208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,16,power_law_1.01,0.12153600454330445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,16,power_law_1.01,0.11809279918670654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,16,power_law_1.01,0.1221951961517334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,16,power_law_1.01,0.12580480575561523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,16,power_law_1.01,0.12855039834976195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,4,balanced,0.05458133419354757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,4,balanced,0.057402665416399636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,4,balanced,0.06044266621271769
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,4,balanced,0.06460799773534139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,4,balanced,0.0809333324432373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,4,balanced,0.12026133139928182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,4,balanced,0.14935466647148132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,4,balanced,0.15107199549674988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,4,balanced,0.1511733333269755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,4,balanced,0.153738667567571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,4,balanced,0.14854933818181357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,4,balanced,0.15226133664449057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,4,balanced,0.15319466590881348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,4,balanced,0.15498666961987814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,4,balanced,0.16370667020479837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,4,balanced,0.16435733437538147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,4,balanced,0.1722559928894043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,4,balanced,0.18530666828155518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,4,balanced,0.20068265994389853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,4,balanced,0.22630933920542398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,4,balanced,0.2510720094045003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,4,balanced,0.30914666255315143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,4,balanced,0.3633120059967041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,4,balanced,0.48341866334279376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,16,power_law_1.01,0.1301568031311035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,4,balanced,0.6370666821797689
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,4,balanced,0.8613119920094808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,4,balanced,1.154693365097046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,4,balanced,1.4421547253926594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,4,balanced,2.1750453313191733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,16,power_law_1.01,0.14028160572052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,4,balanced,4.183093388875325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,16,power_law_1.01,0.14470399618148805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,16,power_law_1.01,0.14920320510864257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,16,power_law_1.01,0.16888320446014404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,16,power_law_1.01,0.18824959993362428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,16,power_law_1.01,0.234169602394104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,4,1,power_law_1.2,0.0471807986497879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,4,1,power_law_1.2,0.04671359956264496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,16,power_law_1.01,0.25947520732879636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,4,1,power_law_1.2,0.04991360008716583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,4,1,power_law_1.2,0.054841601848602296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,4,1,power_law_1.2,0.059628802537918094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,4,1,power_law_1.2,0.0779263973236084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,4,1,power_law_1.2,0.08426240086555481
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,16,power_law_1.01,0.33133440017700194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,4,1,power_law_1.2,0.08476160168647766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,4,1,power_law_1.2,0.08317440152168273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,4,1,power_law_1.2,0.08552320003509521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,4,1,power_law_1.2,0.08963840007781983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,4,1,power_law_1.2,0.09352959990501404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,16,power_law_1.01,0.39911038875579835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,4,1,power_law_1.2,0.09619200229644775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,4,1,power_law_1.2,0.10035840272903443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,4,1,power_law_1.2,0.10996479988098144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,4,1,power_law_1.2,0.11372159719467163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,4,1,power_law_1.2,0.12625919580459594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,16,power_law_1.01,0.5054656028747558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,4,1,power_law_1.2,0.15063040256500243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,4,1,power_law_1.2,0.1736256003379822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,4,1,power_law_1.2,0.21928319931030274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,4,1,power_law_1.2,0.2634239912033081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,4,1,power_law_1.2,0.36062719821929934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,16,power_law_1.01,0.5978623867034912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,4,1,power_law_1.2,0.46272640228271483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,4,1,power_law_1.2,0.6655935764312744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,4,1,power_law_1.2,0.8587008476257324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,16,power_law_1.01,0.8113216400146485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,4,1,power_law_1.2,1.2604607582092284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,4,1,power_law_1.2,1.6521984100341798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,4,1,power_law_1.2,2.046444892883301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,16,power_law_1.01,1.1240127563476563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,4,1,power_law_1.2,3.200755310058594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,1,16,power_law_1.01,1.3049728393554687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,4,1,power_law_1.2,6.258841705322266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,1,16,power_law_1.01,2.00184326171875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,1,16,power_law_1.01,3.9954113006591796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,8,power_law_1.2,0.04579200148582459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,8,power_law_1.2,0.052153599262237546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,8,power_law_1.2,0.047660800814628604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,8,power_law_1.2,0.04979200065135956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,8,power_law_1.2,0.053363198041915895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,8,power_law_1.2,0.0606719970703125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,8,power_law_1.2,0.07342079877853394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,8,power_law_1.2,0.0741375982761383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,8,power_law_1.2,0.07404159903526306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,8,power_law_1.2,0.0749888002872467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,8,power_law_1.2,0.07915520071983337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,8,power_law_1.2,0.0806656002998352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,8,power_law_1.2,0.08315520286560059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,128,balanced,0.03689600030581156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,128,balanced,0.03735466549793879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,128,balanced,0.03538133452335993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,128,balanced,0.035317334036032356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,128,balanced,0.036981334288915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,8,power_law_1.2,0.08962560296058655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,128,balanced,0.03733866661787033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,128,balanced,0.03687999894221624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,128,balanced,0.0372533326347669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,128,balanced,0.03729599962631861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,128,balanced,0.03718933214743932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,128,balanced,0.03843733419974645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,128,balanced,0.03735466549793879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,128,balanced,0.03921066721280416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,128,balanced,0.038986665507157646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,8,power_law_1.2,0.09992319941520691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,128,balanced,0.04125333329041799
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,128,balanced,0.04274666806062063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,128,balanced,0.04257600009441376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,128,balanced,0.045498669147491455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,128,balanced,0.044906665881474815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,128,balanced,0.048698668678601585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,128,balanced,0.048911998669306435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,128,balanced,0.05482666691144308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,128,balanced,0.05881600081920624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,8,power_law_1.2,0.10334080457687378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,128,balanced,0.07195733487606049
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,128,balanced,0.09417600433031718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,128,balanced,0.13114666938781738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,128,balanced,0.16142933567365012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,128,balanced,0.19129067659378052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,128,balanced,0.28134934107462567
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,128,balanced,0.5195039908091227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,8,power_law_1.2,0.11045119762420655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,8,power_law_1.2,0.1277567982673645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,8,power_law_1.2,0.14094719886779786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,8,power_law_1.2,0.16711679697036744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,8,power_law_1.2,0.20884480476379394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,8,power_law_1.2,0.2705343961715698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,8,power_law_1.2,0.3345279932022095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,8,power_law_1.2,0.46151041984558105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,8,power_law_1.2,0.5786623954772949
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,8,power_law_1.2,0.8179136276245117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,8,power_law_1.2,1.1068608283996582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,8,8,power_law_1.2,0.05495039820671081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,8,power_law_1.2,1.422860813140869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,8,8,power_law_1.2,0.05240319967269898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,8,8,power_law_1.2,0.05127040147781372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,8,power_law_1.2,2.2602624893188477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,8,8,power_law_1.2,0.05400320291519165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,8,8,power_law_1.2,0.055801600217819214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,8,8,power_law_1.2,0.0622655987739563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,8,power_law_1.2,4.289503860473633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,8,8,power_law_1.2,0.06450560092926025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,8,8,power_law_1.2,0.06536319851875305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,8,8,power_law_1.2,0.06683520078659058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,8,8,power_law_1.2,0.06743040084838867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,8,8,power_law_1.2,0.06769279837608337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,8,8,power_law_1.2,0.06906239986419678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,8,8,power_law_1.2,0.07192959785461425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,8,8,power_law_1.2,0.07649919986724854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,8,8,power_law_1.2,0.08492159843444824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,8,8,power_law_1.2,0.08592640161514283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,8,8,power_law_1.2,0.09490560293197632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,8,8,power_law_1.2,0.1090432047843933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,4,power_law_1.2,0.07496960163116455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,8,8,power_law_1.2,0.12314879894256592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,4,power_law_1.2,0.07212799787521362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,8,8,power_law_1.2,0.14826879501342774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,4,power_law_1.2,0.0654911994934082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,4,power_law_1.2,0.08209279775619507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,8,8,power_law_1.2,0.1682752013206482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,4,power_law_1.2,0.09466239809989929
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,8,8,power_law_1.2,0.2393343925476074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,4,power_law_1.2,0.14339840412139893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,4,power_law_1.2,0.15575679540634155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,8,8,power_law_1.2,0.2682496070861816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,4,power_law_1.2,0.1655295968055725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,4,power_law_1.2,0.16712960004806518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,8,8,power_law_1.2,0.39295361042022703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,4,power_law_1.2,0.1664896011352539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,4,power_law_1.2,0.17150720357894897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,8,8,power_law_1.2,0.4705984115600586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,4,power_law_1.2,0.17624319791793824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,4,power_law_1.2,0.18277759552001954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,8,8,power_law_1.2,0.679040002822876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,4,power_law_1.2,0.18570239543914796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,4,power_law_1.2,0.19579520225524902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,8,8,power_law_1.2,0.8573696136474609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,4,power_law_1.2,0.19959039688110353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,4,power_law_1.2,0.21475839614868164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,4,power_law_1.2,0.23816320896148682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,8,8,power_law_1.2,1.1314623832702637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,4,power_law_1.2,0.24316799640655518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,4,power_law_1.2,0.30487680435180664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,4,power_law_1.2,0.31832959651947024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,8,8,power_law_1.2,1.9653568267822266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,4,power_law_1.2,0.40952320098876954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,4,power_law_1.2,0.4738111972808838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,4,power_law_1.2,0.5935679912567139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,4,power_law_1.2,0.7762176036834717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,8,8,power_law_1.2,4.011648178100586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,4,power_law_1.2,1.0997376441955566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,4,power_law_1.2,1.4362943649291993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,4,power_law_1.2,1.6744512557983398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,4,power_law_1.2,2.9240575790405274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,4,power_law_1.2,6.127916717529297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,8,power_law_1.2,0.03431040048599243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,8,power_law_1.2,0.034860798716545106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,8,power_law_1.2,0.03107840120792389
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,8,power_law_1.2,0.03221760094165802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,8,power_law_1.2,0.0345984011888504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,8,power_law_1.2,0.035667198896408084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,8,power_law_1.2,0.03710080087184906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,8,power_law_1.2,0.03678080141544342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,8,power_law_1.2,0.038124799728393555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,8,power_law_1.2,0.0390720009803772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,8,power_law_1.2,0.03951359987258911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,8,power_law_1.2,0.04156160056591034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,32,power_law_1.2,0.058336001634597776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,8,power_law_1.2,0.045023998618125914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,8,power_law_1.2,0.04430719912052154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,8,power_law_1.2,0.05097600221633911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,32,power_law_1.2,0.05355520248413086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,8,power_law_1.2,0.054118400812149046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,8,power_law_1.2,0.056883198022842404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,32,power_law_1.2,0.04876160025596619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,8,power_law_1.2,0.0656063973903656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,8,power_law_1.2,0.0664192020893097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,32,power_law_1.2,0.0500544011592865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,8,power_law_1.2,0.07455360293388366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,8,power_law_1.2,0.08420479893684388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,32,power_law_1.2,0.05251200199127197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,8,power_law_1.2,0.10637439489364624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,8,power_law_1.2,0.13210879564285277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,32,power_law_1.2,0.05960320234298706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,8,power_law_1.2,0.16278400421142578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,8,power_law_1.2,0.21473920345306396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,32,power_law_1.2,0.05914880037307739
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,8,power_law_1.2,0.311078405380249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,32,power_law_1.2,0.05985280275344849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,8,power_law_1.2,0.3446912050247192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,32,power_law_1.2,0.06169599890708923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,2,8,power_law_1.2,0.4919295787811279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,32,power_law_1.2,0.06229760050773621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,2,8,power_law_1.2,0.7404863834381104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,32,power_law_1.2,0.06494719982147217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,2,8,power_law_1.2,1.5486592292785644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,32,power_law_1.2,0.06904320120811462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,32,power_law_1.2,0.0720575988292694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,32,power_law_1.2,0.07640320062637329
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,32,power_law_1.2,0.08365439772605895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,32,power_law_1.2,0.08813440203666686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,8,balanced,0.03561066587766012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,8,balanced,0.035487999518712364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,8,balanced,0.035487999518712364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,8,balanced,0.03737599899371465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,8,balanced,0.041690667470296226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,8,balanced,0.04359466830889384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,8,balanced,0.04363200068473816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,8,balanced,0.04368533194065094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,8,balanced,0.042277331153551735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,8,balanced,0.04385066529115041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,8,balanced,0.04424533247947693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,8,balanced,0.04577599962552389
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,8,balanced,0.04390400151411692
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,8,balanced,0.04570133487383524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,8,balanced,0.05004266897837321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,8,balanced,0.04996799925963084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,8,balanced,0.05190933247407278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,8,balanced,0.05435733497142792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,8,balanced,0.060085331400235496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,8,balanced,0.0620959997177124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,8,balanced,0.06974933544794719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,8,balanced,0.09071466326713562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,8,balanced,0.10280533631642659
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,8,balanced,0.1357919971148173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,32,power_law_1.2,0.09100800156593322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,8,balanced,0.1664906640847524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,8,balanced,0.22588266928990683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,8,balanced,0.2725866635640462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,8,balanced,0.3315253257751465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,8,balanced,0.5034559965133667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,8,balanced,0.9457440376281738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,32,power_law_1.2,0.10570240020751953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,16,balanced,0.03862400104602178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,16,balanced,0.03383466601371765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,16,balanced,0.029887999097506206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,16,balanced,0.033802665770053864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,16,balanced,0.03378133227427801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,32,power_law_1.2,0.11806080341339112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,16,balanced,0.03600533306598663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,16,balanced,0.03775999943415324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,16,balanced,0.03736000011364619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,16,balanced,0.035760000348091125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,16,balanced,0.03637866675853729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,16,balanced,0.036602665980656944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,16,balanced,0.03975466638803482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,16,balanced,0.03783999880154928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,16,balanced,0.0360000009338061
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,16,balanced,0.041589332123597465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,16,balanced,0.04188799858093262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,16,balanced,0.042405332128206887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,16,balanced,0.050986667474110924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,16,balanced,0.0440586656332016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,16,balanced,0.0582239975531896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,16,balanced,0.05440000196297964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,16,balanced,0.06706133484840393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,16,balanced,0.07015466690063477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,16,balanced,0.09733866651852925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,16,balanced,0.09571733077367146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,16,balanced,0.13174399733543396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,16,balanced,0.14004266262054443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,2,16,balanced,0.1616373360157013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,32,power_law_1.2,0.1473088026046753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,2,16,balanced,0.21280000607172647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,2,16,balanced,0.36908801396687824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,32,power_law_1.2,0.17295360565185547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,32,power_law_1.2,0.23510398864746093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,32,power_law_1.2,0.2381824016571045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,16,balanced,0.03741333385308584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,16,balanced,0.0395413339138031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,16,balanced,0.03762666632731756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,16,balanced,0.037733333806196846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,16,balanced,0.04456533491611481
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,16,balanced,0.045935998360315956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,16,balanced,0.04605866471926371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,16,balanced,0.045781334241231285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,16,balanced,0.045824001232783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,32,power_law_1.2,0.30748159885406495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,16,balanced,0.04756266872088114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,16,balanced,0.047877331574757896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,16,balanced,0.04776533444722494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,16,balanced,0.04799999793370565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,16,balanced,0.0497920016447703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,16,balanced,0.053413331508636475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,16,balanced,0.05428266525268555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,16,balanced,0.05668266614278158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,16,balanced,0.0621973325808843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,16,balanced,0.07229866584142049
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,16,balanced,0.0825973351796468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,16,balanced,0.0827466646830241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,16,balanced,0.10028266906738281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,16,balanced,0.11157866319020589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,16,balanced,0.14050133029619852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,16,balanced,0.16380799810091654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,16,balanced,0.21572266022364298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,16,balanced,0.2719893256823222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,16,balanced,0.3308639923731486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,16,balanced,0.4899199803670247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,16,balanced,0.9069653352101644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,32,power_law_1.2,0.39302399158477785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,32,power_law_1.2,0.5973440170288086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,32,power_law_1.2,0.7575551986694335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,32,power_law_1.01,0.07893120050430298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,4,32,power_law_1.2,0.9113856315612793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,32,power_law_1.01,0.07929599881172181
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,32,power_law_1.01,0.055769598484039305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,128,power_law_1.01,0.06122879981994629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,4,32,power_law_1.2,1.4161855697631835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,128,power_law_1.01,0.056883198022842404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,32,power_law_1.01,0.058316802978515624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,128,power_law_1.01,0.04856959879398346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,128,power_law_1.01,0.053491199016571046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,32,power_law_1.01,0.0663424015045166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,128,power_law_1.01,0.05356159806251526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,32,power_law_1.01,0.07706239819526672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,128,power_law_1.01,0.0555840015411377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,4,32,power_law_1.2,3.440275192260742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,128,power_law_1.01,0.05559039711952209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,32,power_law_1.01,0.07843199968338013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,128,power_law_1.01,0.05427839756011963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,8,power_law_1.2,0.0883840024471283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,32,power_law_1.01,0.07654399871826172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,128,power_law_1.01,0.05397760272026062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,8,power_law_1.2,0.11496319770812988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,128,power_law_1.01,0.054764801263809205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,32,power_law_1.01,0.0766975998878479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,8,power_law_1.2,0.09415040016174317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,128,power_law_1.01,0.05489280223846436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,8,power_law_1.2,0.08264319896697998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,32,power_law_1.01,0.07909759879112244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,128,power_law_1.01,0.05806080102920532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,8,power_law_1.2,0.1020799994468689
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,8,power_law_1.2,0.12585599422454835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,128,power_law_1.01,0.06490880250930786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,32,power_law_1.01,0.07966079711914062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,8,power_law_1.2,0.13550080060958863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,128,power_law_1.01,0.06655359864234925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,4,power_law_1.2,0.06853119730949402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,8,power_law_1.2,0.17175040245056153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,4,power_law_1.2,0.05798400044441223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,32,power_law_1.01,0.08142719864845276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,128,power_law_1.01,0.07638400197029113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,4,power_law_1.2,0.0652288019657135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,4,power_law_1.2,0.08131200075149536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,8,power_law_1.2,0.1775871992111206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,4,power_law_1.2,0.0898751974105835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,128,power_law_1.01,0.07809280157089234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,32,power_law_1.01,0.08195199966430664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,4,power_law_1.2,0.12268799543380737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,8,power_law_1.2,0.18435839414596558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,4,power_law_1.2,0.12566399574279785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,128,power_law_1.01,0.08478720188140869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,4,power_law_1.2,0.12785279750823975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,8,power_law_1.2,0.19177600145339965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,4,power_law_1.2,0.1270784020423889
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,32,power_law_1.01,0.08286719918251037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,4,power_law_1.2,0.13438719511032104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,128,power_law_1.01,0.10227199792861938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,8,power_law_1.2,0.20404479503631592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,4,power_law_1.2,0.1379647970199585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,4,power_law_1.2,0.14372479915618896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,8,power_law_1.2,0.20906879901885986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,32,power_law_1.01,0.08858879804611205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,128,power_law_1.01,0.10188800096511841
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,4,power_law_1.2,0.1468608021736145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,4,power_law_1.2,0.14629759788513183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,8,power_law_1.2,0.1997696042060852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,4,power_law_1.2,0.16104320287704468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,128,power_law_1.01,0.12991360425949097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,32,power_law_1.01,0.09144319891929627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,4,power_law_1.2,0.1692031979560852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,8,power_law_1.2,0.21875839233398436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,4,power_law_1.2,0.17084800004959105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,128,power_law_1.01,0.13856639862060546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,4,power_law_1.2,0.19496959447860718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,8,power_law_1.2,0.2070591926574707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,4,power_law_1.2,0.22007040977478026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,32,power_law_1.01,0.10307199954986572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,4,power_law_1.2,0.28345599174499514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,8,power_law_1.2,0.23378560543060303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,128,power_law_1.01,0.18551679849624633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,4,power_law_1.2,0.3043519973754883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,4,power_law_1.2,0.3928960084915161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,8,power_law_1.2,0.2609663963317871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,32,power_law_1.01,0.11679999828338623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,4,power_law_1.2,0.4325439929962158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,128,power_law_1.01,0.18581759929656982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,8,power_law_1.2,0.27660160064697265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,4,power_law_1.2,0.6130688190460205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,32,power_law_1.01,0.11761280298233032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,4,power_law_1.2,0.7826560020446778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,8,power_law_1.2,0.3236991882324219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,128,power_law_1.01,0.23765759468078612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,4,power_law_1.2,1.0616959571838378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,8,power_law_1.2,0.35683839321136473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,32,power_law_1.01,0.143340802192688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,4,power_law_1.2,1.4477760314941406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,128,power_law_1.01,0.3058367967605591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,8,power_law_1.2,0.4465792179107666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,4,power_law_1.2,1.5966272354125977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,8,power_law_1.2,0.48937602043151857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,32,power_law_1.01,0.17171839475631714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,128,power_law_1.01,0.3891648054122925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,8,power_law_1.2,0.6418367862701416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,4,power_law_1.2,2.574380874633789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,32,power_law_1.01,0.2119999885559082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,8,power_law_1.2,0.7980288028717041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,128,power_law_1.01,0.46967039108276365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,32,power_law_1.01,0.22660479545593262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,8,power_law_1.2,1.0684096336364746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,4,power_law_1.2,5.722092819213867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,2,128,power_law_1.01,0.5824448108673096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,8,power_law_1.2,1.4438015937805175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,32,power_law_1.01,0.2907776117324829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,8,power_law_1.2,1.8410879135131837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,32,power_law_1.01,0.3506367921829224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,2,128,power_law_1.01,0.8828800201416016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,8,power_law_1.2,2.3545343399047853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,32,power_law_1.01,0.4830848217010498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,8,power_law_1.2,5.518656158447266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,8,balanced,0.049882665276527405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,8,balanced,0.05388266841570536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,8,balanced,0.05215999980767568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,8,balanced,0.056202664971351624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,8,balanced,0.06619200110435486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,8,balanced,0.09538132945696513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,8,balanced,0.11623467008272807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,8,balanced,0.11787733435630798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,8,balanced,0.11733333269755046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,8,balanced,0.11880000432332356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,8,balanced,0.12146133184432983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,8,balanced,0.12114666899045308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,32,power_law_1.01,0.5587584018707276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,8,balanced,0.123690664768219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,8,balanced,0.12395733594894409
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,8,balanced,0.1323946714401245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,8,balanced,0.13316800196965536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,8,balanced,0.1430400013923645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,8,balanced,0.15154133240381876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,8,balanced,0.15956266721089682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,8,balanced,0.17955732345581055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,8,balanced,0.1999573310216268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,8,balanced,0.24184532960255942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,8,balanced,0.2919626633326213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,2,128,power_law_1.01,1.7938304901123048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,8,balanced,0.3725279966990153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,8,balanced,0.4788479804992676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,8,balanced,0.6426080067952474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,8,balanced,0.835594654083252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,8,balanced,1.0290026664733887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,8,balanced,1.5662345886230469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,8,balanced,3.024229367574056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,4,32,power_law_1.01,0.6821824073791504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,4,32,power_law_1.01,1.114732837677002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,16,power_law_1.01,0.05361279845237732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,16,power_law_1.01,0.04095999896526337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,16,power_law_1.01,0.03532159924507141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,16,power_law_1.01,0.03831680119037628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,2,balanced,0.06273599962393443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,2,balanced,0.0682826687892278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,2,balanced,0.07287466526031494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,2,balanced,0.08267199993133545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,2,balanced,0.10896000266075134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,2,balanced,0.16402133305867514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,16,power_law_1.01,0.04379520118236542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,2,balanced,0.16551466782887778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,2,balanced,0.16378666957219443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,2,balanced,0.16451733311017355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,2,balanced,0.16826667388280234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,2,balanced,0.16683733463287354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,2,balanced,0.16885334253311157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,2,balanced,0.171834667523702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,16,power_law_1.01,0.05272960066795349
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,2,balanced,0.17359999815622965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,2,balanced,0.1823199987411499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,2,balanced,0.18582399686177573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,2,balanced,0.19463467597961426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,16,balanced,0.043791999419530235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,2,balanced,0.21504000822703043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,16,balanced,0.045647998650868736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,2,balanced,0.2339413364728292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,16,power_law_1.01,0.05425919890403748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,16,balanced,0.04155199974775314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,2,balanced,0.27215999364852905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,16,balanced,0.045797333121299744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,16,balanced,0.051669334371884666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,2,balanced,0.3090239961942037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,16,balanced,0.06340266764163971
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,16,balanced,0.06568533182144165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,2,balanced,0.3973546822865804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,16,balanced,0.06512000163396199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,16,balanced,0.06588266789913177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,16,balanced,0.06677866478761037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,2,balanced,0.48309866587320965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,16,power_law_1.01,0.05231999754905701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,16,balanced,0.06858666737874348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,4,32,power_law_1.01,2.2878528594970704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,16,balanced,0.06824533144632976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,16,balanced,0.07143466671307881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,2,balanced,0.7035466829935709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,16,balanced,0.07328000168005626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,16,balanced,0.07707199951012929
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,16,balanced,0.08069866895675659
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,2,balanced,0.8832053343454996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,16,balanced,0.08188266555468242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,16,balanced,0.09121599793434143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,16,balanced,0.09965866804122925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,16,power_law_1.01,0.055769598484039305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,16,balanced,0.11930666367212932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,2,balanced,1.282698631286621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,16,balanced,0.13763200243314108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,16,balanced,0.17726399501164755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,16,balanced,0.19817066192626953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,16,balanced,0.2752000093460083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,2,balanced,1.6745707194010417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,16,balanced,0.32585599025090534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,16,power_law_1.01,0.0548799991607666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,16,balanced,0.4517279863357544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,16,balanced,0.570746660232544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,2,balanced,2.090949376424154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,16,balanced,0.6783040364583334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,16,power_law_1.01,0.05483520030975342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,16,balanced,1.053594668706258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,2,balanced,3.1958773930867515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,16,power_law_1.01,0.05619840025901794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,16,balanced,2.0419467290242515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,16,power_law_1.01,0.05681920051574707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,2,balanced,6.215552012125651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,16,power_law_1.01,0.05811200141906738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,16,power_law_1.01,0.06700159907341004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,16,power_law_1.01,0.06680319905281067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,16,power_law_1.01,0.06850559711456299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,16,power_law_1.01,0.07528319954872131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,16,power_law_1.01,0.0785152018070221
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,16,power_law_1.01,0.08784000277519226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,16,power_law_1.01,0.09815679788589478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,16,power_law_1.01,0.12487679719924927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,16,power_law_1.01,0.1529088020324707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,16,power_law_1.01,0.19267840385437013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,16,power_law_1.01,0.23665919303894042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,16,power_law_1.01,0.34362239837646485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,1,128,power_law_1.2,0.03386240005493164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,16,power_law_1.01,0.4326848030090332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,1,128,power_law_1.2,0.03367680013179779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,16,power_law_1.01,0.46613759994506837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,1,128,power_law_1.2,0.03407999873161316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,1,128,power_law_1.2,0.03443840146064758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,16,power_law_1.01,0.8007552146911621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,1,128,power_law_1.2,0.03342719972133636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,32,4,power_law_1.2,0.04565120041370392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,1,128,power_law_1.2,0.035571199655532834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,8,power_law_1.2,0.04184960126876831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,32,4,power_law_1.2,0.048844799399375916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,1,128,power_law_1.2,0.03496319949626923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,8,power_law_1.2,0.034176000952720643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,32,4,power_law_1.2,0.04695680141448975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,16,power_law_1.01,1.5475263595581055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,8,power_law_1.2,0.03342080116271973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,1,128,power_law_1.2,0.034867200255393985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,8,power_law_1.2,0.03468160033226013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,32,4,power_law_1.2,0.052723199129104614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,1,128,power_law_1.2,0.035071998834609985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,8,power_law_1.2,0.03600000143051148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,8,power_law_1.2,0.040294399857521056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,1,128,power_law_1.2,0.03461759984493255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,32,4,power_law_1.2,0.05889279842376709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,8,power_law_1.2,0.04097279906272888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,1,128,power_law_1.2,0.036083200573921205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,32,4,power_law_1.2,0.06172159910202026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,8,power_law_1.2,0.0412416011095047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,8,power_law_1.2,0.04161919951438904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,1,128,power_law_1.2,0.03542400002479553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,32,4,power_law_1.2,0.0766207993030548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,8,power_law_1.2,0.04203520119190216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,1,128,power_law_1.2,0.03583999872207642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,8,power_law_1.2,0.04330880045890808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,32,4,power_law_1.2,0.08115199804306031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,1,128,power_law_1.2,0.03729279935359955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,8,power_law_1.2,0.0450111985206604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,32,4,power_law_1.2,0.08606719970703125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,8,power_law_1.2,0.045459198951721194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,1,128,power_law_1.2,0.03938559889793396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,8,power_law_1.2,0.04776319861412048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,32,4,power_law_1.2,0.08577920198440551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,1,128,power_law_1.2,0.03948160111904144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,8,power_law_1.2,0.052102398872375486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,8,power_law_1.2,0.054451197385787964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,32,4,power_law_1.2,0.08812159895896912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,1,128,power_law_1.2,0.03948160111904144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,8,power_law_1.2,0.05628799796104431
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,1,128,power_law_1.2,0.04177919924259186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,32,4,power_law_1.2,0.09041919708251953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,8,power_law_1.2,0.06607360243797303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,128,power_law_1.2,0.044627198576927186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,8,power_law_1.2,0.07031040191650391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,32,4,power_law_1.2,0.09409279823303222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,8,power_law_1.2,0.08046720027923585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,128,power_law_1.2,0.05001599788665771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,32,4,power_law_1.2,0.09675520062446594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,8,power_law_1.2,0.1003775954246521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,128,power_law_1.2,0.05944960117340088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,8,power_law_1.2,0.12789119482040406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,32,4,power_law_1.2,0.10486400127410889
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,128,power_law_1.2,0.06903679966926575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,8,power_law_1.2,0.153164803981781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,32,4,power_law_1.2,0.10865919589996338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,8,power_law_1.2,0.22092161178588868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,128,power_law_1.2,0.08435840010643006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,32,4,power_law_1.2,0.12526079416275024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,8,power_law_1.2,0.2873471975326538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,128,power_law_1.2,0.11390080451965331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,32,4,power_law_1.2,0.14286719560623168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,8,power_law_1.2,0.42773118019104006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,128,power_law_1.2,0.14513280391693115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,32,4,power_law_1.2,0.15639679431915282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,8,power_law_1.2,0.49313921928405763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,128,power_law_1.2,0.19861760139465331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,32,4,power_law_1.2,0.18428159952163697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,8,power_law_1.2,0.6902272224426269
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,32,4,power_law_1.2,0.20565760135650635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,128,power_law_1.2,0.2530816078186035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,32,4,power_law_1.2,0.2583679914474487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,8,power_law_1.2,1.1362431526184082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,128,power_law_1.2,0.3096127986907959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,32,4,power_law_1.2,0.33327360153198243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,32,4,power_law_1.2,0.46274561882019044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,128,power_law_1.2,0.47605118751525877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,8,power_law_1.2,2.197862434387207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,32,4,power_law_1.2,0.6137599945068359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,32,4,power_law_1.2,0.9263360023498535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,128,power_law_1.2,0.9267583847045898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,32,4,power_law_1.2,1.1546560287475587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,32,4,power_law_1.2,1.3800512313842774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,32,4,power_law_1.2,2.295903968811035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,32,4,power_law_1.2,4.747872161865234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,1,1,balanced,0.13646933436393738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,1,1,balanced,0.34829866886138916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,1,1,balanced,0.4386880000432332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,1,1,balanced,0.43879465262095135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,1,1,balanced,0.4410613377888997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,1,1,balanced,0.4420586824417114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,1,1,balanced,0.4437813361485799
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,1,1,balanced,0.44570668538411456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,1,1,balanced,0.4442880153656006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,1,1,balanced,0.44737064838409424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,8,4,balanced,0.03862933317820231
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,1,1,balanced,0.45394666989644367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,8,4,balanced,0.03697066754102707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,1,1,balanced,0.45255998770395917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,8,4,balanced,0.05459733307361603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,1,1,balanced,0.4574933449427287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,8,4,balanced,0.05358933409055074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,1,1,balanced,0.4778453509012858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,8,4,balanced,0.05691733459631602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,8,4,balanced,0.05522133409976959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,1,1,balanced,0.4826879898707072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,8,4,balanced,0.056799997886021934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,1,1,balanced,0.4890933434168498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,8,4,balanced,0.05585066477457682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,1,1,balanced,0.5542133251825968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,8,4,balanced,0.05541866521040598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,8,4,balanced,0.05691733459631602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,1,1,balanced,0.5645546515782675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,8,4,balanced,0.06004266440868378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,1,1,balanced,0.609658678372701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,8,4,balanced,0.05921066800753275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,8,4,balanced,0.059392000238100685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,1,1,balanced,0.7928106784820557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,8,4,balanced,0.0613013356924057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,8,4,balanced,0.06596266726652782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,1,1,balanced,0.8757867018381754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,8,4,balanced,0.06478400031725566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,8,4,balanced,0.07308266560236613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,8,4,balanced,0.076773335536321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,1,1,balanced,1.221120039621989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,8,4,balanced,0.08382933338483174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,8,4,balanced,0.09504000345865886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,8,4,balanced,0.10477866729100545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,1,1,balanced,1.4778666496276855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,8,4,balanced,0.12761599818865457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,8,4,balanced,0.1521440049012502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,8,4,balanced,0.19820266962051392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,8,4,balanced,0.25094934304555255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,1,1,balanced,2.180415948232015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,8,4,balanced,0.3407040039698283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,8,4,balanced,0.41302935282389325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,8,4,balanced,0.5054346720377604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,1,1,balanced,2.8626187642415366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,8,4,balanced,0.7637973626454672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,8,4,balanced,1.4347200393676758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,1,1,balanced,4.250405311584473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,1,1,balanced,5.58949343363444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,1,1,balanced,6.986805597941081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,1,1,balanced,11.191610972086588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,16,balanced,0.05234666665395101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,16,balanced,0.05003199974695841
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,16,balanced,0.07032533486684163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,16,balanced,0.0958133339881897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,16,balanced,0.1458239952723185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,16,balanced,0.23988266785939535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,16,balanced,0.2429386576016744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,16,balanced,0.24372265736262003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,16,balanced,0.24569600820541382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,16,balanced,0.24597867329915366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,16,balanced,0.2461706598599752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,16,balanced,0.24785067637761435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,16,balanced,0.24901332457860312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,16,balanced,0.24905065695444742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,16,balanced,0.25380265712738037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,16,balanced,0.2566133340199788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,16,balanced,0.25757332642873126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,16,balanced,0.26522133747736615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,16,balanced,0.27406400442123413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,16,balanced,0.2833706736564636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,16,balanced,0.3086133400599162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,16,balanced,0.33353598912556964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,16,balanced,0.35443198680877686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,16,balanced,0.4317866563796997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,16,balanced,0.46990398565928143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,16,balanced,0.6191039880116781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,16,balanced,0.6973973115285238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,1,1,balanced,23.209915161132812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,16,balanced,0.9227840105692545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,16,balanced,1.2521706422170003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,16,balanced,2.4113547007242837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,16,power_law_1.2,0.07605119943618774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,4,power_law_1.01,0.032646399736404416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,4,power_law_1.01,0.033055999875068666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,16,power_law_1.2,0.09784960150718688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,4,power_law_1.01,0.0339711993932724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,4,power_law_1.01,0.03728640079498291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,16,power_law_1.2,0.05363199710845947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,4,power_law_1.01,0.04019840061664581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,4,power_law_1.01,0.0515392005443573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,16,power_law_1.2,0.059468799829483034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,4,power_law_1.01,0.054092800617218016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,16,power_law_1.2,0.0663424015045166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,4,power_law_1.01,0.05472000241279602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,4,power_law_1.01,0.055961602926254274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,16,power_law_1.2,0.07539839744567871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,4,power_law_1.01,0.05697280168533325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,4,power_law_1.01,0.05912320017814636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,16,power_law_1.2,0.08609279990196228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,4,power_law_1.01,0.060224002599716185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,4,power_law_1.01,0.06117119789123535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,16,power_law_1.2,0.09208319783210754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,4,power_law_1.01,0.06435840129852295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,16,power_law_1.2,0.08747519850730896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,4,power_law_1.01,0.07066879868507385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,4,power_law_1.01,0.0759552001953125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,16,power_law_1.2,0.09354239702224731
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,4,power_law_1.01,0.07585279941558838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,4,power_law_1.01,0.08485119938850402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,16,power_law_1.2,0.09813759922981262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,4,power_law_1.01,0.09315840005874634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,16,power_law_1.2,0.0973695993423462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,4,power_law_1.01,0.11304960250854493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,4,power_law_1.01,0.1266816020011902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,16,power_law_1.2,0.10156799554824829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,4,power_law_1.01,0.1674239993095398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,16,power_law_1.2,0.10469119548797608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,4,power_law_1.01,0.2192768096923828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,16,power_law_1.2,0.11156480312347412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,4,power_law_1.01,0.28072319030761717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,16,power_law_1.2,0.11807359457015991
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,4,power_law_1.01,0.3371135950088501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,16,power_law_1.2,0.12284159660339355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,4,power_law_1.01,0.5028672218322754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,16,power_law_1.2,0.13921279907226564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,4,power_law_1.01,0.6674496173858643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,16,power_law_1.2,0.15552639961242676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,4,power_law_1.01,0.8241984367370605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,16,power_law_1.2,0.18770560026168823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,16,power_law_1.2,0.20804479122161865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,4,power_law_1.01,1.1695167541503906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,16,power_law_1.2,0.2731839895248413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,16,power_law_1.2,0.32062718868255613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,4,power_law_1.01,2.3795391082763673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,16,power_law_1.2,0.4817215919494629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,16,power_law_1.2,0.5841023921966553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,16,power_law_1.2,0.8456128120422364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,16,power_law_1.2,0.9986751556396485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,16,power_law_1.2,1.3132032394409179
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,16,power_law_1.2,1.8607423782348633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,16,power_law_1.2,3.8824512481689455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,64,balanced,0.05433600147565206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,64,balanced,0.05097599824269613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,64,balanced,0.048672000567118325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,64,balanced,0.053871999184290566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,64,balanced,0.05406933526198069
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,64,balanced,0.05407999952634176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,64,balanced,0.05585599939028422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,64,balanced,0.055733333031336464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,64,balanced,0.05604266623655955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,64,balanced,0.05613866448402405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,64,balanced,0.05789333085219065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,64,balanced,0.06017066538333893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,64,balanced,0.060138667623202004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,64,balanced,0.060693333546320595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,64,balanced,0.06622933348019917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,64,balanced,0.06671466430028279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,64,balanced,0.06849599877993266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,64,balanced,0.08889599641164143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,64,balanced,0.08738133311271667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,64,balanced,0.11346133550008138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,64,balanced,0.1299253304799398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,64,balanced,0.16927466789881387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,64,balanced,0.15822399655977884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,64,balanced,0.20458134015401205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,64,balanced,0.21161067485809326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,64,balanced,0.2858346700668335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,64,balanced,0.3325226704279582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,4,64,balanced,0.4071520169576009
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,4,64,balanced,0.5744533141454061
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,4,64,balanced,1.2477440039316814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,64,power_law_1.01,0.03635199964046478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,64,power_law_1.01,0.035283198952674864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,64,power_law_1.01,0.032262399792671204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,64,power_law_1.01,0.03281919956207276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,64,power_law_1.01,0.033062401413917544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,16,power_law_1.2,0.03735679984092712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,64,power_law_1.01,0.034534400701522826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,64,power_law_1.01,0.03460479974746704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,16,power_law_1.2,0.031193599104881287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,64,power_law_1.01,0.03390080034732819
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,16,power_law_1.2,0.02876800000667572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,64,power_law_1.01,0.03408640027046204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,16,power_law_1.2,0.030668801069259642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,64,power_law_1.01,0.03446399867534637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,16,power_law_1.2,0.03075839877128601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,16,power_law_1.2,0.033657601475715636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,64,power_law_1.01,0.03521279990673065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,16,power_law_1.2,0.03394559919834137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,64,power_law_1.01,0.03531520068645477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,16,power_law_1.2,0.03439359962940216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,64,power_law_1.01,0.03592320084571839
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,16,power_law_1.2,0.03431040048599243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,64,power_law_1.01,0.036800000071525577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,16,power_law_1.2,0.03553279936313629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,64,power_law_1.01,0.039724799990653994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,16,power_law_1.2,0.0358271986246109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,64,power_law_1.01,0.040217599272727965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,16,power_law_1.2,0.03653759956359863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,64,power_law_1.01,0.04051840007305145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,16,power_law_1.2,0.03871360123157501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,64,power_law_1.01,0.04170880019664765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,16,power_law_1.2,0.04106239974498749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,64,power_law_1.01,0.04420480132102966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,16,power_law_1.2,0.04714879989624023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,64,power_law_1.01,0.0493120014667511
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,16,power_law_1.2,0.047891199588775635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,64,power_law_1.01,0.055379199981689456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,16,power_law_1.2,0.050387197732925416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,64,power_law_1.01,0.06270719766616821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,16,power_law_1.2,0.05738239884376526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,16,power_law_1.2,0.059628802537918094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,64,power_law_1.01,0.07555840015411378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,16,power_law_1.2,0.0702015995979309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,64,power_law_1.01,0.08828160166740417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,16,power_law_1.2,0.0695680022239685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,64,power_law_1.01,0.11853439807891845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,16,power_law_1.2,0.09564160108566284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,16,power_law_1.2,0.1114240050315857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,64,power_law_1.01,0.1657663941383362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,16,power_law_1.2,0.13983999490737914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,64,power_law_1.01,0.2188096046447754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,16,power_law_1.2,0.17830400466918944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,16,power_law_1.2,0.23612160682678224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,64,power_law_1.01,0.25077760219573975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,16,power_law_1.2,0.2838207960128784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,64,power_law_1.01,0.3814271926879883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,2,16,power_law_1.2,0.3840320110321045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,2,16,power_law_1.2,0.6090879917144776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,64,power_law_1.01,0.7377535820007324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,2,16,power_law_1.2,1.4309120178222656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,64,power_law_1.2,0.06647679805755616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,2,power_law_1.01,0.06778240203857422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,64,power_law_1.2,0.06617599725723267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,2,power_law_1.01,0.07394559979438782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,2,power_law_1.01,0.08261759877204895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,64,power_law_1.2,0.05529599785804749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,2,power_law_1.01,0.08543360233306885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,64,power_law_1.2,0.05608320236206055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,2,power_law_1.01,0.1028223991394043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,64,power_law_1.2,0.05775359869003296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,2,power_law_1.01,0.13337600231170654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,2,power_law_1.01,0.14428800344467163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,64,power_law_1.2,0.059654402732849124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,2,power_law_1.01,0.16192640066146852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,64,power_law_1.2,0.05973759889602661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,2,power_law_1.01,0.16784640550613403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,64,power_law_1.2,0.05990399718284607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,2,power_law_1.01,0.17301119565963746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,64,power_law_1.2,0.06024320125579834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,2,power_law_1.01,0.1776576042175293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,64,power_law_1.2,0.0603007972240448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,2,power_law_1.01,0.18491519689559938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,2,power_law_1.01,0.19124480485916137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,64,power_law_1.2,0.06054400205612183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,2,power_law_1.01,0.19841920137405394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,64,power_law_1.2,0.06375679969787598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,2,power_law_1.01,0.21185920238494874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,64,power_law_1.2,0.06472960114479065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,2,power_law_1.01,0.2138688087463379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,2,power_law_1.01,0.23113598823547363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,64,power_law_1.2,0.06592640280723572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,2,power_law_1.01,0.2503040075302124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,64,power_law_1.2,0.07404800057411194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,2,power_law_1.01,0.27962241172790525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,64,power_law_1.2,0.07640320062637329
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,2,power_law_1.01,0.334604811668396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,64,power_law_1.2,0.08557440042495727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,2,power_law_1.01,0.39064960479736327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,2,power_law_1.01,0.49431681632995605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,64,power_law_1.2,0.09976959824562073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,2,power_law_1.01,0.6005951881408691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,64,power_law_1.2,0.11310720443725586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,2,power_law_1.01,0.8261247634887695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,64,power_law_1.2,0.14176000356674195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,2,power_law_1.01,1.0747520446777343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,64,power_law_1.2,0.1631999969482422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,2,power_law_1.01,1.5283583641052245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,64,power_law_1.2,0.22417280673980713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,2,power_law_1.01,1.9104192733764649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,64,power_law_1.2,0.2256319999694824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,2,power_law_1.01,2.329324722290039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,64,power_law_1.2,0.3177023887634277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,64,power_law_1.2,0.40474882125854494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,2,power_law_1.01,3.7267902374267576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,64,power_law_1.2,0.558835220336914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,2,power_law_1.01,7.110451507568359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,64,power_law_1.2,0.698252820968628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,2,64,power_law_1.2,0.9149951934814453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,2,64,power_law_1.2,1.6548608779907226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,2,64,power_law_1.2,3.0351871490478515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,2,power_law_1.2,0.06756479740142822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,2,power_law_1.2,0.07077760100364686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,2,power_law_1.2,0.08106240034103393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,2,power_law_1.2,0.09344000220298768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,2,power_law_1.2,0.12804479598999025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,2,power_law_1.2,0.18480000495910645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,2,power_law_1.2,0.19820159673690796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,2,power_law_1.2,0.20436480045318603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,2,power_law_1.2,0.21062400341033935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,2,power_law_1.2,0.2174976110458374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,2,power_law_1.2,0.2203200101852417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,2,power_law_1.2,0.2294015884399414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,2,power_law_1.2,0.2318272113800049
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,2,power_law_1.2,0.24008960723876954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,2,power_law_1.2,0.25274879932403566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,2,power_law_1.2,0.2609407901763916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,2,power_law_1.2,0.27628159523010254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,2,power_law_1.2,0.3049344062805176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,2,power_law_1.2,0.35491199493408204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,2,power_law_1.2,0.43079681396484376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,2,power_law_1.2,0.4761472225189209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,2,power_law_1.2,0.5888448238372803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,2,power_law_1.2,0.7388160228729248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,2,power_law_1.2,0.9209535598754883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,2,power_law_1.2,1.158937644958496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,2,power_law_1.2,1.5186752319335937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,2,power_law_1.2,2.190048027038574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,2,power_law_1.2,2.5828351974487305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,2,power_law_1.2,3.912793731689453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,2,power_law_1.2,7.804653167724609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,16,power_law_1.2,0.03362559974193573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,16,power_law_1.2,0.03048959970474243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,16,power_law_1.2,0.030073601007461547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,16,power_law_1.2,0.03057279884815216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,16,power_law_1.2,0.031033599376678468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,16,power_law_1.2,0.03325439989566803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,16,power_law_1.2,0.03377279937267304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,16,power_law_1.2,0.034227201342582704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,16,power_law_1.2,0.034892800450325015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,16,power_law_1.2,0.03516159951686859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,16,power_law_1.2,0.035571199655532834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,16,power_law_1.2,0.03633280098438263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,16,power_law_1.2,0.03734399974346161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,16,power_law_1.2,0.04016000032424927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,16,power_law_1.2,0.0450111985206604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,16,power_law_1.2,0.04540160000324249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,16,power_law_1.2,0.047737601399421695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,16,power_law_1.2,0.05294719934463501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,16,power_law_1.2,0.058963197469711306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,16,power_law_1.2,0.07255039811134338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,16,power_law_1.2,0.08672000169754028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,16,power_law_1.2,0.11459200382232666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,16,power_law_1.2,0.15148160457611085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,16,power_law_1.2,0.22179839611053467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,16,power_law_1.2,0.300601601600647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,16,power_law_1.2,0.4233151912689209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,16,power_law_1.2,0.5110784053802491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,16,power_law_1.2,0.6681536197662353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,16,power_law_1.2,1.11778564453125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,16,power_law_1.2,2.908006477355957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,64,power_law_1.01,0.06279039978981019
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,64,power_law_1.01,0.05568000078201294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,64,power_law_1.01,0.05235199928283692
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,64,power_law_1.01,0.05135999917984009
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,64,power_law_1.01,0.05220479965209961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,64,power_law_1.01,0.052339202165603636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,64,power_law_1.01,0.05360640287399292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,64,power_law_1.01,0.055103999376296994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,64,power_law_1.01,0.054553598165512085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,64,power_law_1.01,0.055027198791503903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,64,power_law_1.01,0.05532159805297852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,64,power_law_1.01,0.05804799795150757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,64,power_law_1.01,0.05777919888496399
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,64,power_law_1.01,0.05980799794197082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,64,power_law_1.01,0.06488959789276123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,64,power_law_1.01,0.06607360243797303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,64,power_law_1.01,0.07077760100364686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,64,power_law_1.01,0.07917439937591553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,64,power_law_1.01,0.08729599714279175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,64,power_law_1.01,0.1084671974182129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,64,power_law_1.01,0.12658560276031494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,64,power_law_1.01,0.16406400203704835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,64,power_law_1.01,0.18803839683532714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,64,power_law_1.01,0.2569727897644043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,64,power_law_1.01,0.3374783992767334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,64,power_law_1.01,0.4547904014587402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,64,power_law_1.01,0.5797056198120117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,64,power_law_1.01,0.699782419204712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,64,power_law_1.01,1.1107711791992188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,64,power_law_1.01,2.2177024841308595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,4,2,power_law_1.2,0.04359680116176605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,4,2,power_law_1.2,0.03896960020065308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,4,2,power_law_1.2,0.04073599874973297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,4,2,power_law_1.2,0.04418559968471527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,4,2,power_law_1.2,0.04663679897785187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,4,2,power_law_1.2,0.05552639961242676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,4,2,power_law_1.2,0.05790079832077026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,4,2,power_law_1.2,0.05809919834136963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,4,2,power_law_1.2,0.05913599729537964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,4,2,power_law_1.2,0.06108160018920898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,4,2,power_law_1.2,0.06094719767570496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,4,2,power_law_1.2,0.06521599888801574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,4,2,power_law_1.2,0.06718720197677612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,4,2,power_law_1.2,0.06801279783248901
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,4,2,power_law_1.2,0.07486720085144043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,4,2,power_law_1.2,0.0778880000114441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,4,2,power_law_1.2,0.08359040021896362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,4,2,power_law_1.2,0.10032000541687011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,4,2,power_law_1.2,0.10872960090637207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,4,2,power_law_1.2,0.13653119802474975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,4,2,power_law_1.2,0.16236799955368042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,4,2,power_law_1.2,0.22211840152740478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,4,2,power_law_1.2,0.27231359481811523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,4,2,power_law_1.2,0.40291199684143064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,4,2,power_law_1.2,0.4960319995880127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,4,2,power_law_1.2,0.7530176162719726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,4,2,power_law_1.2,1.0327808380126953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,4,2,power_law_1.2,1.3638400077819823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,4,2,power_law_1.2,2.0174400329589846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,4,2,power_law_1.2,3.738809585571289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,8,8,power_law_1.2,0.03438720107078552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,8,8,power_law_1.2,0.035820800065994265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,8,8,power_law_1.2,0.03320960104465485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,8,8,power_law_1.2,0.03342080116271973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,8,8,power_law_1.2,0.03361920118331909
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,8,8,power_law_1.2,0.03436160087585449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,8,8,power_law_1.2,0.034918400645256045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,8,8,power_law_1.2,0.03561600148677826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,8,8,power_law_1.2,0.03591040074825287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,8,8,power_law_1.2,0.036032000184059144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,8,8,power_law_1.2,0.03592959940433502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,8,8,power_law_1.2,0.03689599931240082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,8,8,power_law_1.2,0.037427198886871335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,8,8,power_law_1.2,0.038252800703048706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,8,8,power_law_1.2,0.04423680007457733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,8,8,power_law_1.2,0.047603198885917665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,8,8,power_law_1.2,0.04960640072822571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,8,8,power_law_1.2,0.05508480072021484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,8,8,power_law_1.2,0.06051200032234192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,8,8,power_law_1.2,0.07362560033798218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,8,8,power_law_1.2,0.08035200238227844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,8,8,power_law_1.2,0.10669440031051636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,8,8,power_law_1.2,0.12956160306930542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,8,8,power_law_1.2,0.17857919931411742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,8,8,power_law_1.2,0.21552639007568358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,8,8,power_law_1.2,0.30471680164337156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,8,8,power_law_1.2,0.38769919872283937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,8,8,power_law_1.2,0.4776576042175293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,8,8,power_law_1.2,0.724076795578003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,8,8,power_law_1.2,1.3758272171020507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,16,1,balanced,0.03756800045569738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,16,1,balanced,0.041690667470296226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,16,1,balanced,0.04382933179537455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,16,1,balanced,0.04295999805132548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,16,1,balanced,0.04572266836961111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,16,1,balanced,0.04502933224042257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,16,1,balanced,0.04762133459250132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,16,1,balanced,0.045978665351867676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,16,1,balanced,0.04574933151404063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,16,1,balanced,0.04781333108743032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,16,1,balanced,0.04841599861780802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,16,1,balanced,0.04977599779764811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,16,1,balanced,0.049695998430252075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,16,1,balanced,0.05576533575852712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,16,1,balanced,0.06435733536879222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,16,1,balanced,0.0642080008983612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,16,1,balanced,0.0765173335870107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,16,1,balanced,0.0895253320535024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,16,1,balanced,0.10072533289591472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,16,1,balanced,0.11957866946856181
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,16,1,balanced,0.1397706667582194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,16,1,balanced,0.18146133422851562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,16,1,balanced,0.21633599201838175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,16,1,balanced,0.3001760045687358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,16,1,balanced,0.3856213490168254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,16,1,balanced,0.5541386604309082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,16,1,balanced,0.7194506327311198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,16,1,balanced,0.8838026523590088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,16,1,balanced,1.371247927347819
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,16,1,balanced,2.6804478963216147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,2,power_law_1.2,0.05690240263938904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,2,power_law_1.2,0.06059520244598389
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,2,power_law_1.2,0.06939520239830017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,2,power_law_1.2,0.08092799782752991
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,2,power_law_1.2,0.09824000000953674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,2,power_law_1.2,0.13304959535598754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,2,power_law_1.2,0.1407871961593628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,2,power_law_1.2,0.14404480457305907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,2,power_law_1.2,0.15160319805145264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,2,power_law_1.2,0.15251840353012086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,2,power_law_1.2,0.16017919778823853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,2,power_law_1.2,0.16634880304336547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,2,power_law_1.2,0.16887680292129517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,2,power_law_1.2,0.18150399923324584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,2,power_law_1.2,0.18903679847717286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,2,power_law_1.2,0.202508807182312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,2,power_law_1.2,0.2006592035293579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,2,power_law_1.2,0.22924160957336426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,2,power_law_1.2,0.2638592004776001
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,2,power_law_1.2,0.3054464101791382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,2,power_law_1.2,0.36894080638885496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,2,power_law_1.2,0.4567423820495605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,2,power_law_1.2,0.5850240230560303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,2,power_law_1.2,0.7568575859069824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,2,power_law_1.2,1.0605504035949707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,2,power_law_1.2,1.4787903785705567
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,2,power_law_1.2,1.772857666015625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,2,power_law_1.2,2.284121513366699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,2,power_law_1.2,3.4432193756103517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,2,power_law_1.2,7.602278137207032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,16,power_law_1.01,0.08574720025062561
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,16,power_law_1.01,0.1283519983291626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,16,power_law_1.01,0.07077119946479797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,16,power_law_1.01,0.07474560141563416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,16,power_law_1.01,0.07919999957084656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,16,power_law_1.01,0.09086080193519593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,16,power_law_1.01,0.10232319831848144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,16,power_law_1.01,0.1028480052947998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,16,power_law_1.01,0.10424959659576416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,16,power_law_1.01,0.10435839891433715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,16,power_law_1.01,0.1093440055847168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,16,power_law_1.01,0.10956799983978271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,16,power_law_1.01,0.11050239801406861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,16,power_law_1.01,0.1145408034324646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,16,power_law_1.01,0.11887999773025512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,16,power_law_1.01,0.12409600019454955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,16,power_law_1.01,0.13030400276184081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,16,power_law_1.01,0.14120320081710816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,16,power_law_1.01,0.15400960445404052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,16,power_law_1.01,0.18490240573883057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,16,power_law_1.01,0.2070847988128662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,16,power_law_1.01,0.2691263914108276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,16,power_law_1.01,0.28838400840759276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,16,power_law_1.01,0.3868544101715088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,16,power_law_1.01,0.4601408004760742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,16,power_law_1.01,0.6413055896759033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,16,power_law_1.01,0.8192895889282227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,16,power_law_1.01,1.0162943840026855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,16,power_law_1.01,1.5748096466064454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,16,power_law_1.01,2.888422393798828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,32,4,power_law_1.01,0.045664000511169436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,32,4,power_law_1.01,0.05047680139541626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,32,4,power_law_1.01,0.0482367992401123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,32,4,power_law_1.01,0.05169919729232788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,32,4,power_law_1.01,0.058297598361968996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,32,4,power_law_1.01,0.06679040193557739
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,32,4,power_law_1.01,0.07650560140609741
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,32,4,power_law_1.01,0.08138239979743958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,32,4,power_law_1.01,0.08631680011749268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,32,4,power_law_1.01,0.08507519960403442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,32,4,power_law_1.01,0.08956159949302674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,32,4,power_law_1.01,0.08981119990348815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,32,4,power_law_1.01,0.09283199906349182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,32,4,power_law_1.01,0.0957759976387024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,32,4,power_law_1.01,0.10661760568618775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,32,4,power_law_1.01,0.10729600191116333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,32,4,power_law_1.01,0.12071679830551148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,32,4,power_law_1.01,0.13596160411834718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,32,4,power_law_1.01,0.14852479696273804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,32,4,power_law_1.01,0.18227200508117675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,32,4,power_law_1.01,0.20921599864959717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,32,4,power_law_1.01,0.2624959945678711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,32,4,power_law_1.01,0.32977919578552245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,32,4,power_law_1.01,0.4417856216430664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,32,4,power_law_1.01,0.5556032180786132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,32,4,power_law_1.01,0.7725183963775635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,32,4,power_law_1.01,1.0490559577941894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,4,power_law_1.2,0.03681919872760773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,4,power_law_1.2,0.03500800132751465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,32,4,power_law_1.01,1.2411968231201171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,4,power_law_1.2,0.039392000436782836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,4,power_law_1.2,0.04672000110149384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,4,power_law_1.2,0.049395200610160825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,4,power_law_1.2,0.07390720248222352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,32,4,power_law_1.01,1.8659328460693358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,4,power_law_1.2,0.07638400197029113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,4,power_law_1.2,0.07786880135536194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,4,power_law_1.2,0.08083840012550354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,4,power_law_1.2,0.08339200019836426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,4,power_law_1.2,0.08494719862937927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,4,power_law_1.2,0.08637440204620361
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,32,4,power_law_1.01,3.8486080169677734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,4,power_law_1.2,0.08701440095901489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,4,power_law_1.2,0.09233279824256897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,4,power_law_1.2,0.09841920137405395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,4,power_law_1.2,0.1014143943786621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,4,power_law_1.2,0.1083840012550354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,4,power_law_1.2,0.12293119430541992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,4,power_law_1.2,0.13871999979019164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,4,power_law_1.2,0.1808575987815857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,4,power_law_1.2,0.1996351957321167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,4,power_law_1.2,0.24700160026550294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,4,power_law_1.2,0.3252928018569946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,4,power_law_1.2,0.3968127965927124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,4,power_law_1.2,0.5800576210021973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,4,power_law_1.2,0.7881855964660645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,4,power_law_1.2,1.0945856094360351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,4,power_law_1.2,1.2390975952148438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,1,balanced,0.03769599894682566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,1,balanced,0.03972800076007843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,1,balanced,0.057999998331069946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,1,balanced,0.08250133196512859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,1,balanced,0.13750933607419333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,1,balanced,0.133242666721344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,1,balanced,0.13237333297729492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,1,balanced,0.1341546674569448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,1,balanced,0.1376106639703115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,1,balanced,0.14189866185188293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,1,balanced,0.13910399874051413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,1,balanced,0.14355199535687765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,1,balanced,0.1476959983507792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,1,balanced,0.15477866927782694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,1,balanced,0.1601706643899282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,1,balanced,0.16637333234151205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,1,balanced,0.17631999651590982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,1,balanced,0.19964800278345743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,1,balanced,0.21980265776316324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,4,power_law_1.2,2.0609792709350585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,1,balanced,0.2807839910189311
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,1,balanced,0.3039253354072571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,1,balanced,0.42054398854573566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,1,balanced,0.49425601959228516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,1,balanced,0.6816159884134928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,1,balanced,0.8697066307067871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,1,balanced,1.2731520334879558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,1,balanced,1.635503927866618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,1,balanced,2.009061336517334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,1,balanced,3.0769761403401694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,4,power_law_1.2,4.897375869750976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,1,balanced,5.936789194742839
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,4,balanced,0.03336533407370249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,4,balanced,0.0372533326347669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,4,balanced,0.05364799996217092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,4,balanced,0.07648000121116638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,4,balanced,0.12651733557383218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,4,balanced,0.11962133646011353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,4,balanced,0.12014933427174886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,4,balanced,0.12092266480127971
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,4,balanced,0.12130133310953777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,4,balanced,0.12313066919644673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,4,balanced,0.12389333049456279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,4,balanced,0.12361066540082295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,4,balanced,0.12395733594894409
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,4,balanced,0.12556800246238708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,4,balanced,0.13434666395187378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,4,balanced,0.13193066914876303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,16,power_law_1.2,0.09032959938049316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,4,balanced,0.13518399993578592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,4,balanced,0.1406613290309906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,4,balanced,0.14884266257286072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,4,balanced,0.1698453426361084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,4,balanced,0.17601066827774048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,4,balanced,0.21930134296417236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,4,balanced,0.24203733603159586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,4,balanced,0.3234613339106242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,4,balanced,0.37012799580891925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,16,power_law_1.2,0.04152320027351379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,4,balanced,0.4965386788050334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,4,balanced,0.6378186543782552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,4,balanced,0.7623466650644938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,4,balanced,1.1715199947357178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,16,power_law_1.2,0.03395200073719025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,4,balanced,2.222346623738607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,16,power_law_1.2,0.03608959913253784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,16,power_law_1.2,0.03950079977512359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,16,power_law_1.2,0.049593600630760196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,16,power_law_1.2,0.052211201190948485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,16,power_law_1.2,0.05057920217514038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,16,power_law_1.2,0.054585599899291994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,16,power_law_1.2,0.05331839919090271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,16,power_law_1.2,0.05383679866790771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,16,power_law_1.2,0.056985598802566526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,16,power_law_1.2,0.05785599946975708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,16,power_law_1.2,0.059552001953125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,16,power_law_1.2,0.06480640172958374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,16,power_law_1.2,0.06399359703063964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,16,power_law_1.2,0.06913279891014099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,16,power_law_1.2,0.0816703975200653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,16,power_law_1.2,0.07518720030784606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,16,power_law_1.2,0.09186559915542603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,16,power_law_1.2,0.09401599764823913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,16,power_law_1.2,0.11537280082702636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,16,power_law_1.2,0.1362239956855774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,64,balanced,0.0641599992911021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,64,balanced,0.06439466774463654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,16,power_law_1.2,0.18833279609680176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,64,balanced,0.06486933430035909
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,64,balanced,0.06411733229955037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,64,balanced,0.07491733133792877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,64,balanced,0.09106666843096416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,64,balanced,0.08980799714724223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,64,balanced,0.09214400251706441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,64,balanced,0.09311466415723164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,64,balanced,0.09104532996813457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,64,balanced,0.0997866690158844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,64,balanced,0.09471467137336731
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,64,balanced,0.10311999917030334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,64,balanced,0.09803199768066406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,16,power_law_1.2,0.24696319103240966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,64,balanced,0.1032480001449585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,64,balanced,0.10522666573524475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,64,balanced,0.10578133662541707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,64,balanced,0.11734933654467265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,64,balanced,0.12232533097267151
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,64,balanced,0.1381333371003469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,64,balanced,0.15457600355148315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,64,balanced,0.19100799163182577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,64,balanced,0.22387200593948364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,64,balanced,0.3009066581726074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,16,power_law_1.2,0.35620479583740233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,64,balanced,0.36510932445526123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,64,balanced,0.512719988822937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,64,balanced,0.6547626654307047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,64,balanced,0.8009706338246664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,64,balanced,1.2248480319976807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,16,power_law_1.2,0.34104959964752196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,64,balanced,2.413952032725016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,1,16,power_law_1.2,0.48210558891296384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,1,16,power_law_1.2,0.6454527854919434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,1,16,power_law_1.2,1.4421055793762207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,2,2,power_law_1.2,0.07802879810333252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,2,2,power_law_1.2,0.0902079999446869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,2,2,power_law_1.2,0.09750400185585022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,2,2,power_law_1.2,0.1347584009170532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,2,2,power_law_1.2,0.1342527985572815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,2,2,power_law_1.2,0.13605120182037353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,2,2,power_law_1.2,0.14236799478530884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,2,2,power_law_1.2,0.1434816002845764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,2,2,power_law_1.2,0.1436735987663269
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,2,2,power_law_1.2,0.14316799640655517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,2,2,power_law_1.2,0.14447360038757323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,2,2,power_law_1.2,0.15386879444122314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,2,2,power_law_1.2,0.15443840026855468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,2,2,power_law_1.2,0.16611839532852174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,2,2,power_law_1.2,0.20259199142456055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,2,2,power_law_1.2,0.20562560558319093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,2,2,power_law_1.2,0.2101439952850342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,2,2,power_law_1.2,0.23466880321502687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,2,2,power_law_1.2,0.266048002243042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,2,2,power_law_1.2,0.34311039447784425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,2,2,power_law_1.2,0.3734463930130005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,8,power_law_1.2,0.033983999490737916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,2,2,power_law_1.2,0.5321216106414794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,2,2,power_law_1.2,0.7249855995178223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,8,power_law_1.2,0.0293503999710083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,2,2,power_law_1.2,0.9928192138671875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,8,power_law_1.2,0.029414400458335876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,2,2,power_law_1.2,1.1147711753845215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,8,power_law_1.2,0.030112001299858093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,2,2,power_law_1.2,2.0746559143066405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,8,power_law_1.2,0.031206399202346802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,8,power_law_1.2,0.03451519906520843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,2,2,power_law_1.2,2.718489646911621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,8,power_law_1.2,0.03551360070705414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,8,power_law_1.2,0.03523840010166168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,2,2,power_law_1.2,3.0092031478881838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,8,power_law_1.2,0.03503359854221344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,8,power_law_1.2,0.03646720051765442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,2,2,power_law_1.2,5.274105453491211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,8,power_law_1.2,0.03612799942493439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,8,power_law_1.2,0.037868800759315493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,8,power_law_1.2,0.03947519958019256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,8,power_law_1.2,0.041247999668121337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,8,power_law_1.2,0.0473471999168396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,2,2,power_law_1.2,10.002035522460938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,8,power_law_1.2,0.0490880012512207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,8,power_law_1.2,0.05065600275993347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,8,power_law_1.2,0.056390398740768434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,8,power_law_1.2,0.06576640009880066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,8,power_law_1.2,0.08001919984817504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,8,power_law_1.2,0.10005120038986207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,8,power_law_1.2,0.11827199459075928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,8,power_law_1.2,0.16428159475326537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,8,power_law_1.2,0.2352128028869629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,8,power_law_1.2,0.27820160388946535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,8,power_law_1.2,0.3861567974090576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,8,power_law_1.2,0.4678080081939697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,8,power_law_1.2,0.7076863765716552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,8,power_law_1.2,1.0572735786437988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,16,8,power_law_1.01,0.039238399267196654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,16,8,power_law_1.01,0.03551360070705414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,16,8,power_law_1.01,0.0328575998544693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,16,8,power_law_1.01,0.032915198802948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,16,8,power_law_1.01,0.03304319977760315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,16,8,power_law_1.01,0.03408640027046204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,16,8,power_law_1.01,0.03405439853668213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,16,8,power_law_1.01,0.035097599029541016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,16,8,power_law_1.01,0.03531520068645477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,8,power_law_1.2,2.201785659790039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,16,8,power_law_1.01,0.03612799942493439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,16,8,power_law_1.01,0.03608959913253784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,16,8,power_law_1.01,0.03718400001525879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,16,8,power_law_1.01,0.037273600697517395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,16,8,power_law_1.01,0.03704319894313812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,16,8,power_law_1.01,0.04162560105323791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,16,8,power_law_1.01,0.044223999977111815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,16,8,power_law_1.01,0.04524799883365631
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,16,8,power_law_1.01,0.05021439790725708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,16,8,power_law_1.01,0.057171201705932616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,16,8,power_law_1.01,0.06437119841575623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,16,8,power_law_1.01,0.06908159852027893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,16,8,power_law_1.01,0.09132159948348999
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,16,8,power_law_1.01,0.11404160261154175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,16,8,power_law_1.01,0.1565824031829834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,16,8,power_law_1.01,0.19051519632339478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,16,8,power_law_1.01,0.2659392118453979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,16,8,power_law_1.01,0.346233606338501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,16,8,power_law_1.01,0.42543997764587405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,16,8,power_law_1.01,0.632428789138794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,16,8,power_law_1.01,1.1966464042663574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,128,balanced,0.037802666425704956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,128,balanced,0.03863999992609024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,128,balanced,0.03697066754102707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,128,balanced,0.03869866579771042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,128,balanced,0.03775466730197271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,128,balanced,0.03797333439191183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,128,balanced,0.039493332306543984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,128,balanced,0.038165333370367684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,128,balanced,0.03985599925120672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,128,balanced,0.0383146678407987
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,128,balanced,0.03828266759713491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,128,balanced,0.04029333343108495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,128,balanced,0.04005333284536997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,128,balanced,0.04035733391841253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,128,balanced,0.042410666743914284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,128,balanced,0.04407466451327006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,128,balanced,0.04394666850566864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,128,balanced,0.04402133325735728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,128,balanced,0.046021332343419395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,128,balanced,0.048783997694651283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,128,balanced,0.050426666935284935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,128,balanced,0.05406400064627329
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,128,balanced,0.060266668597857155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,128,balanced,0.06851199766000111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,128,balanced,0.08006933331489563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,128,balanced,0.10195733110109965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,128,balanced,0.1218239963054657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,128,balanced,0.1423786679903666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,128,balanced,0.19934932390848795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,128,balanced,0.3577333291371663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,128,power_law_1.2,0.03372159898281098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,128,power_law_1.2,0.03158400058746338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,128,power_law_1.2,0.0320576012134552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,128,power_law_1.2,0.03159680068492889
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,128,power_law_1.2,0.0320576012134552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,128,power_law_1.2,0.032793599367141726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,128,power_law_1.2,0.034195199608802795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,128,power_law_1.2,0.033542400598526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,128,power_law_1.2,0.03423359990119934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,128,power_law_1.2,0.03436160087585449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,128,power_law_1.2,0.03457919955253601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,128,power_law_1.2,0.0344895988702774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,128,power_law_1.2,0.03492479920387268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,128,power_law_1.2,0.036108800768852235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,128,power_law_1.2,0.038899201154708865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,128,power_law_1.2,0.039315199851989745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,2,power_law_1.01,0.04631040096282959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,128,power_law_1.2,0.040064001083374025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,128,power_law_1.2,0.04348799884319306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,128,power_law_1.2,0.046291199326515195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,2,power_law_1.01,0.05241600275039673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,128,power_law_1.2,0.050367999076843264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,128,power_law_1.2,0.05694720149040222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,2,power_law_1.01,0.05807999968528747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,128,power_law_1.2,0.07022079825401306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,128,power_law_1.2,0.07935360074043274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,2,power_law_1.01,0.060153597593307497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,128,power_law_1.2,0.12902400493621827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,128,power_law_1.2,0.151910400390625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,2,power_law_1.01,0.0781823992729187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,128,power_law_1.2,0.22151679992675782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,2,power_law_1.01,0.0997376024723053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,4,balanced,0.03419200082619985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,128,power_law_1.2,0.28533120155334474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,4,balanced,0.033759998778502144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,4,balanced,0.03877866764863332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,4,balanced,0.05630399783452352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,4,balanced,0.0820853312810262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,4,balanced,0.08320533235867818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,4,balanced,0.08217599987983704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,4,balanced,0.08347200353940327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,4,balanced,0.08354666829109192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,4,balanced,0.08583999673525493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,4,balanced,0.08566932876904805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,4,balanced,0.08745599786440532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,2,power_law_1.01,0.11090559959411621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,4,balanced,0.09047466516494751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,4,balanced,0.09288000067075093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,4,balanced,0.09974400202433269
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,128,power_law_1.2,0.3543423891067505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,4,balanced,0.10022399822870891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,4,balanced,0.10411199927330017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,4,balanced,0.11522133151690166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,4,balanced,0.12185600399971008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,4,balanced,0.14659733573595682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,4,balanced,0.15827199816703796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,4,balanced,0.20476800203323364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,4,balanced,0.23924267292022705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,4,balanced,0.33179734150568646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,4,balanced,0.4095199902852376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,2,power_law_1.01,0.13213440179824829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,4,balanced,0.574399987856547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,4,balanced,0.744704008102417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,128,power_law_1.2,0.5633471965789795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,4,balanced,0.9079946676890055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,4,balanced,1.402517318725586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,2,power_law_1.01,0.13281919956207275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,4,balanced,2.6971305211385093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,2,power_law_1.01,0.1403264045715332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,128,power_law_1.2,1.106611156463623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,2,power_law_1.01,0.14426239728927612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,2,power_law_1.01,0.15493119955062867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,2,power_law_1.01,0.15740159749984742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,2,power_law_1.01,0.17080960273742676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,2,power_law_1.01,0.18572800159454345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,2,power_law_1.01,0.19884159564971923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,2,power_law_1.01,0.19567359685897828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,2,power_law_1.01,0.22554879188537597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,2,power_law_1.2,0.04240640103816986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,2,power_law_1.01,0.25696640014648436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,2,power_law_1.2,0.04119040071964264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,2,power_law_1.01,0.31754240989685056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,2,power_law_1.2,0.04782719910144806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,2,power_law_1.01,0.3716223955154419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,2,power_law_1.2,0.049779200553894044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,2,power_law_1.01,0.489024019241333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,2,power_law_1.2,0.05326719880104065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,2,power_law_1.2,0.07255679965019227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,2,power_law_1.01,0.5972991943359375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,2,power_law_1.2,0.07511680126190186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,2,power_law_1.01,0.8387392044067383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,2,power_law_1.2,0.077811199426651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,2,power_law_1.2,0.07628160119056701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,2,power_law_1.01,1.033676815032959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,2,power_law_1.2,0.07783039808273315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,2,power_law_1.2,0.08184959888458251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,2,power_law_1.01,1.446720027923584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,2,power_law_1.2,0.08271999955177307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,2,power_law_1.2,0.08572800159454345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,2,power_law_1.2,0.09031040072441102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,2,power_law_1.01,1.932851219177246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,2,power_law_1.2,0.09413120150566101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,2,power_law_1.2,0.09758719801902771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,2,power_law_1.2,0.10448000431060792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,2,power_law_1.01,2.3285823822021485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,2,power_law_1.2,0.11818879842758179
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,2,power_law_1.2,0.1301632046699524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,2,power_law_1.2,0.15630719661712647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,2,power_law_1.01,3.734022521972656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,2,power_law_1.2,0.19059840440750123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,2,power_law_1.2,0.2425663948059082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,2,power_law_1.2,0.2935935974121094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,2,power_law_1.2,0.4480703830718994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,2,power_law_1.2,0.5447360038757324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,2,power_law_1.01,7.027974700927734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,2,power_law_1.2,0.7820288181304932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,2,power_law_1.2,1.082688045501709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,2,power_law_1.2,1.2992575645446778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,2,power_law_1.2,2.008844757080078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,2,power_law_1.2,4.143705749511719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,4,power_law_1.01,0.08684800267219543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,4,power_law_1.01,0.15047039985656738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,4,power_law_1.01,0.07749760150909424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,4,power_law_1.01,0.10485759973526002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,4,power_law_1.01,0.13687679767608643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,128,power_law_1.2,0.06096000075340271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,4,power_law_1.01,0.19764480590820313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,128,power_law_1.2,0.059987199306488034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,4,power_law_1.01,0.20567679405212402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,128,power_law_1.2,0.05829120278358459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,4,power_law_1.01,0.21791999340057372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,128,power_law_1.2,0.05891839861869812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,4,power_law_1.01,0.2296191930770874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,128,power_law_1.2,0.05882880091667175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,4,power_law_1.01,0.22826240062713624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,128,power_law_1.2,0.060736000537872314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,4,power_law_1.01,0.23328640460968017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,128,power_law_1.2,0.06121600270271301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,128,power_law_1.2,0.06183680295944214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,4,power_law_1.01,0.23972480297088622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,128,power_law_1.2,0.061247998476028444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,4,power_law_1.01,0.24350080490112305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,128,power_law_1.2,0.061875200271606444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,4,power_law_1.01,0.2501120090484619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,128,power_law_1.2,0.06177279949188232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,4,power_law_1.01,0.2628479957580566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,128,power_law_1.2,0.0633791983127594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,4,power_law_1.01,0.27619199752807616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,128,power_law_1.2,0.06462079882621766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,128,power_law_1.2,0.06321280002593994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,4,power_law_1.01,0.28261120319366456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,128,power_law_1.2,0.07069439888000488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,4,power_law_1.01,0.3213696002960205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,128,power_law_1.2,0.07036160230636597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,4,power_law_1.01,0.3382528066635132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,128,power_law_1.2,0.07585920095443725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,128,power_law_1.2,0.08266239762306213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,4,power_law_1.01,0.40823678970336913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,128,power_law_1.2,0.08906239867210389
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,4,power_law_1.01,0.4537087917327881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,128,power_law_1.2,0.11207040548324584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,128,power_law_1.2,0.12985600233078004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,4,power_law_1.01,0.5751232147216797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,128,power_law_1.2,0.16522879600524903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,4,power_law_1.01,0.6350272178649903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,128,power_law_1.2,0.2187648057937622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,128,power_law_1.2,0.29114880561828616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,4,power_law_1.01,0.8216768264770508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,128,power_law_1.2,0.3692800045013428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,4,power_law_1.01,1.0869312286376953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,128,power_law_1.2,0.5373504161834717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,128,power_law_1.2,0.7542975902557373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,4,power_law_1.01,1.4774975776672363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,128,power_law_1.2,0.8802047729492187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,4,power_law_1.01,1.7493247985839844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,128,power_law_1.2,1.449452781677246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,2,4,power_law_1.01,2.2374975204467775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,128,power_law_1.2,2.9586496353149414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,2,4,power_law_1.01,3.2501758575439452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,2,4,power_law_1.01,6.511634826660156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,32,1,power_law_1.2,0.09500799775123596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,32,1,power_law_1.2,0.09815679788589478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,32,1,power_law_1.2,0.10609920024871826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,32,1,power_law_1.2,0.12333439588546753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,32,1,power_law_1.2,0.15352319478988646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,4,balanced,0.0461706668138504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,4,balanced,0.048767998814582825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,4,balanced,0.050250664353370667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,4,balanced,0.05605333546797434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,4,balanced,0.07674133280913036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,4,balanced,0.11183466513951619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,4,balanced,0.11157866319020589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,4,balanced,0.11206400394439697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,4,balanced,0.11383466919263203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,4,balanced,0.11573333541552226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,4,balanced,0.11623467008272807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,4,balanced,0.11808533469835918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,4,balanced,0.1181706686814626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,4,balanced,0.11952533324559529
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,4,balanced,0.12946666280428568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,4,balanced,0.130213330189387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,32,1,power_law_1.2,0.1821120023727417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,4,balanced,0.13897066315015158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,4,balanced,0.1497760017712911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,4,balanced,0.1579093337059021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,4,balanced,0.18341867129007974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,4,balanced,0.2023093303044637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,4,balanced,0.2629759907722473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,4,balanced,0.29843733708063763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,4,balanced,0.4145066738128662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,4,balanced,0.515717347462972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,4,balanced,0.7294133504231771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,4,balanced,0.9208213488260905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,4,balanced,1.1399253209431965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,4,balanced,1.7316479682922363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,32,1,power_law_1.2,0.20288000106811524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,4,balanced,3.2849013010660806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,32,1,power_law_1.2,0.24406399726867675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,32,1,power_law_1.2,0.25490560531616213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,32,1,power_law_1.2,0.2602751970291138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,32,1,power_law_1.2,0.2806976079940796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,32,1,power_law_1.2,0.2857151985168457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,32,1,power_law_1.2,0.2984256029129028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,32,2,balanced,0.047781333327293396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,32,1,power_law_1.2,0.3209023952484131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,32,2,balanced,0.051818668842315674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,32,2,balanced,0.05381333331267039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,32,2,balanced,0.060234665870666504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,32,2,balanced,0.0763733337322871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,32,2,balanced,0.11852799852689107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,32,2,balanced,0.1206719974676768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,32,2,balanced,0.11576533317565918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,32,2,balanced,0.11746133367220561
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,32,2,balanced,0.11808533469835918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,32,2,balanced,0.12060800194740295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,32,2,balanced,0.12274666627248128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,32,2,balanced,0.1256480018297831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,32,2,balanced,0.12963733077049255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,32,2,balanced,0.13780799508094788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,32,2,balanced,0.14218133687973022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,32,2,balanced,0.15599999825159708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,32,2,balanced,0.17774399121602377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,32,2,balanced,0.19305066267649332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,32,2,balanced,0.23609066009521484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,32,2,balanced,0.27271467447280884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,32,2,balanced,0.3962719837824504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,32,2,balanced,0.45896001656850177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,32,1,power_law_1.2,0.3243135929107666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,32,2,balanced,0.6770666440327963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,32,2,balanced,0.8501066366831461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,32,2,balanced,1.2454453309377034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,32,2,balanced,1.6236640612284343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,32,2,balanced,2.0047839482625327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,32,1,power_law_1.2,0.3342720031738281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,32,2,balanced,3.1379626592000327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,32,2,balanced,6.159082412719727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,32,1,power_law_1.2,0.3587775945663452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,32,1,power_law_1.2,0.41413121223449706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,32,1,power_law_1.2,0.4543168067932129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,32,1,power_law_1.2,0.5553855895996094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,32,1,power_law_1.2,0.6443327903747559
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,32,1,power_law_1.2,0.835654354095459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,32,1,power_law_1.2,1.0085503578186035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,64,balanced,0.07962133487065633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,64,balanced,0.08341866731643677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,64,balanced,0.07944533228874207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,64,balanced,0.07913066446781158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,64,balanced,0.094458669424057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,64,balanced,0.12590932846069336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,64,balanced,0.12006933490435283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,64,balanced,0.12017599741617839
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,64,balanced,0.12956800063451132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,64,balanced,0.13395200173060098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,64,balanced,0.1267519990603129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,64,balanced,0.12618666887283325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,64,balanced,0.13613866766293845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,64,balanced,0.12853866815567017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,64,balanced,0.13726933797200522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,64,balanced,0.14328533411026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,64,balanced,0.1488746702671051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,64,balanced,0.1368160049120585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,64,balanced,0.14311466614405313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,64,balanced,0.14659200112024942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,32,1,power_law_1.2,1.380185604095459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,64,balanced,0.15493333339691162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,64,balanced,0.1752799948056539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,64,balanced,0.19599467515945435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,64,balanced,0.2584639986356099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,64,balanced,0.29331199328104657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,64,balanced,0.3999573389689128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,64,balanced,0.49906134605407715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,64,balanced,0.6021866798400879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,64,balanced,0.8961439927419027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,64,balanced,1.7436587015787761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,32,1,power_law_1.2,1.7376575469970703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,32,1,power_law_1.2,2.4575103759765624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,32,1,power_law_1.2,3.189958381652832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,32,1,power_law_1.2,3.9045951843261717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,32,1,power_law_1.2,6.184896087646484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,16,power_law_1.01,0.07879679799079894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,16,power_law_1.01,0.09212160110473633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,16,power_law_1.01,0.06313599944114685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,32,1,power_law_1.2,11.857984161376953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,16,power_law_1.01,0.06739839911460876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,16,power_law_1.01,0.07544959783554077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,16,power_law_1.01,0.08664320111274719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,16,power_law_1.01,0.08828160166740417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,16,power_law_1.01,0.09319679737091065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,16,power_law_1.01,0.09388160109519958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,16,power_law_1.01,0.09159680008888245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,16,power_law_1.01,0.09587839841842652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,16,2,power_law_1.2,0.033344000577926636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,16,2,power_law_1.2,0.03276160061359405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,16,2,power_law_1.2,0.03187839984893799
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,16,power_law_1.01,0.09751679897308349
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,16,2,power_law_1.2,0.03391999900341034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,16,2,power_law_1.2,0.03404799997806549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,16,2,power_law_1.2,0.03386240005493164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,16,2,power_law_1.2,0.03535360097885132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,16,2,power_law_1.2,0.036320000886917114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,16,power_law_1.01,0.10055680274963379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,16,2,power_law_1.2,0.037254399061203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,16,2,power_law_1.2,0.03780480027198792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,16,2,power_law_1.2,0.03858560025691986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,16,2,power_law_1.2,0.040294399857521056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,16,power_law_1.01,0.10419199466705323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,16,2,power_law_1.2,0.04095999896526337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,16,2,power_law_1.2,0.042982399463653564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,16,2,power_law_1.2,0.0494271993637085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,16,2,power_law_1.2,0.05041279792785645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,16,2,power_law_1.2,0.05848960280418396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,16,power_law_1.01,0.11141760349273681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,16,2,power_law_1.2,0.06865919828414917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,16,2,power_law_1.2,0.07667840123176575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,16,2,power_law_1.2,0.09413759708404541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,16,2,power_law_1.2,0.12774399518966675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,16,2,power_law_1.2,0.1515071988105774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,16,power_law_1.01,0.1145792007446289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,16,2,power_law_1.2,0.1756608009338379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,16,2,power_law_1.2,0.2598207950592041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,16,2,power_law_1.2,0.3179903984069824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,16,2,power_law_1.2,0.4961728096008301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,16,power_law_1.01,0.12136960029602051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,16,2,power_law_1.2,0.6691008090972901
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,16,2,power_law_1.2,0.8419839859008789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,16,power_law_1.01,0.13886719942092896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,16,2,power_law_1.2,1.2645440101623535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,16,2,power_law_1.2,2.203539276123047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,16,power_law_1.01,0.15808639526367188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,16,power_law_1.01,0.19297280311584472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,16,power_law_1.01,0.21938560009002686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,16,power_law_1.01,0.29354879856109617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,16,power_law_1.01,0.32385280132293703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,16,power_law_1.01,0.4377151966094971
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,16,power_law_1.01,0.5306560039520264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,32,power_law_1.2,0.09283199906349182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,16,power_law_1.01,0.7485439777374268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,32,power_law_1.2,0.11141760349273681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,16,power_law_1.01,0.9433728218078613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,32,power_law_1.2,0.10582400560379028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,2,16,power_law_1.01,1.175603199005127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,32,power_law_1.2,0.06965759992599488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,32,power_law_1.2,0.0773311972618103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,2,16,power_law_1.01,1.6909696578979492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,32,power_law_1.2,0.08273280262947083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,32,power_law_1.2,0.08737919926643371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,32,power_law_1.2,0.09154559969902039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,32,power_law_1.2,0.09265279769897461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,2,16,power_law_1.01,3.6033985137939455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,32,power_law_1.2,0.10098559856414795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,32,power_law_1.2,0.09809920191764832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,32,power_law_1.2,0.1085055947303772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,32,power_law_1.2,0.10003199577331542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,32,power_law_1.2,0.1081279993057251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,32,power_law_1.2,0.11817599534988403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,32,power_law_1.2,0.12095359563827515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,32,power_law_1.2,0.1266495943069458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,8,power_law_1.01,0.08188160061836243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,32,power_law_1.2,0.14063359498977662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,8,power_law_1.01,0.129094398021698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,32,power_law_1.2,0.15457279682159425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,8,power_law_1.01,0.07055360078811646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,32,power_law_1.2,0.17851519584655762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,8,power_law_1.01,0.08221439719200134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,32,power_law_1.2,0.2081984043121338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,8,power_law_1.01,0.09225599765777588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,32,power_law_1.2,0.2703999996185303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,8,power_law_1.01,0.11555839776992798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,32,power_law_1.2,0.315283203125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,8,power_law_1.01,0.11940480470657348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,32,power_law_1.2,0.45551362037658694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,8,power_law_1.01,0.12160639762878418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,32,power_law_1.2,0.5784383773803711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,8,power_law_1.01,0.12152960300445556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,8,power_law_1.01,0.12943999767303466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,32,power_law_1.2,0.7776959896087646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,8,power_law_1.01,0.13354239463806153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,32,power_law_1.2,1.0776512145996093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,8,power_law_1.01,0.13686399459838866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,2,power_law_1.01,0.048665601015090945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,8,power_law_1.01,0.13631999492645264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,2,power_law_1.01,0.07127040028572082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,32,power_law_1.2,1.3231743812561034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,2,power_law_1.01,0.06327040195465088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,8,power_law_1.01,0.14431359767913818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,2,power_law_1.01,0.08691200017929077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,2,power_law_1.01,0.11909760236740112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,32,power_law_1.2,2.077996826171875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,8,power_law_1.01,0.1522304058074951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,2,power_law_1.01,0.15091840028762818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,2,power_law_1.01,0.1905727982521057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,8,power_law_1.01,0.16083840131759644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,2,power_law_1.01,0.1910272002220154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,2,power_law_1.01,0.19879039525985717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,8,power_law_1.01,0.16656639575958251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,2,power_law_1.01,0.21298561096191407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,32,power_law_1.2,4.3081214904785154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,2,power_law_1.01,0.21546878814697265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,8,power_law_1.01,0.19337600469589233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,2,power_law_1.01,0.22734720706939698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,2,power_law_1.01,0.23558399677276612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,8,power_law_1.01,0.21916160583496094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,2,power_law_1.01,0.23897600173950195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,2,power_law_1.01,0.25991039276123046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,8,power_law_1.01,0.27692160606384275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,2,power_law_1.01,0.2669055938720703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,2,power_law_1.01,0.27473280429840086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,8,power_law_1.01,0.3082304000854492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,2,power_law_1.01,0.31025280952453616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,2,power_law_1.01,0.3309695959091187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,8,power_law_1.01,0.41565442085266113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,2,power_law_1.01,0.3805567979812622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,2,power_law_1.01,0.45041918754577637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,8,power_law_1.01,0.4527103900909424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,2,power_law_1.01,0.544159984588623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,2,power_law_1.01,0.6535168170928956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,8,power_law_1.01,0.6490623950958252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,1,1,power_law_1.01,0.137990403175354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,2,power_law_1.01,0.843289566040039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,1,1,power_law_1.01,0.18193279504776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,1,1,power_law_1.01,0.2065727949142456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,1,1,power_law_1.01,0.42725119590759275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,2,power_law_1.01,1.0806143760681153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,1,1,power_law_1.01,0.4311552047729492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,8,power_law_1.01,0.7445312023162842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,1,1,power_law_1.01,0.43671040534973143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,1,1,power_law_1.01,0.4517632007598877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,1,1,power_law_1.01,0.45204482078552244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,2,power_law_1.01,1.4104319572448731
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,1,1,power_law_1.01,0.4620543956756592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,1,1,power_law_1.01,0.4689216136932373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,1,1,power_law_1.01,0.4646143913269043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,1,1,power_law_1.01,0.5007487773895264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,2,power_law_1.01,1.7689664840698243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,8,power_law_1.01,1.0699775695800782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,1,1,power_law_1.01,0.4902336120605469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,1,1,power_law_1.01,0.4912703990936279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,1,1,power_law_1.01,0.5273215770721436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,1,1,power_law_1.01,0.5341504096984864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,1,1,power_law_1.01,0.6057663917541504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,2,power_law_1.01,2.167980766296387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,1,1,power_law_1.01,0.6788352012634278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,1,1,power_law_1.01,0.7555071830749511
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,1,1,power_law_1.01,0.8995200157165527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,8,power_law_1.01,1.4070719718933105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,1,1,power_law_1.01,1.0649791717529298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,1,1,power_law_1.01,1.4438207626342774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,2,power_law_1.01,3.2437633514404296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,1,1,power_law_1.01,1.7063295364379882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,1,1,power_law_1.01,2.4118783950805662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,1,1,power_law_1.01,3.1152767181396483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,2,8,power_law_1.01,1.6388416290283203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,1,1,power_law_1.01,4.52476806640625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,2,power_law_1.01,6.410578918457031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,1,1,power_law_1.01,5.923948669433594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,2,8,power_law_1.01,2.597433662414551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,1,1,power_law_1.01,7.398834991455078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,1,1,power_law_1.01,11.770572662353516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,2,8,power_law_1.01,5.335257720947266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,1,1,power_law_1.01,23.444313049316406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,32,8,balanced,0.03621333340803782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,32,8,balanced,0.03573333223660787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,32,8,balanced,0.03355200091997782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,32,8,balanced,0.033887999753157295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,32,8,balanced,0.03344533344109853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,32,8,balanced,0.03344533344109853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,32,8,balanced,0.035445332527160645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,32,8,balanced,0.03367999941110611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,32,8,balanced,0.03571200122435888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,32,8,balanced,0.03393599887688955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,32,8,balanced,0.035536001125971474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,32,8,balanced,0.03460799902677536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,32,8,balanced,0.0354666660229365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,32,8,balanced,0.03545066714286804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,32,8,balanced,0.039861333866914116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,32,8,balanced,0.039861333866914116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,32,8,balanced,0.04161600023508072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,32,8,balanced,0.043824002146720886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,32,8,balanced,0.04382933179537455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,32,8,balanced,0.04808533191680908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,32,8,balanced,0.050069332122802734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,32,8,balanced,0.062352001667022705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,32,8,balanced,0.0681333343187968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,32,8,balanced,0.08880000313123067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,32,8,balanced,0.11147200067838033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,32,8,balanced,0.14803199966748556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,32,8,balanced,0.16895999511082968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,32,8,balanced,0.2053119937578837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,32,8,balanced,0.297925333182017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,32,8,balanced,0.5478293498357137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,64,power_law_1.01,0.06113280057907104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,32,power_law_1.2,0.04951040148735046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,64,power_law_1.01,0.059462398290634155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,2,4,power_law_1.01,0.0643392026424408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,2,4,power_law_1.01,0.057011198997497556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,2,4,power_law_1.01,0.05960320234298706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,2,4,power_law_1.01,0.0651968002319336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,64,power_law_1.01,0.055731201171875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,2,4,power_law_1.01,0.06607360243797303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,2,4,power_law_1.01,0.06635519862174988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,2,4,power_law_1.01,0.06692479848861695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,32,power_law_1.2,0.05030400156974792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,2,4,power_law_1.01,0.06744959950447083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,64,power_law_1.01,0.05432320237159729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,2,4,power_law_1.01,0.06766080260276794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,2,4,power_law_1.01,0.06922240257263183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,2,4,power_law_1.01,0.0706496000289917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,2,4,power_law_1.01,0.07390080094337463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,2,4,power_law_1.01,0.07530239820480347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,64,power_law_1.01,0.05669119954109192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,16,power_law_1.2,0.04108160138130188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,2,4,power_law_1.01,0.08065919876098633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,2,4,power_law_1.01,0.08762239813804626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,2,4,power_law_1.01,0.09326720237731934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,16,power_law_1.2,0.03273600041866302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,2,4,power_law_1.01,0.10328320264816285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,32,power_law_1.2,0.050195199251174924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,64,power_law_1.01,0.05753600001335144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,2,4,power_law_1.01,0.11619199514389038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,2,4,power_law_1.01,0.13838720321655273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,16,power_law_1.2,0.030067199468612672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,2,4,power_law_1.01,0.19407999515533447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,2,4,power_law_1.01,0.19886080026626587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,16,power_law_1.2,0.030079999566078187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,64,power_law_1.01,0.05923839807510376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,2,4,power_law_1.01,0.2886847972869873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,2,4,power_law_1.01,0.3936511993408203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,16,power_law_1.2,0.030777600407600404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,2,4,power_law_1.01,0.6402048110961914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,64,power_law_1.01,0.06058880090713501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,16,power_law_1.2,0.03194240033626557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,32,power_law_1.2,0.0455808013677597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,2,4,power_law_1.01,0.689024019241333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,16,power_law_1.2,0.032518398761749265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,2,4,power_law_1.01,0.926534366607666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,64,power_law_1.01,0.06087039709091187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,16,power_law_1.2,0.032953599095344545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,2,4,power_law_1.01,1.2482175827026367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,16,power_law_1.2,0.03361279964447021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,64,power_law_1.01,0.06156799793243408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,2,4,power_law_1.01,1.6236095428466797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,32,power_law_1.2,0.0474047988653183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,16,power_law_1.2,0.0353408008813858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,64,power_law_1.01,0.0606719970703125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,16,power_law_1.2,0.03700479865074158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,2,4,power_law_1.01,2.8394048690795897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,16,power_law_1.2,0.03783040046691895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,64,power_law_1.01,0.06247680187225342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,16,power_law_1.2,0.03845759928226471
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,32,power_law_1.2,0.04991360008716583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,16,power_law_1.2,0.039103999733924866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,2,4,power_law_1.01,4.411078262329101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,64,power_law_1.01,0.06222079992294312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,16,power_law_1.2,0.04481280148029328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,64,power_law_1.01,0.0643455982208252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,16,power_law_1.2,0.045977601408958436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,32,power_law_1.2,0.050329601764678954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,16,power_law_1.2,0.04799999892711639
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,64,power_law_1.01,0.06912639737129211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,16,power_law_1.2,0.05234559774398804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,16,power_law_1.2,0.055225598812103274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,64,power_law_1.01,0.07068799734115601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,16,power_law_1.2,0.06754559874534607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,32,power_law_1.2,0.05056639909744263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,64,power_law_1.01,0.07429760098457336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,16,power_law_1.2,0.08329600095748901
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,16,power_law_1.2,0.10046080350875855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,64,power_law_1.01,0.08472959995269776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,16,power_law_1.2,0.13211519718170167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,32,power_law_1.2,0.05314559936523437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,64,power_law_1.01,0.09550719857215881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,16,power_law_1.2,0.16155519485473632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,16,power_law_1.2,0.24175360202789306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,64,power_law_1.01,0.12023040056228637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,32,power_law_1.2,0.05397760272026062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,16,power_law_1.2,0.3792639970779419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,64,power_law_1.01,0.1416640043258667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,16,power_law_1.2,0.46511359214782716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,64,power_law_1.01,0.18137600421905517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,32,power_law_1.2,0.05553280115127564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,64,power_law_1.01,0.22283520698547363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,16,power_law_1.2,0.6317503929138184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,64,power_law_1.01,0.30609281063079835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,32,power_law_1.2,0.05738880038261414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,16,power_law_1.2,0.839628791809082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,64,power_law_1.01,0.38331520557403564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,2,power_law_1.01,0.06498559713363647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,32,power_law_1.2,0.05795199871063232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,2,power_law_1.01,0.07678719758987426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,64,power_law_1.01,0.5589951992034912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,2,power_law_1.01,0.08382080197334289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,16,power_law_1.2,2.0300224304199217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,2,power_law_1.01,0.10339839458465576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,32,power_law_1.2,0.06316159963607788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,64,power_law_1.01,0.7521024227142334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,2,power_law_1.01,0.1331264019012451
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,2,power_law_1.01,0.18794239759445192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,32,power_law_1.2,0.06917759776115417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,2,power_law_1.01,0.19708800315856934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,64,power_law_1.01,0.8886336326599121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,2,power_law_1.01,0.19733760356903077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,2,power_law_1.01,0.20631680488586426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,32,power_law_1.2,0.07167360186576843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,2,power_law_1.01,0.20380160808563233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,64,power_law_1.01,1.420684814453125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,2,power_law_1.01,0.21722240447998048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,2,power_law_1.01,0.2244096040725708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,32,power_law_1.2,0.07904000282287597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,2,power_law_1.01,0.22520320415496825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,2,power_law_1.01,0.23484160900115966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,32,power_law_1.2,0.09633920192718506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,2,power_law_1.01,0.2516160011291504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,64,power_law_1.01,2.8954240798950197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,2,power_law_1.01,0.2559231996536255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,32,power_law_1.2,0.1062656044960022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,2,power_law_1.01,0.2668159961700439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,2,power_law_1.01,0.3047744035720825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,2,power_law_1.01,0.3356800079345703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,32,power_law_1.2,0.13831679821014403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,2,power_law_1.01,0.41383681297302244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,2,power_law_1.01,0.4443391799926758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,32,power_law_1.2,0.15594880580902098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,2,power_law_1.01,0.5769472122192383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,4,power_law_1.01,0.046291199326515195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,2,power_law_1.01,0.6939712047576905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,4,power_law_1.01,0.05668479800224304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,32,power_law_1.2,0.20872960090637208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,2,power_law_1.01,0.9409152030944824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,4,power_law_1.01,0.05056639909744263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,2,power_law_1.01,1.1571711540222167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,32,power_law_1.2,0.25671679973602296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,4,power_law_1.01,0.05735679864883423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,2,power_law_1.01,1.5861311912536622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,4,power_law_1.01,0.06728960275650024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,32,power_law_1.2,0.3558207988739014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,4,power_law_1.01,0.08269439935684204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,2,power_law_1.01,1.9775680541992187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,4,power_law_1.01,0.09269760251045227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,16,1,balanced,0.0371573343873024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,16,1,balanced,0.037920000652472176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,16,1,balanced,0.0397173340121905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,16,1,balanced,0.03942933430274328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,16,1,balanced,0.04160533348719279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,16,1,balanced,0.039733332892258964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,16,1,balanced,0.039733332892258964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,4,power_law_1.01,0.09887359738349914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,16,1,balanced,0.04152533411979675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,16,1,balanced,0.04195733368396759
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,16,1,balanced,0.043712000052134194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,32,power_law_1.2,0.45555839538574217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,16,1,balanced,0.04182399809360504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,2,power_law_1.01,2.3803264617919924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,16,1,balanced,0.04393066465854645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,16,1,balanced,0.04593066871166229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,16,1,balanced,0.04785066843032837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,16,1,balanced,0.05399466554323832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,16,1,balanced,0.05378133555253347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,16,1,balanced,0.06011733412742615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,16,1,balanced,0.07064533233642578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,16,1,balanced,0.08058133224646251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,16,1,balanced,0.10476799805959065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,16,1,balanced,0.11808000008265178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,16,1,balanced,0.15025066335995993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,16,1,balanced,0.1782026688257853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,4,power_law_1.01,0.0986624002456665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,16,1,balanced,0.2539253234863281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,16,1,balanced,0.32131733496983844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,16,1,balanced,0.46378131707509357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,16,1,balanced,0.6030400196711222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,16,1,balanced,0.7427999973297119
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,16,1,balanced,1.1439839998881023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,4,power_law_1.01,0.10254720449447632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,16,1,balanced,2.2327253023783364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,32,power_law_1.2,0.6862080097198486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,2,power_law_1.01,3.7249919891357424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,4,power_law_1.01,0.10710400342941284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,4,power_law_1.01,0.11278079748153687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,4,power_law_1.01,0.11649279594421387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,32,power_law_1.2,0.8555775642395019
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,4,power_law_1.01,0.12595839500427247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,2,power_law_1.01,7.136326599121094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,4,power_law_1.01,0.13949439525604249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,32,power_law_1.2,1.0194175720214844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,4,power_law_1.01,0.1428096055984497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,4,power_law_1.01,0.14753279685974122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,4,power_law_1.01,0.16524800062179565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,32,power_law_1.2,1.6318527221679688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,4,power_law_1.01,0.1852671980857849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,4,power_law_1.01,0.2198784112930298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,4,power_law_1.01,0.25199360847473146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,4,power_law_1.01,0.34662399291992185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,4,power_law_1.01,0.40900478363037107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,32,power_law_1.2,3.40211181640625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,32,4,power_law_1.2,0.03614720106124878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,32,4,power_law_1.2,0.03427839875221252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,32,4,power_law_1.2,0.03297280073165894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,32,4,power_law_1.2,0.03331199884414673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,32,4,power_law_1.2,0.033529600501060484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,32,4,power_law_1.2,0.03406080007553101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,4,power_law_1.01,0.5595712184906005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,32,4,power_law_1.2,0.03459199965000152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,32,4,power_law_1.2,0.03546879887580871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,32,4,power_law_1.2,0.03585279881954193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,32,4,power_law_1.2,0.03591679930686951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,32,4,power_law_1.2,0.03617919981479645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,32,4,power_law_1.2,0.037011200189590455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,32,4,power_law_1.2,0.03733119964599609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,4,power_law_1.01,0.7520832061767578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,32,4,power_law_1.2,0.038848000764846805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,32,4,power_law_1.2,0.04445439875125885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,32,4,power_law_1.2,0.04594559967517853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,32,4,power_law_1.2,0.04815999865531921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,32,4,power_law_1.2,0.05258880257606506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,32,4,power_law_1.2,0.057222402095794676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,32,4,power_law_1.2,0.08103039860725403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,4,power_law_1.01,0.9919551849365235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,32,4,power_law_1.2,0.08188160061836243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,4,balanced,0.031583999594052635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,4,balanced,0.029616000751654308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,4,balanced,0.031301334500312805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,32,4,power_law_1.2,0.11228159666061402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,4,balanced,0.0335413341720899
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,4,balanced,0.04905066887537638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,4,balanced,0.05745066702365875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,4,balanced,0.05784533421198527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,4,balanced,0.05689600110054016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,4,balanced,0.05801600217819214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,32,4,power_law_1.2,0.13786879777908326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,4,balanced,0.0565280020236969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,4,balanced,0.058101331194241844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,4,balanced,0.05673066775004069
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,4,balanced,0.057909334699312844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,4,balanced,0.06018666426340739
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,32,4,power_law_1.2,0.18426239490509033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,4,balanced,0.06445333361625671
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,4,balanced,0.06423999865849812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,4,balanced,0.0652159998814265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,4,balanced,0.07645333309968312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,4,balanced,0.07023466626803081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,4,balanced,0.081194669008255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,32,4,power_law_1.2,0.19544320106506347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,4,balanced,0.0846453309059143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,4,balanced,0.10523200035095215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,4,balanced,0.1200213332970937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,4,balanced,0.15944000085194907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,4,balanced,0.16901334126790366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,32,4,power_law_1.2,0.38800640106201173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,4,balanced,0.24039467175801596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,4,power_law_1.01,1.3728511810302735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,4,balanced,0.2822773257891337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,4,4,balanced,0.3429653247197469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,4,4,balanced,0.5053066809972128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,32,4,power_law_1.2,0.43097600936889646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,4,4,balanced,0.94486403465271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,32,4,power_law_1.2,0.4701888084411621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,32,4,power_law_1.2,0.8863295555114746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,4,power_law_1.01,1.6306880950927733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,32,4,power_law_1.2,1.75598087310791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,4,power_law_1.01,2.4729984283447264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,16,balanced,0.03879466652870178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,16,balanced,0.03773866593837738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,16,balanced,0.03842133283615112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,16,balanced,0.03878399978081385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,16,balanced,0.038586666186650596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,16,balanced,0.04224533339341482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,16,balanced,0.041946664452552795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,16,balanced,0.04109866668780645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,16,balanced,0.0420959989229838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,16,balanced,0.04043733328580856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,16,balanced,0.04393066465854645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,16,balanced,0.04333333174387614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,16,balanced,0.04246933261553446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,16,balanced,0.04435733457406362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,16,balanced,0.04771733283996582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,16,balanced,0.04854399959246317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,16,balanced,0.048698668678601585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,16,balanced,0.05243200063705444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,16,balanced,0.05726933479309082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,16,balanced,0.06070399781068166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,16,balanced,0.06674666702747345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,16,balanced,0.09053333600362141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,4,power_law_1.01,4.8504383087158205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,16,balanced,0.10160533587137859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,16,balanced,0.12662399808565775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,16,balanced,0.1560533344745636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,16,balanced,0.20503999789555868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,16,balanced,0.2472426692644755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,16,balanced,0.2972426613171895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,16,balanced,0.44440531730651855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,16,balanced,0.8301546573638916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,8,power_law_1.2,0.03857280015945434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,8,power_law_1.2,0.03315840065479279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,8,power_law_1.2,0.03294720053672791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,8,power_law_1.2,0.03370240032672882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,4,64,power_law_1.01,0.023123200237751006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,8,power_law_1.2,0.03431040048599243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,8,power_law_1.2,0.03763200044631958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,4,64,power_law_1.01,0.02306559979915619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,8,power_law_1.2,0.03816959857940674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,4,64,power_law_1.01,0.02338559925556183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,8,power_law_1.2,0.03798399865627289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,4,64,power_law_1.01,0.02375040054321289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,8,power_law_1.2,0.03883520066738129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,4,64,power_law_1.01,0.023660799860954283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,8,power_law_1.2,0.039468801021575926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,4,64,power_law_1.01,0.02510719895362854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,8,power_law_1.2,0.040243199467658995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,4,64,power_law_1.01,0.02507520020008087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,4,64,power_law_1.01,0.023443199694156647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,8,power_law_1.2,0.04126079976558685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,16,balanced,0.05232533315817515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,16,balanced,0.04896000027656555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,16,balanced,0.04990933338801066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,16,balanced,0.051829333106676735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,4,64,power_law_1.01,0.027699199318885804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,16,balanced,0.05406400064627329
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,16,balanced,0.05650666852792104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,8,power_law_1.2,0.042419201135635375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,16,balanced,0.058789332707722984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,16,balanced,0.058335999647776283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,16,balanced,0.0580320010582606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,16,balanced,0.06018133461475372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,16,balanced,0.05810666580994924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,16,balanced,0.06063466767470042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,16,balanced,0.04597333570321401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,16,balanced,0.06029333174228668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,4,64,power_law_1.01,0.02784000039100647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,16,balanced,0.062421331803003945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,16,balanced,0.048623998959859215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,8,power_law_1.2,0.043756800889968875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,16,balanced,0.06625600159168243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,16,balanced,0.06688533226648967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,16,balanced,0.047466665506362915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,16,balanced,0.07020799815654755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,16,balanced,0.07523199915885925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,16,balanced,0.05061866839726766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,16,balanced,0.08337066570917766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,16,balanced,0.054133335749308266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,16,balanced,0.11011733611424764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,16,balanced,0.1129866639773051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,4,64,power_law_1.01,0.029344001412391664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,16,balanced,0.07601066430409749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,16,balanced,0.14193066954612732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,8,power_law_1.2,0.049542400240898135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,16,balanced,0.08259200056393941
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,16,balanced,0.1625493367513021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,16,balanced,0.0796800007422765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,16,balanced,0.2184213399887085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,16,balanced,0.08070399860541026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,16,balanced,0.26109333833058673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,16,balanced,0.08117866516113281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,16,balanced,0.3534506559371948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,4,64,power_law_1.01,0.030118399858474733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,16,balanced,0.08164266745249431
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,8,power_law_1.2,0.04848639965057373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,16,balanced,0.4382079839706421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,16,balanced,0.0819893330335617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,16,balanced,0.0860693355401357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,16,balanced,0.5182079871495565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,16,balanced,0.0860586663087209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,16,balanced,0.09092266360918681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,4,64,power_law_1.01,0.029951998591423036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,16,balanced,0.7991147041320801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,16,balanced,0.09429333607355754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,8,power_law_1.2,0.05326719880104065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,16,balanced,0.09784533580144246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,16,balanced,0.10821333527565002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,16,balanced,1.5436479250590007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,16,balanced,0.11646933356920879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,4,64,power_law_1.01,0.029977598786354066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,16,balanced,0.13502933581670126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,8,power_law_1.2,0.06058880090713501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,16,balanced,0.16088533401489258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,16,balanced,0.1983413298924764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,16,balanced,0.22719999154408774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,16,balanced,0.3051039973894755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,4,64,power_law_1.01,0.03224959969520569
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,16,balanced,0.3755199909210205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,8,power_law_1.2,0.06854400038719177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,16,balanced,0.5324480136235555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,16,balanced,0.6821653048197428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,4,64,power_law_1.01,0.03304319977760315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,8,power_law_1.2,0.08410239815711976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,16,balanced,0.8315306504567465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,16,balanced,1.2678133646647136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,4,64,power_law_1.01,0.03289600014686585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,8,power_law_1.2,0.08361600041389465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,16,balanced,2.4631519317626953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,4,64,power_law_1.01,0.03439359962940216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,8,power_law_1.2,0.12049280405044556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,64,power_law_1.01,0.03585920035839081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,8,power_law_1.2,0.138099205493927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,64,power_law_1.01,0.04115839898586273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,8,power_law_1.2,0.21395199298858641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,64,power_law_1.01,0.045900800824165346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,64,power_law_1.01,0.0497408002614975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,8,power_law_1.2,0.24708480834960939
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,64,power_law_1.01,0.059935998916625974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,8,power_law_1.2,0.3596735954284668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,64,power_law_1.01,0.07087360024452209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,8,power_law_1.2,0.5179647922515869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,64,power_law_1.01,0.08291199803352356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,64,power_law_1.01,0.10842880010604858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,8,power_law_1.2,0.6615359783172607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,64,power_law_1.01,0.1458624005317688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,8,power_law_1.2,1.1974143981933594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,64,power_law_1.01,0.17562240362167358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,64,power_law_1.01,0.24508159160614013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,1,balanced,0.09218133489290874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,1,balanced,0.09939199686050415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,1,balanced,0.11199466387430827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,1,balanced,0.14622933665911356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,1,balanced,0.20858667294184366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,1,balanced,0.3195679982503255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,1,balanced,0.43085332711537677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,1,balanced,0.42960532506306964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,1,balanced,0.43301331996917725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,1,balanced,0.4345279932022095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,1,balanced,0.670805295308431
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,8,power_law_1.2,2.6685951232910154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,1,balanced,0.6824533144632975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,1,balanced,0.6755946477254232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,1,balanced,0.678383986155192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,1,balanced,0.6891360282897949
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,1,balanced,0.6922399997711182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,1,balanced,0.7031520207722982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,1,balanced,0.7455146312713623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,1,balanced,0.7826826572418213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,64,power_law_1.01,0.499129581451416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,1,balanced,0.8584480285644531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,1,balanced,0.9229013125101725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,1,balanced,1.0727946758270264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,8,power_law_1.01,0.03619840145111084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,1,balanced,1.2190346717834473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,1,balanced,1.5139733950297039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,8,power_law_1.01,0.03555839955806732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,1,balanced,2.2325013478597007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,1,balanced,2.8052053451538086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,8,power_law_1.01,0.030995199084281923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,1,balanced,3.7935358683268228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,1,balanced,4.6280161539713545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,8,power_law_1.01,0.03110400140285492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,8,power_law_1.01,0.031839999556541446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,1,balanced,6.846895853678386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,8,power_law_1.01,0.03500159978866577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,8,power_law_1.01,0.034585601091384886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,8,power_law_1.01,0.035078400373458864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,1,balanced,12.828629811604818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,8,power_law_1.01,0.03460479974746704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,8,power_law_1.01,0.035436800122261046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,8,power_law_1.01,0.03760640025138855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,8,power_law_1.01,0.03937920033931732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,8,power_law_1.01,0.04080640077590943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,8,power_law_1.01,0.04110719859600067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,8,power_law_1.01,0.04450559914112091
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,4,8,balanced,0.025301332275072735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,4,8,balanced,0.022976001103719074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,4,8,balanced,0.02333866556485494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,4,8,balanced,0.02533866713444392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,4,8,balanced,0.025306666890780132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,4,8,balanced,0.02733866622050603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,4,8,balanced,0.027669332921504974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,4,8,balanced,0.029338667790095013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,8,power_law_1.01,0.04625279903411865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,4,8,balanced,0.027301333844661713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,4,8,balanced,0.027349332968393963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,4,8,balanced,0.029482667644818623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,4,8,balanced,0.02977599948644638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,4,8,balanced,0.02956266701221466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,4,8,balanced,0.033413333197434746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,4,8,balanced,0.03551999976237615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,4,8,balanced,0.035642666121323906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,4,8,balanced,0.03926933308442434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,4,8,balanced,0.039333333571751915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,8,balanced,0.04574933151404063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,8,balanced,0.047584002216657005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,8,balanced,0.05199466645717621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,8,balanced,0.06016000111897787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,8,balanced,0.07031466563542683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,8,power_law_1.01,0.04832639992237091
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,8,balanced,0.09611200292905171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,8,balanced,0.11752000451087952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,8,balanced,0.15743466218312582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,8,balanced,0.18277867635091147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,8,balanced,0.22644799947738647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,2,balanced,0.1520906686782837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,8,balanced,0.3262079954147339
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,2,balanced,0.19366933902104697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,8,balanced,0.6058346827824911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,2,balanced,0.3286293347676595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,8,power_law_1.01,0.054201602935791016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,2,balanced,0.6074453194936117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,2,balanced,1.1223680178324382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,2,balanced,1.6127039591471355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,8,power_law_1.01,0.06113280057907104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,2,balanced,1.6199520428975422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,2,balanced,1.6203625996907551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,8,power_law_1.01,0.06871039867401123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,2,balanced,1.6211679776509602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,2,balanced,1.6257279713948567
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,2,balanced,1.6399626731872559
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,8,power_law_1.01,0.07976959943771363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,2,balanced,1.644213358561198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,2,balanced,1.6528426806132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,8,power_law_1.01,0.10710400342941284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,2,balanced,1.662234624226888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,2,balanced,1.6814239819844563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,2,balanced,1.6901493072509766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,8,power_law_1.01,0.13613439798355104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,2,balanced,1.7156799634297688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,2,balanced,1.902517318725586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,8,power_law_1.01,0.1843135952949524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,2,balanced,1.8088746070861816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,2,balanced,2.35042667388916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,2,balanced,1.9873654047648113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,8,power_law_1.01,0.2231679916381836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,2,balanced,2.9106667836507163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,2,balanced,2.338618596394857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,8,power_law_1.01,0.3319744110107422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,2,balanced,3.532848040262858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,2,balanced,3.4645118713378906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,1,power_law_1.2,0.24241280555725098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,8,power_law_1.01,0.4544896125793457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,2,balanced,5.798432032267253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,8,power_law_1.01,0.5337215900421143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,1,power_law_1.2,0.2178368091583252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,2,balanced,5.403535842895508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,8,power_law_1.01,0.8028927803039551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,1,2,balanced,7.228623708089192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,1,power_law_1.2,0.41797761917114257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,128,power_law_1.01,0.08775039911270141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,1,2,balanced,10.0469601949056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,1,power_law_1.2,0.700108814239502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,128,power_law_1.01,0.08313599824905396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,8,power_law_1.01,1.658745574951172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,1,power_law_1.2,0.9676671981811523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,128,power_law_1.01,0.08477439880371093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,1,2,balanced,19.219248453776043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,128,power_law_1.01,0.07482879757881164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,1,power_law_1.2,2.16760311126709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,128,power_law_1.01,0.07379199862480164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,1,power_law_1.2,2.3644287109375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,128,power_law_1.01,0.07868160009384155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,128,power_law_1.01,0.08328959941864014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,1,power_law_1.2,2.5336896896362306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,128,power_law_1.01,0.08453119993209839
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,128,power_law_1.01,0.08335360288619995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,1,power_law_1.2,2.6322240829467773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,128,power_law_1.01,0.08191360235214233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,1,power_law_1.2,2.6533376693725588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,128,power_law_1.01,0.08576639890670776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,128,power_law_1.01,0.08817920088768005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,1,power_law_1.2,2.7847616195678713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,128,power_law_1.01,0.08562560081481933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,2,power_law_1.2,0.04085760116577149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,128,power_law_1.01,0.0893887996673584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,1,power_law_1.2,2.917068862915039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,2,power_law_1.2,0.04431360065937042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,2,power_law_1.2,0.05048959851264954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,2,power_law_1.2,0.05699840188026428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,128,power_law_1.01,0.09719679951667785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,2,power_law_1.2,0.06835839748382569
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,2,power_law_1.2,0.09593600034713745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,1,power_law_1.2,3.0540544509887697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,128,power_law_1.01,0.09426559805870056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,2,power_law_1.2,0.10021120309829712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,2,power_law_1.2,0.10278400182723998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,128,power_law_1.01,0.09979519844055176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,2,power_law_1.2,0.10501760244369507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,2,power_law_1.2,0.10759680271148682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,1,power_law_1.2,3.1507200241088866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,128,power_law_1.01,0.10882560014724732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,2,power_law_1.2,0.10823040008544922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,2,power_law_1.2,0.1155519962310791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,2,power_law_1.2,0.11919360160827637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,128,power_law_1.01,0.1189695954322815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,2,power_law_1.2,0.12078720331192017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,1,power_law_1.2,3.244287872314453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,2,power_law_1.2,0.1276927947998047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,128,power_law_1.01,0.1500864028930664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,2,power_law_1.2,0.13304320573806763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,2,power_law_1.2,0.13956480026245116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,128,power_law_1.01,0.17505279779434205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,2,power_law_1.2,0.15363199710845948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,1,power_law_1.2,3.3284225463867188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,2,power_law_1.2,0.18125439882278443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,128,power_law_1.01,0.22724480628967286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,2,power_law_1.2,0.21892480850219725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,2,power_law_1.2,0.24844160079956054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,128,power_law_1.01,0.2748863935470581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,1,power_law_1.2,3.4496318817138674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,2,power_law_1.2,0.3276927947998047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,2,power_law_1.2,0.39883520603179934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,128,power_law_1.01,0.37966079711914064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,2,power_law_1.2,0.5780608177185058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,1,power_law_1.2,3.841139221191406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,2,power_law_1.2,0.6677375793457031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,128,power_law_1.01,0.4759232044219971
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,2,power_law_1.2,1.1347647666931153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,128,power_law_1.01,0.6871935844421386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,1,power_law_1.2,3.936671829223633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,2,power_law_1.2,1.278553581237793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,2,power_law_1.2,1.8053760528564453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,128,power_law_1.01,0.9080448150634766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,1,power_law_1.2,4.426009750366211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,2,power_law_1.2,2.444051170349121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,128,power_law_1.01,1.049177646636963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,1,power_law_1.2,4.660249710083008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,128,power_law_1.01,1.6881471633911134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,2,power_law_1.2,5.138950347900391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,1,power_law_1.2,5.597241592407227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,128,power_law_1.01,3.392985534667969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,1,power_law_1.2,5.5230976104736325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,1,power_law_1.2,6.600895690917969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,1,power_law_1.2,8.784690856933594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,32,balanced,0.05709866682688395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,32,balanced,0.0588319996992747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,32,balanced,0.05843733251094818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,32,balanced,0.0621013343334198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,32,balanced,0.06289066871007283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,32,balanced,0.07110400001207988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,32,balanced,0.0801333338022232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,32,balanced,0.08019733428955078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,32,balanced,0.07930666704972585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,32,balanced,0.08077333370844524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,32,balanced,0.07835199932257335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,32,balanced,0.08063466846942902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,32,balanced,0.08079466720422109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,32,balanced,0.08185066779454549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,32,balanced,0.09121066331863403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,32,balanced,0.08941866954167683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,32,balanced,0.09618666768074036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,32,balanced,0.10412800312042236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,32,balanced,0.11126400033632915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,32,balanced,0.12989866733551025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,32,balanced,0.14642133315404257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,32,balanced,0.17922665675481161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,32,balanced,0.20686399936676025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,32,balanced,0.26874667406082153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,32,balanced,0.3479199806849162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,32,balanced,0.4713493188222249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,32,balanced,0.6055039962132772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,32,balanced,0.7424639860788981
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,32,balanced,1.1329066753387451
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,8,16,power_law_1.2,0.03856639862060547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,32,balanced,2.2130026817321777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,8,16,power_law_1.2,0.0358271986246109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,1,power_law_1.2,11.265644836425782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,8,16,power_law_1.2,0.03377279937267304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,8,16,power_law_1.2,0.03383679986000061
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,8,16,power_law_1.2,0.036396801471710205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,8,16,power_law_1.2,0.03749760091304779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,8,16,power_law_1.2,0.038124799728393555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,8,16,power_law_1.2,0.03847039937973022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,8,16,power_law_1.2,0.03935999870300293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,8,16,power_law_1.2,0.03938559889793396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,8,16,power_law_1.2,0.040908798575401306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,8,16,power_law_1.2,0.041222399473190306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,1,power_law_1.2,12.745945739746094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,8,16,power_law_1.2,0.04410879909992218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,8,16,power_law_1.2,0.043424001336097716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,8,16,power_law_1.2,0.04862079918384552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,8,16,power_law_1.2,0.05134720206260681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,8,16,power_law_1.2,0.05335680246353149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,8,16,power_law_1.2,0.057657599449157715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,8,16,power_law_1.2,0.06454399824142457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,8,16,power_law_1.2,0.07332479953765869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,8,16,power_law_1.2,0.08675199747085571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,8,16,power_law_1.2,0.11403520107269287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,1,balanced,0.11477333307266235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,1,balanced,0.17458132902781168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,1,balanced,0.2046133279800415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,8,16,power_law_1.2,0.1538432002067566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,1,balanced,0.3257066607475281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,1,balanced,0.5628639856974283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,1,1,power_law_1.2,15.452076721191407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,1,balanced,1.040122667948405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,1,balanced,1.5243679682413738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,8,16,power_law_1.2,0.18812160491943358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,1,balanced,1.5258240699768066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,1,balanced,1.5283466974894206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,8,16,power_law_1.2,0.2579008102416992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,1,balanced,1.533631960550944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,1,balanced,1.538010597229004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,1,balanced,1.5443785985310872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,8,16,power_law_1.2,0.4225791931152344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,1,balanced,1.551146666208903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,1,balanced,1.5667519569396973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,1,balanced,1.5798293749491374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,1,balanced,1.5903894106547039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,8,16,power_law_1.2,0.5280255794525146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,1,balanced,1.6148640314737956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,1,balanced,1.6638026237487793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,1,balanced,1.7051199277242024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,1,balanced,1.7912906010945637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,8,16,power_law_1.2,0.8174528121948242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,1,balanced,1.8732105890909831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,1,balanced,2.07315731048584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,1,balanced,2.2570080757141113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,1,balanced,2.6112160682678223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,8,16,power_law_1.2,1.2311039924621583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,1,balanced,3.2205867767333984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,1,balanced,3.9428160985310874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,1,balanced,5.364976247151692
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,1,1,power_law_1.2,21.68732147216797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,8,16,power_law_1.2,2.1590911865234377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,1,balanced,6.831626892089844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,1,balanced,9.560202916463217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,1,balanced,17.65838368733724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,8,balanced,0.05271466573079427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,8,balanced,0.07100800176461537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,8,balanced,0.09611200292905171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,8,balanced,0.14696000019709268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,8,balanced,0.24259734153747559
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,8,balanced,0.43857598304748535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,8,balanced,0.6364159981409708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,8,balanced,0.636410673459371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,8,balanced,0.6350880066553751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,8,balanced,0.6395039955774943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,8,balanced,0.6389013528823853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,8,balanced,0.6398986577987671
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,8,balanced,0.6437013149261475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,8,balanced,0.6409013271331787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,8,balanced,0.6521173318227133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,8,balanced,0.653711994489034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,8,balanced,0.6610560019810995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,8,balanced,0.6739359696706136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,8,balanced,0.6819946765899658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,8,balanced,0.7048426469167074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,1,1,power_law_1.2,39.542022705078125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,8,balanced,0.7277119954427084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,8,balanced,0.7709120114644369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,8,balanced,0.8248746395111084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,8,balanced,0.8907626469930013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,8,balanced,1.0244320233662922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,8,balanced,1.1659519672393799
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,8,balanced,1.5141654014587402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,8,balanced,1.8909974098205566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,8,balanced,2.426640033721924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,16,2,power_law_1.01,0.03116160035133362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,16,2,power_law_1.01,0.031430399417877196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,16,2,power_law_1.01,0.031123200058937074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,16,2,power_law_1.01,0.03221119940280914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,8,balanced,4.668314615885417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,16,2,power_law_1.01,0.03292160034179688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,16,2,power_law_1.01,0.032678401470184325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,16,2,power_law_1.01,0.0353983998298645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,16,2,power_law_1.01,0.03521279990673065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,16,2,power_law_1.01,0.04100480079650879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,16,2,power_law_1.01,0.04266240000724793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,16,2,power_law_1.01,0.040249601006507874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,16,2,power_law_1.01,0.04704639911651611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,16,2,power_law_1.01,0.047884801030159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,16,2,power_law_1.01,0.04391680061817169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,16,2,power_law_1.01,0.05607680082321167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,16,2,power_law_1.01,0.06044160127639771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,16,2,power_law_1.01,0.05912960171699524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,16,2,power_law_1.01,0.06859520077705383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,16,2,power_law_1.01,0.08133760094642639
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,16,2,power_law_1.01,0.08877440094947815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,16,2,power_law_1.01,0.10910719633102417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,16,2,power_law_1.01,0.15542399883270264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,16,2,power_law_1.01,0.16427520513534546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,16,2,power_law_1.01,0.2218240022659302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,16,2,power_law_1.01,0.28487040996551516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,16,2,power_law_1.01,0.3940927982330322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,16,2,power_law_1.01,0.48689918518066405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,16,2,power_law_1.01,0.6596928119659424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,16,2,power_law_1.01,0.8994943618774414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,16,2,power_law_1.01,1.8338239669799805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,16,power_law_1.01,0.08204159736633301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,16,power_law_1.01,0.1257151961326599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,8,power_law_1.2,0.032339200377464294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,8,power_law_1.2,0.03107840120792389
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,8,power_law_1.2,0.028262400627136232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,16,power_law_1.01,0.05739520192146301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,8,power_law_1.2,0.029574400186538695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,8,power_law_1.2,0.03054719865322113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,8,power_law_1.2,0.03519999980926514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,16,power_law_1.01,0.07004799842834472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,8,power_law_1.2,0.035571199655532834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,8,power_law_1.2,0.03437440097332001
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,8,power_law_1.2,0.03447679877281189
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,16,power_law_1.01,0.0753279983997345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,8,power_law_1.2,0.03493120074272156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,8,power_law_1.2,0.037561601400375365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,8,power_law_1.2,0.03803519904613495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,16,power_law_1.01,0.09085440039634704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,8,2,power_law_1.01,0.03699840009212494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,8,2,power_law_1.01,0.04164479970932007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,8,power_law_1.2,0.03895680010318756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,8,2,power_law_1.01,0.03758080005645752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,8,2,power_law_1.01,0.04183039963245392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,8,2,power_law_1.01,0.04168959856033325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,8,power_law_1.2,0.043782401084899905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,8,2,power_law_1.01,0.042393600940704344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,8,2,power_law_1.01,0.04214400053024292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,8,2,power_law_1.01,0.0440064013004303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,8,power_law_1.2,0.04659200012683869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,8,2,power_law_1.01,0.04506239891052246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,16,power_law_1.01,0.09406080245971679
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,8,2,power_law_1.01,0.049292799830436704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,8,2,power_law_1.01,0.048281601071357726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,8,power_law_1.2,0.04936319887638092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,8,2,power_law_1.01,0.05235840082168579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,8,2,power_law_1.01,0.0542464017868042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,8,2,power_law_1.01,0.057651197910308837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,8,power_law_1.2,0.05151360034942627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,8,2,power_law_1.01,0.0645695984363556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,8,2,power_law_1.01,0.07057279944419861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,8,2,power_law_1.01,0.08171520233154297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,8,power_law_1.2,0.05857279896736145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,16,power_law_1.01,0.09927039742469787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,8,2,power_law_1.01,0.09119359850883484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,8,2,power_law_1.01,0.10178560018539429
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,8,power_law_1.2,0.05686399936676025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,8,2,power_law_1.01,0.11610239744186401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,8,2,power_law_1.01,0.14375679492950438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,8,2,power_law_1.01,0.16958080530166625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,8,power_law_1.2,0.06995199918746949
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,8,2,power_law_1.01,0.2125823974609375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,8,2,power_law_1.01,0.2905407905578613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,8,power_law_1.2,0.07427840232849121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,16,power_law_1.01,0.09742720127105713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,8,2,power_law_1.01,0.3820672035217285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,8,power_law_1.2,0.08885120153427124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,8,2,power_law_1.01,0.5202623844146729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,8,2,power_law_1.01,0.6643648147583008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,8,power_law_1.2,0.10682239532470703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,8,2,power_law_1.01,0.8997759819030762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,16,power_law_1.01,0.09850879907608032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,8,power_law_1.2,0.13341439962387086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,8,2,power_law_1.01,1.230777645111084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,8,power_law_1.2,0.17439359426498413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,16,power_law_1.01,0.10127359628677368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,8,2,power_law_1.01,2.5618751525878904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,8,power_law_1.2,0.24753921031951903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,8,power_law_1.2,0.32798080444335936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,16,power_law_1.01,0.10173439979553223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,4,8,power_law_1.2,0.3624576091766357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,16,power_law_1.01,0.10712319612503052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,4,8,power_law_1.2,0.6060031890869141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,16,power_law_1.01,0.10849920511245728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,4,8,power_law_1.2,0.9617471694946289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,16,power_law_1.01,0.1137727975845337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,16,power_law_1.01,0.11648000478744507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,16,power_law_1.01,0.12099200487136841
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,16,power_law_1.01,0.139136004447937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,8,power_law_1.01,0.035385599732398985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,16,power_law_1.01,0.15029759407043458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,8,power_law_1.01,0.03988479971885681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,8,power_law_1.01,0.03219839930534363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,8,power_law_1.01,0.03553920090198517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,8,power_law_1.01,0.03911679983139038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,16,power_law_1.01,0.18280960321426393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,8,power_law_1.01,0.05264639854431152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,8,power_law_1.01,0.05370879769325256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,8,power_law_1.01,0.05331839919090271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,16,power_law_1.01,0.21521279811859131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,8,power_law_1.01,0.05133439898490906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,8,power_law_1.01,0.054201602935791016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,8,power_law_1.01,0.052851200103759766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,2,power_law_1.01,0.04644480049610138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,16,power_law_1.01,0.2711679935455322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,8,power_law_1.01,0.05427200198173523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,2,power_law_1.01,0.05248640179634094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,8,power_law_1.01,0.055232000350952146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,2,power_law_1.01,0.051551997661590576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,8,power_law_1.01,0.05688959956169128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,2,power_law_1.01,0.06562560200691223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,8,power_law_1.01,0.06319360136985779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,16,power_law_1.01,0.27973759174346924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,2,power_law_1.01,0.08106880187988282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,8,power_law_1.01,0.06212480068206787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,2,power_law_1.01,0.10825599431991577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,8,power_law_1.01,0.06412799954414368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,8,power_law_1.01,0.06929919719696045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,2,power_law_1.01,0.11251840591430665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,8,power_law_1.01,0.07032960057258605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,2,power_law_1.01,0.11667840480804444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,16,power_law_1.01,0.36515839099884034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,8,power_law_1.01,0.07925119996070862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,2,power_law_1.01,0.1174015998840332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,8,power_law_1.01,0.08601599931716919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,2,power_law_1.01,0.12103040218353271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,8,power_law_1.01,0.10167039632797241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,2,power_law_1.01,0.12739839553833007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,8,power_law_1.01,0.12064640522003174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,2,power_law_1.01,0.12952959537506104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,16,power_law_1.01,0.46359682083129883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,8,power_law_1.01,0.15266560316085814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,2,power_law_1.01,0.13446400165557862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,2,power_law_1.01,0.13895679712295533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,8,power_law_1.01,0.1909824013710022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,2,power_law_1.01,0.14914560317993164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,8,power_law_1.01,0.25116159915924074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,2,power_law_1.01,0.15251840353012086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,16,power_law_1.01,0.6880576133728027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,8,power_law_1.01,0.2743936061859131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,2,power_law_1.01,0.1646783947944641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,2,power_law_1.01,0.19035520553588867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,2,8,power_law_1.01,0.3557760000228882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,2,power_law_1.01,0.20638079643249513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,2,power_law_1.01,0.2622848033905029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,2,8,power_law_1.01,0.48380160331726074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,16,power_law_1.01,0.7774208068847657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,2,power_law_1.01,0.3233344078063965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,2,power_law_1.01,0.38435840606689453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,2,8,power_law_1.01,0.9915072441101074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,2,power_law_1.01,0.500383996963501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,2,power_law_1.01,0.6779520034790039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,4,16,power_law_1.01,0.900057601928711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,2,power_law_1.01,0.9110591888427735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,2,power_law_1.01,1.2038335800170898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,2,power_law_1.01,1.470137596130371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,2,power_law_1.01,1.8045312881469726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,4,16,power_law_1.01,1.43571195602417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,2,power_law_1.01,3.0040639877319335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,2,power_law_1.01,5.564012908935547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,8,power_law_1.01,0.04166400134563446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,8,power_law_1.01,0.04101119935512543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,8,power_law_1.01,0.03433600068092346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,4,16,power_law_1.01,2.8683584213256834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,8,power_law_1.01,0.03925119936466217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,8,power_law_1.01,0.04043520092964172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,8,power_law_1.01,0.045363199710845944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,8,power_law_1.01,0.04478079974651337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,8,power_law_1.01,0.04664320051670075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,8,power_law_1.01,0.04757120013237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,8,power_law_1.01,0.04833920001983642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,8,power_law_1.01,0.04716159999370575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,8,power_law_1.01,0.04904960095882416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,8,power_law_1.01,0.04922240078449249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,8,power_law_1.01,0.05304960012435913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,8,power_law_1.01,0.05813120007514953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,8,power_law_1.01,0.0586624026298523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,8,power_law_1.01,0.06388480067253113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,8,power_law_1.01,0.07109119892120361
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,8,power_law_1.01,0.07585920095443725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,8,power_law_1.01,0.08664320111274719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,8,power_law_1.01,0.10094720125198364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,8,power_law_1.01,0.12447359561920165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,8,power_law_1.01,0.13975679874420166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,16,power_law_1.01,0.048563200235366824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,8,power_law_1.01,0.1915071964263916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,16,power_law_1.01,0.054604798555374146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,8,8,balanced,0.03718400001525879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,8,8,balanced,0.03808533400297165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,8,power_law_1.01,0.2285759925842285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,8,8,balanced,0.036517334481080375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,8,8,balanced,0.040037333965301514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,8,8,balanced,0.04645333190759023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,8,8,balanced,0.0444106658299764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,8,8,balanced,0.04566933214664459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,16,power_law_1.01,0.04254080057144165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,8,8,balanced,0.04587733248869578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,8,8,balanced,0.04567466676235199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,8,8,balanced,0.04580266773700714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,8,8,balanced,0.04578666885693868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,8,8,balanced,0.04779199759165446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,8,8,balanced,0.04806933303674062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,8,8,balanced,0.04773333172003428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,8,8,balanced,0.0518453319867452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,8,8,balanced,0.05329066514968872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,8,8,balanced,0.0563679983218511
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,8,8,balanced,0.060191998879114784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,8,power_law_1.01,0.31247360706329347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,8,8,balanced,0.06622933348019917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,16,power_law_1.01,0.04359680116176605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,8,8,balanced,0.07242133220036824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,8,8,balanced,0.08504533767700195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,8,8,balanced,0.10951466361681621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,8,8,balanced,0.127920001745224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,8,8,balanced,0.16761600971221924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,8,8,balanced,0.20218666394551596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,8,8,balanced,0.2850506703058879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,16,power_law_1.01,0.04475519955158234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,8,8,balanced,0.3561386664708455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,8,8,balanced,0.4360426664352417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,8,power_law_1.01,0.4226816177368164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,8,8,balanced,0.6670400301615397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,8,8,balanced,1.2722506523132324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,16,power_law_1.01,0.04750080108642578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,16,power_law_1.01,0.048691201210021975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,2688,1856,6,128,1,8,power_law_1.01,0.4924799919128418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,16,power_law_1.01,0.04948480129241943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,16,power_law_1.01,0.05004159808158874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,2688,1856,6,128,1,8,power_law_1.01,0.6757184028625488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,16,power_law_1.01,0.0505407989025116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,16,power_law_1.01,0.05310720205307007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,16,power_law_1.01,0.054092800617218016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,2688,1856,6,128,1,8,power_law_1.01,1.4269824028015137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,16,power_law_1.01,0.05666559934616089
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,16,power_law_1.01,0.055769598484039305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,16,power_law_1.01,0.06173440217971802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,16,power_law_1.01,0.063155198097229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,16,power_law_1.01,0.06773759722709656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,16,power_law_1.01,0.07571200132369996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,16,power_law_1.01,0.09068800210952759
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,16,power_law_1.01,0.11230720281600952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,16,power_law_1.01,0.14073599576950074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,16,power_law_1.01,0.18535679578781128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,16,power_law_1.01,0.21857280731201173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,16,power_law_1.01,0.2977920055389404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,2,power_law_1.2,0.03704319894313812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,16,power_law_1.01,0.4089536190032959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,2,power_law_1.2,0.03816959857940674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,2,power_law_1.2,0.03949440121650696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,16,power_law_1.01,0.5795584201812745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,2,power_law_1.2,0.043673598766326906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,2,power_law_1.2,0.05030400156974792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,2,power_law_1.2,0.07288320064544677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,16,power_law_1.01,0.7319680213928222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,2,power_law_1.2,0.07601919770240784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,2,power_law_1.2,0.07793279886245727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,16,power_law_1.01,0.8962688446044922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,256,power_law_1.01,0.056953597068786624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,2,power_law_1.2,0.08153600096702576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,2,power_law_1.2,0.08483840227127075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,256,power_law_1.01,0.057126402854919434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,2,power_law_1.2,0.08542720079421998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,16,power_law_1.01,1.3854335784912108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,2,power_law_1.2,0.08648959994316101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,256,power_law_1.01,0.05687680244445801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,2,power_law_1.2,0.09065600037574768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,2,power_law_1.2,0.09492480158805847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,256,power_law_1.01,0.057625597715377806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,2,power_law_1.2,0.10405759811401367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,2,power_law_1.2,0.1078336000442505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,256,power_law_1.01,0.05726079940795899
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,2,power_law_1.2,0.10883840322494506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,16,power_law_1.01,2.656070327758789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,256,power_law_1.01,0.05928959846496582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,2,power_law_1.2,0.12629120349884032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,2,power_law_1.2,0.1407807946205139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,256,power_law_1.01,0.05850239992141724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,2,power_law_1.2,0.1672320008277893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,2,power_law_1.2,0.19292160272598266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,256,power_law_1.01,0.05984640121459961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,2,power_law_1.2,0.2645695924758911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,256,power_law_1.01,0.060083198547363284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,2,power_law_1.2,0.30135040283203124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,256,power_law_1.01,0.06094719767570496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,2,power_law_1.2,0.4622655868530273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,256,power_law_1.01,0.060499197244644164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,2,power_law_1.2,0.584281587600708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,256,power_law_1.01,0.06213759779930115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,2,power_law_1.2,0.8091839790344239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,256,power_law_1.01,0.06199679970741272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,2,power_law_1.2,1.0602175712585449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,256,power_law_1.01,0.06328319907188415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,2,power_law_1.2,1.3957632064819336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,256,power_law_1.01,0.06800640225410462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,256,power_law_1.01,0.06773759722709656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,2,power_law_1.2,2.125574493408203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,256,power_law_1.01,0.07110400199890136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,256,power_law_1.01,0.078847998380661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,256,power_law_1.01,0.08326399922370911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,1,power_law_1.01,0.04837760031223297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,1,power_law_1.01,0.07306879758834839
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,256,power_law_1.01,0.10391680002212525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,2,power_law_1.2,3.827372741699219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,1,power_law_1.01,0.08255360126495362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,256,power_law_1.01,0.11960320472717285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,1,power_law_1.01,0.11938560009002686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,1,power_law_1.01,0.15210880041122438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,256,power_law_1.01,0.15530879497528077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,1,power_law_1.01,0.24594559669494628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,1,power_law_1.01,0.26762239933013915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,256,power_law_1.01,0.18379520177841185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,1,power_law_1.01,0.2741055965423584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,256,power_law_1.01,0.24591360092163086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,1,power_law_1.01,0.2803967952728271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,1,power_law_1.01,0.2861824035644531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,256,power_law_1.01,0.3081471920013428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,1,power_law_1.01,0.2984639883041382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,1,power_law_1.01,0.3056576013565063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,256,power_law_1.01,0.4416959762573242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,1,power_law_1.01,0.3135040044784546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,1,power_law_1.01,0.32376320362091066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,256,power_law_1.01,0.5609792232513428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,1,power_law_1.01,0.3361792087554932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,1,power_law_1.01,0.34532480239868163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,1,power_law_1.01,0.36854400634765627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,256,power_law_1.01,0.7071296215057373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,1,power_law_1.01,0.426035213470459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,1,power_law_1.01,0.45114879608154296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,1,power_law_1.01,0.5387008190155029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,256,power_law_1.01,1.1358400344848634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,8,2,power_law_1.01,0.049670401215553286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,8,2,power_law_1.01,0.06537600159645081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,8,2,power_law_1.01,0.06082559823989868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,1,power_law_1.01,0.6448768138885498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,8,2,power_law_1.01,0.07259520292282104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,8,2,power_law_1.01,0.07415680289268493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,8,2,power_law_1.01,0.07482879757881164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,8,2,power_law_1.01,0.07802240252494812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,8,2,power_law_1.01,0.07966079711914062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,1,power_law_1.01,0.8101696014404297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,8,2,power_law_1.01,0.08054400086402894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,8,2,power_law_1.01,0.08016639947891235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,8,2,power_law_1.01,0.08387200236320495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,8,2,power_law_1.01,0.08395519852638245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,8,2,power_law_1.01,0.08209279775619507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,1,power_law_1.01,1.0444160461425782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,8,2,power_law_1.01,0.09491199851036072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,8,2,power_law_1.01,0.1017024040222168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,8,2,power_law_1.01,0.10315519571304321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,8,2,power_law_1.01,0.12922879457473754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,256,power_law_1.01,2.2049407958984375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,8,2,power_law_1.01,0.13287680149078368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,1,power_law_1.01,1.3360896110534668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,8,2,power_law_1.01,0.14142080545425414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,8,2,power_law_1.01,0.1807360053062439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,8,2,power_law_1.01,0.20212481021881104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,8,2,power_law_1.01,0.30960640907287595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,8,2,power_law_1.01,0.3239295959472656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,1,power_law_1.01,1.6757951736450196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,8,2,power_law_1.01,0.4187007904052734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,8,2,power_law_1.01,0.5766592025756836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,8,2,power_law_1.01,0.8744192123413086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,1,power_law_1.01,2.227481651306152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,8,2,power_law_1.01,1.1324735641479493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,8,2,power_law_1.01,1.3827327728271483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,8,2,power_law_1.01,1.7428800582885742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,1,power_law_1.01,2.7396352767944334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,8,2,power_law_1.01,4.230195236206055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,1,power_law_1.01,3.331135940551758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,1,power_law_1.01,5.306496047973633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,1,balanced,0.06891199946403503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,1,balanced,0.09270933270454407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,1,balanced,0.10930666327476501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,1,balanced,0.14646933476130167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,1,balanced,0.2217493255933126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,1,balanced,0.2642666697502136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,1,balanced,0.2715839942296346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,1,balanced,0.27421865860621136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,1,balanced,0.2774133284886678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,1,balanced,0.27770666281382245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,1,balanced,0.2832319935162862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,1,balanced,0.28810666004816693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,1,balanced,0.29098665714263916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,1,power_law_1.01,9.495308685302735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,1,balanced,0.29552000761032104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,1,balanced,0.3088533282279968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,1,balanced,0.3148319919904073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,1,balanced,0.32972800731658936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,1,balanced,0.3726079861323039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,1,balanced,0.40187732378641766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,1,balanced,0.4917440017064412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,1,balanced,0.5475039879480997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,1,balanced,0.7751573721567789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,1,balanced,0.8922932942708334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,1,balanced,1.2800906499226887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,1,balanced,1.6257012685139973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,1,balanced,2.475173314412435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,1,balanced,3.037007967631022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,1,balanced,3.6025654474894204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,1,balanced,5.65178108215332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,4,power_law_1.01,0.050361597537994386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,4,power_law_1.01,0.051641601324081424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,1,balanced,10.971268971761068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,4,power_law_1.01,0.07735679745674133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,4,power_law_1.01,0.05854079723358154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,4,power_law_1.01,0.07340800166130065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,4,power_law_1.01,0.08711680173873901
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,1,power_law_1.01,0.1516991972923279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,4,power_law_1.01,0.09887359738349914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,1,power_law_1.01,0.2215359926223755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,4,power_law_1.01,0.10989439487457275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,1,power_law_1.01,0.33934080600738525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,1,power_law_1.01,0.6107200145721435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,4,power_law_1.01,0.1146239995956421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,1,power_law_1.01,0.9212672233581543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,1,power_law_1.01,1.603980827331543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,4,power_law_1.01,0.11741440296173096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,1,power_law_1.01,1.71331844329834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,4,power_law_1.01,0.12023680210113526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,1,power_law_1.01,1.7709184646606446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,8,32,power_law_1.01,0.0387584000825882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,4,power_law_1.01,0.12703360319137574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,8,32,power_law_1.01,0.03760640025138855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,1,power_law_1.01,1.8319040298461915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,8,32,power_law_1.01,0.03375360071659088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,8,32,power_law_1.01,0.034329599142074584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,1,power_law_1.01,1.8969856262207032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,4,power_law_1.01,0.13285759687423707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,8,32,power_law_1.01,0.03447040021419525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,1,power_law_1.01,1.9270719528198241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,8,32,power_law_1.01,0.03598720133304596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,4,power_law_1.01,0.13931519985198976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,8,32,power_law_1.01,0.03628160059452057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,1,power_law_1.01,2.000268745422363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,8,32,power_law_1.01,0.036185601353645326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,4,power_law_1.01,0.15232640504837036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,8,32,power_law_1.01,0.036601600050926206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,1,power_law_1.01,2.056787109375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,8,32,power_law_1.01,0.03674240112304687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,8,32,power_law_1.01,0.03768959939479828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,4,power_law_1.01,0.15897599458694459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,1,power_law_1.01,2.0377792358398437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,8,32,power_law_1.01,0.03823359906673431
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,8,32,power_law_1.01,0.038848000764846805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,1,power_law_1.01,2.1308671951293947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,4,power_law_1.01,0.17487360239028932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,8,32,power_law_1.01,0.040659201145172116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,8,32,power_law_1.01,0.04373759925365448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,1,power_law_1.01,2.2047487258911134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,4,power_law_1.01,0.20279040336608886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,8,32,power_law_1.01,0.045049598813056944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,1,power_law_1.01,2.2311168670654298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,8,32,power_law_1.01,0.04705280065536499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,4,power_law_1.01,0.2060159921646118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,8,32,power_law_1.01,0.050400000810623166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,1,power_law_1.01,2.4204288482666017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,8,32,power_law_1.01,0.05464959740638733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,4,power_law_1.01,0.24828801155090333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,8,32,power_law_1.01,0.06007680296897888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,1,power_law_1.01,2.6275264739990236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,8,32,power_law_1.01,0.07032319903373718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,8,32,power_law_1.01,0.09057919979095459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,4,power_law_1.01,0.2909888029098511
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,1,power_law_1.01,3.038412857055664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,8,32,power_law_1.01,0.10207359790802002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,4,power_law_1.01,0.37149438858032224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,8,32,power_law_1.01,0.15292799472808838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,1,power_law_1.01,3.1423103332519533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,8,32,power_law_1.01,0.1863935947418213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,4,power_law_1.01,0.47447681427001953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,8,32,power_law_1.01,0.2508671998977661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,1,power_law_1.01,3.8316673278808593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,8,32,power_law_1.01,0.33245439529418946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,4,power_law_1.01,0.6531648159027099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,1,power_law_1.01,3.8499519348144533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,8,32,power_law_1.01,0.3439039945602417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,4,power_law_1.01,0.7729216098785401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,1,power_law_1.01,4.711846542358399
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,8,32,power_law_1.01,0.5921408176422119
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,4,power_law_1.01,1.149465560913086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,1,power_law_1.01,6.099276733398438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,8,32,power_law_1.01,1.2183168411254883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,4,power_law_1.01,1.4149696350097656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,1,power_law_1.01,7.818246459960937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,4,power_law_1.01,1.7363775253295899
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,1,power_law_1.01,9.717600250244141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,4,power_law_1.01,2.626367950439453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,1,power_law_1.01,12.055628967285156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,4,power_law_1.01,5.175251388549805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,1,balanced,0.08086933195590973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,1,balanced,0.11248532931009929
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,1,balanced,0.13170133034388223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,1,balanced,0.2008906602859497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,1,balanced,0.3351626793543498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,1,balanced,0.46614933013916016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,1,balanced,0.47229333718617755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,1,balanced,0.4719626506169637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,1,balanced,0.476911981900533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,1,balanced,0.475877324740092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,1,power_law_1.01,15.532730102539062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,1,balanced,0.4846080144246419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,1,balanced,0.4891306559244792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,1,balanced,0.49193068345387775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,1,balanced,0.5007839997609457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,1,balanced,0.5106773376464844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,1,balanced,0.5213919878005981
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,1,balanced,0.54093865553538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,1,balanced,0.6184213161468506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,1,balanced,0.6216853459676107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,1,balanced,0.8633120059967041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,1,balanced,0.8499893347422282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,1,balanced,1.3619893391927083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,1,balanced,1.3700373967488606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,1,balanced,2.046885331471761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,1,balanced,2.3092479705810547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,1,balanced,3.4487892786661782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,1,balanced,4.269237200419108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,4,1,balanced,5.435893376668294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,4,1,balanced,8.073925018310547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,1,power_law_1.01,28.658444213867188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,4,power_law_1.01,0.050988799333572386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,4,power_law_1.01,0.050323200225830075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,4,1,balanced,15.372394561767578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,4,power_law_1.01,0.05017600059509277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,4,power_law_1.01,0.05617280006408691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,4,power_law_1.01,0.06280320286750793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,4,power_law_1.01,0.07565439939498901
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,4,power_law_1.01,0.08127359747886657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,4,power_law_1.01,0.0840831995010376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,4,power_law_1.01,0.08568320274353028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,4,power_law_1.01,0.08629760146141052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,4,power_law_1.01,0.09039999842643738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,4,power_law_1.01,0.09191039800643921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,4,power_law_1.01,0.09416959881782531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,4,power_law_1.01,0.09523199796676636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,4,power_law_1.01,0.10404479503631592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,4,power_law_1.01,0.11105920076370239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,4,power_law_1.01,0.11251840591430665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,4,power_law_1.01,0.12977919578552247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,4,power_law_1.01,0.1415743947029114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,32,2,balanced,0.048570667703946434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,32,2,balanced,0.052501335740089417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,4,power_law_1.01,0.17262719869613646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,32,2,balanced,0.05643199880917867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,32,2,balanced,0.06419200201829274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,32,2,balanced,0.09205866853396098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,32,2,balanced,0.13662933309872946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,32,2,balanced,0.1776853402455648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,32,2,balanced,0.1691360076268514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,32,2,balanced,0.16873067617416382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,32,2,balanced,0.17009600003560385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,32,2,balanced,0.1705120007197062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,32,2,balanced,0.1765120029449463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,4,power_law_1.01,0.2018496036529541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,32,2,balanced,0.17731734116872153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,32,2,balanced,0.17988799015680948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,32,2,balanced,0.18992533286412558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,2,balanced,0.033728001018365227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,32,2,balanced,0.19261866807937622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,2,balanced,0.03525333354870478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,32,2,balanced,0.20298133293787637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,2,balanced,0.041690667470296226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,32,2,balanced,0.2246826688448588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,2,balanced,0.06062933305899302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,2,balanced,0.08644800384839375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,32,2,balanced,0.24031466245651245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,2,balanced,0.08739200234413147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,2,balanced,0.0899786651134491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,32,2,balanced,0.28218666712443036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,4,power_law_1.01,0.25806078910827634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,2,balanced,0.08902933200200398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,2,balanced,0.09273599584897359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,32,2,balanced,0.3187733292579651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,2,balanced,0.09471999605496724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,2,balanced,0.09426666299502055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,32,2,balanced,0.4031466643015544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,2,balanced,0.1013866662979126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,2,balanced,0.1011253297328949
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,32,2,balanced,0.5038986603418986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,2,balanced,0.10331199566523235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,2,balanced,0.11051199833552043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,2,balanced,0.11497066418329875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,32,2,balanced,0.6861333052317301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,2,balanced,0.11910399794578552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,2,balanced,0.1313973367214203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,2,balanced,0.1477226714293162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,32,2,balanced,0.8567787011464437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,4,power_law_1.01,0.3279551982879639
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,2,balanced,0.18383999665578207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,2,balanced,0.19835732380549112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,2,balanced,0.27499733368555707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,32,2,balanced,1.231829325358073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,2,balanced,0.3227519989013672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,2,balanced,0.4671573241551717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,32,2,balanced,1.6047946612040203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,2,balanced,0.5903253157933553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,2,balanced,0.8512319723765055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,4,power_law_1.01,0.421235179901123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,32,2,balanced,2.000885327657064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,2,balanced,1.1053226788838704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,2,balanced,1.3642293612162273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,32,2,balanced,3.0120693842569985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,2,balanced,2.0944320360819497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,4,power_law_1.01,0.5322048187255859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,2,balanced,4.071370760599772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,32,2,balanced,5.871072133382161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,4,power_law_1.01,0.7587647914886475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,4,power_law_1.01,1.0286975860595704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,4,power_law_1.01,1.229478359222412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,4,power_law_1.01,1.8828159332275392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,4,power_law_1.01,3.878803253173828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,8,balanced,0.03813866774241129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,8,balanced,0.03801066676775614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,8,balanced,0.03583999971548716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,32,8,balanced,0.0444106658299764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,8,balanced,0.039813332259655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,8,balanced,0.0452106644709905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,32,8,balanced,0.0439573327700297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,8,balanced,0.049125333627065025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,32,8,balanced,0.04572266836961111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,8,balanced,0.05011733373006185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,32,8,balanced,0.04773333172003428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,8,balanced,0.04775466521581014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,32,8,balanced,0.05189333359400431
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,8,balanced,0.0495306650797526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,32,8,balanced,0.060175999999046326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,8,balanced,0.050160000721613564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,32,8,balanced,0.06252266466617584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,8,balanced,0.048010667165120445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,32,8,balanced,0.0642133355140686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,8,balanced,0.05026133358478546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,32,8,balanced,0.0649599979321162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,8,balanced,0.051541333397229515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,32,8,balanced,0.06423466900984447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,8,balanced,0.05342933535575867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,32,8,balanced,0.062181333700815834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,8,balanced,0.057215998570124306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,32,8,balanced,0.06446399788061778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,8,balanced,0.05596800148487091
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,32,8,balanced,0.06668800115585327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,8,balanced,0.062394668658574425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,32,8,balanced,0.06884799897670746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,8,balanced,0.06281066437562306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,32,8,balanced,0.07690666615962982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,8,balanced,0.06865066786607106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,32,8,balanced,0.07472533484299977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,8,balanced,0.08268266419569652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,32,8,balanced,0.08265600105126698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,8,balanced,0.08528000116348267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,32,8,balanced,0.09289600451787312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,8,balanced,0.1060640017191569
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,32,8,balanced,0.09541333715120952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,8,balanced,0.1179253359635671
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,32,8,balanced,0.11347732941309611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,8,balanced,0.15437333782513937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,32,8,balanced,0.13797332843144736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,8,balanced,0.1757813294728597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,32,8,balanced,0.16481066743532816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,8,balanced,0.24033600091934204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,32,8,balanced,0.19508800903956094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,8,balanced,0.28436799844106037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,32,8,balanced,0.2695573369661967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,2688,1856,6,128,1,8,balanced,0.3490560054779053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,32,8,balanced,0.3202986717224121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,2688,1856,6,128,1,8,balanced,0.5182666778564453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,32,8,balanced,0.4526666800181071
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,32,8,balanced,0.5747573375701904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,2688,1856,6,128,1,8,balanced,0.9721759955088297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,32,8,balanced,0.7194506327311198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,32,8,balanced,1.0609973271687825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,32,8,balanced,2.0743680000305176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,4,balanced,0.04409599800904592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,4,balanced,0.05053866902987162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,4,balanced,0.0499946673711141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,4,balanced,0.052655999859174095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,4,balanced,0.06366933385531108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,4,balanced,0.0937493344148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,4,balanced,0.11608533064524333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,4,balanced,0.11143466830253601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,4,balanced,0.11117333173751831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,4,balanced,0.11061867078145345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,4,balanced,0.11343999703725179
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,4,balanced,0.11807466546694438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,4,balanced,0.11850666999816895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,4,balanced,0.12166399757067363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,4,balanced,0.12965866923332214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,4,balanced,0.13241066535313925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,4,balanced,0.1418880025545756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,4,balanced,0.16074666380882263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,4,balanced,0.17674134174982706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,4,balanced,0.20869332551956177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,4,balanced,0.2379093368848165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,4,balanced,0.30270399649937946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,4,balanced,0.3782026767730713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,4,balanced,0.5083786646525065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,4,balanced,0.6693173249562582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,4,balanced,0.9492053190867106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,4,balanced,1.2311413288116455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,4,balanced,1.5405866305033367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,4,balanced,2.327232042948405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,4,balanced,4.546597480773926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,4,power_law_1.2,0.14998400211334229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,4,power_law_1.2,0.10888960361480712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,4,power_law_1.2,0.23883519172668458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,4,power_law_1.2,0.1054527997970581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,4,power_law_1.2,0.2180351972579956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,4,power_law_1.2,0.20521600246429444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,4,power_law_1.2,0.09621760249137878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,4,power_law_1.2,0.33724799156188967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,4,power_law_1.2,0.13279999494552613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,4,power_law_1.2,0.469868803024292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,4,balanced,0.04576000074545542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,4,balanced,0.05527999997138977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,4,power_law_1.2,0.16444159746170045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,4,balanced,0.07382399837176006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,4,balanced,0.1086293359597524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,4,balanced,0.15602133671442667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,4,balanced,0.18030399084091187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,4,balanced,0.1809706687927246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,4,balanced,0.18338666359583536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,4,balanced,0.18440000216166177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,4,balanced,0.18330132961273193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,4,balanced,0.18684266010920206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,4,balanced,0.18780267238616943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,4,power_law_1.2,0.5771584033966064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,4,balanced,0.18994667132695517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,4,balanced,0.19086933135986328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,4,balanced,0.1981119910875956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,4,balanced,0.19930134216944376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,4,balanced,0.2049600084622701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,4,balanced,0.22077866395314535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,4,balanced,0.22409067551294962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,4,balanced,0.2646506627400716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,4,balanced,0.26605333884557086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,4,balanced,0.3375626802444458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,4,balanced,0.359877347946167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,4,power_law_1.2,0.24599039554595947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,4,balanced,0.5246773163477579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,4,balanced,0.5794719854990641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,4,balanced,0.8411680062611898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,4,power_law_1.2,0.7911935806274414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,4,balanced,1.0046506722768147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,4,balanced,1.1547359625498455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,4,balanced,1.8196372985839844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,4,power_law_1.2,0.26708478927612306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,4,power_law_1.2,0.7678016185760498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,4,balanced,3.398122787475586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,4,power_law_1.2,0.2881727933883667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,4,power_law_1.2,0.8069503784179688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,4,power_law_1.2,0.8914943695068359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,4,power_law_1.2,0.28726398944854736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,4,power_law_1.2,0.8972031593322753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,4,power_law_1.2,0.3021951913833618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,4,power_law_1.2,0.9432640075683594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,4,power_law_1.2,0.304966402053833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,4,power_law_1.2,0.9761343955993652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,4,power_law_1.2,0.3155263900756836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,4,power_law_1.2,0.9983679771423339
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,4,power_law_1.2,0.3227135896682739
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,4,power_law_1.2,1.0308544158935546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,4,power_law_1.2,0.3307904005050659
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,4,power_law_1.2,1.0735744476318358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,16,8,power_law_1.2,0.039027199149131775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,16,8,power_law_1.2,0.03868159949779511
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,16,8,power_law_1.2,0.03153280019760132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,16,8,power_law_1.2,0.031481599807739256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,16,8,power_law_1.2,0.031839999556541446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,4,power_law_1.2,0.3470592021942139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,16,8,power_law_1.2,0.03265919983386993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,16,8,power_law_1.2,0.033792001008987424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,16,8,power_law_1.2,0.033913600444793704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,4,power_law_1.2,1.1336576461791992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,16,8,power_law_1.2,0.03497599959373474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,16,8,power_law_1.2,0.03488639891147614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,16,8,power_law_1.2,0.03589119911193848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,16,8,power_law_1.2,0.037567999958992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,16,8,power_law_1.2,0.03854719996452331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,16,8,power_law_1.2,0.040863999724388124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,4,power_law_1.2,0.34771199226379396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,16,8,power_law_1.2,0.05297920107841492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,16,8,power_law_1.2,0.05220479965209961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,4,power_law_1.2,1.2184191703796388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,16,8,power_law_1.2,0.05482239723205566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,16,8,power_law_1.2,0.06643199920654297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,16,8,power_law_1.2,0.06883839964866638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,16,8,power_law_1.2,0.08833919763565064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,16,8,power_law_1.2,0.10484479665756226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,16,8,power_law_1.2,0.1325376033782959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,4,power_law_1.2,0.3858943939208984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,16,8,power_law_1.2,0.16832000017166138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,4,power_law_1.2,1.482470417022705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,16,8,power_law_1.2,0.23544321060180665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,16,8,power_law_1.2,0.2806976079940796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,16,8,power_law_1.2,0.40309758186340333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,16,8,power_law_1.2,0.47586560249328613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,4,power_law_1.2,1.4891648292541504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,4,power_law_1.2,0.4135615825653076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,16,8,power_law_1.2,0.5897471904754639
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,16,8,power_law_1.2,0.9012543678283691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,16,8,power_law_1.2,1.6631168365478515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,4,power_law_1.2,1.7760704040527344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,4,power_law_1.2,0.4693759918212891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,4,power_law_1.2,1.8244863510131837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,4,power_law_1.2,0.583244800567627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,4,power_law_1.2,2.2332544326782227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,4,power_law_1.2,0.6822847843170166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,4,power_law_1.2,0.8364928245544434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,4,power_law_1.2,2.3355199813842775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,4,power_law_1.2,1.0058048248291016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,4,power_law_1.2,2.926316833496094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,4,power_law_1.2,1.2992256164550782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,4,power_law_1.2,3.770265579223633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,128,power_law_1.01,0.039059200882911684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,4,power_law_1.2,1.6187519073486327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,128,power_law_1.01,0.03928320109844208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,4,power_law_1.2,4.528351974487305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,128,power_law_1.01,0.03894400000572205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,128,power_law_1.01,0.03909760117530823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,128,power_law_1.01,0.03969280123710632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,128,power_law_1.01,0.04054400026798248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,4,power_law_1.2,2.414521598815918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,128,power_law_1.01,0.03999359905719757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,128,power_law_1.01,0.04044800102710724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,4,power_law_1.2,6.056364822387695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,128,power_law_1.01,0.04080640077590943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,128,power_law_1.01,0.04040960073471069
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,128,power_law_1.01,0.04177280068397522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,128,power_law_1.01,0.04137600064277649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,128,power_law_1.01,0.04164479970932007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,4,power_law_1.2,3.027769660949707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,128,power_law_1.01,0.042028799653053284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,128,power_law_1.01,0.04411520063877106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,128,power_law_1.01,0.04615040123462677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,128,power_law_1.01,0.04725759923458099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,128,power_law_1.01,0.04872959852218628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,128,power_law_1.01,0.05162879824638367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,4,power_law_1.2,11.28784637451172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,128,power_law_1.01,0.05745279788970947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,1,4,power_law_1.2,3.340345764160156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,128,power_law_1.01,0.06503040194511414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,128,power_law_1.01,0.08104959726333619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,128,power_law_1.01,0.09687039852142335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,128,power_law_1.01,0.11791360378265381
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,128,power_law_1.01,0.16281599998474122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,128,power_law_1.01,0.22271358966827393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,128,power_law_1.01,0.27628800868988035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,1,4,power_law_1.2,5.774284744262696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,128,power_law_1.01,0.39385600090026857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,128,power_law_1.01,0.5666816234588623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,128,power_law_1.01,1.1249664306640625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,8,2,power_law_1.2,0.03671680092811584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,8,2,power_law_1.2,0.04870400130748749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,8,2,power_law_1.2,0.04520959854125976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,8,2,power_law_1.2,0.055155199766159055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,8,2,power_law_1.2,0.05534719824790955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,8,2,power_law_1.2,0.05475199818611145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,8,2,power_law_1.2,0.0575872004032135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,8,2,power_law_1.2,0.058937597274780276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,8,2,power_law_1.2,0.06232320070266724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,8,2,power_law_1.2,0.06323840022087097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,8,2,power_law_1.2,0.06207360029220581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,8,2,power_law_1.2,0.06437119841575623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,8,2,power_law_1.2,0.06614400148391723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,8,2,power_law_1.2,0.06830080151557923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,8,2,power_law_1.2,0.07415680289268493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,8,2,power_law_1.2,0.08273919820785522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,1,4,power_law_1.2,11.621657562255859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,8,2,power_law_1.2,0.0821183979511261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,8,2,power_law_1.2,0.09685760140419006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,8,2,power_law_1.2,0.10615040063858032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,8,2,power_law_1.2,0.12269439697265624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,8,2,power_law_1.2,0.14379520416259767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,8,2,power_law_1.2,0.20072319507598876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,8,2,power_law_1.2,0.22561919689178467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,8,2,power_law_1.2,0.3401792049407959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,8,2,power_law_1.2,0.4109951972961426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,8,2,power_law_1.2,0.565664005279541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,8,2,power_law_1.2,0.7103487968444824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,8,2,power_law_1.2,0.9287360191345215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,8,2,power_law_1.2,1.5478143692016602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,8,2,power_law_1.2,3.032153511047363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,64,balanced,0.054383998115857445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,64,balanced,0.04569066564242045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,64,balanced,0.04595733185609182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,64,balanced,0.04584000011285146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,64,balanced,0.045791998505592346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,64,balanced,0.04982399940490723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,64,balanced,0.052015999952952065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,64,balanced,0.05242133140563965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,64,balanced,0.05031466484069824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,64,balanced,0.052015999952952065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,64,balanced,0.05223466455936432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,64,balanced,0.05431999762852987
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,64,balanced,0.05423999826113383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,64,balanced,0.0565226674079895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,64,balanced,0.06057066718737284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,64,balanced,0.0620959997177124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,64,balanced,0.06446933249632518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,64,balanced,0.06846933563550313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,64,balanced,0.07457066575686137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,64,balanced,0.08760533730189006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,64,balanced,0.09699199597040813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,64,balanced,0.11386666695276897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,64,balanced,0.1302880048751831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,64,balanced,0.159770667552948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,64,balanced,0.1959786613782247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,64,balanced,0.27244265874226886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,64,balanced,0.314794659614563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,64,balanced,0.3961706558863322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,64,balanced,0.5589600006739298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,64,balanced,1.0772266387939453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,16,4,power_law_1.2,0.037241598963737486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,16,4,power_law_1.2,0.033657601475715636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,16,4,power_law_1.2,0.033081600069999696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,16,4,power_law_1.2,0.03348479866981506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,16,4,power_law_1.2,0.03374080061912536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,16,4,power_law_1.2,0.03390080034732819
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,16,4,power_law_1.2,0.03493120074272156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,16,4,power_law_1.2,0.03501439988613129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,16,4,power_law_1.2,0.035180801153182985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,16,4,power_law_1.2,0.0355648010969162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,16,4,power_law_1.2,0.0359935998916626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,16,4,power_law_1.2,0.03736959993839264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,16,4,power_law_1.2,0.037254399061203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,16,4,power_law_1.2,0.03830400109291077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,16,4,power_law_1.2,0.043705600500106814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,16,4,power_law_1.2,0.045363199710845944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,16,4,power_law_1.2,0.05112959742546082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,16,4,power_law_1.2,0.056992000341415404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,16,4,power_law_1.2,0.06904320120811462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,16,4,power_law_1.2,0.07905279994010925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,16,4,power_law_1.2,0.08783360123634339
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,16,4,power_law_1.2,0.11964160203933716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,16,4,power_law_1.2,0.14309760332107543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,16,4,power_law_1.2,0.21356160640716554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,16,4,power_law_1.2,0.22200961112976075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,16,4,power_law_1.2,0.3288640022277832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,16,4,power_law_1.2,0.44627838134765624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,16,power_law_1.01,0.05533440113067627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,16,4,power_law_1.2,0.6634751796722412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,16,power_law_1.01,0.06693120002746582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,16,4,power_law_1.2,0.7301631927490234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,16,power_law_1.01,0.046854400634765626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,16,4,power_law_1.2,1.6567615509033202
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,16,power_law_1.01,0.0484607994556427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,16,power_law_1.01,0.0498879998922348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,16,power_law_1.01,0.055846399068832396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,16,power_law_1.01,0.061459201574325564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,16,power_law_1.01,0.05969280004501343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,32,8,power_law_1.01,0.05727360248565674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,16,power_law_1.01,0.06450560092926025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,16,power_law_1.01,0.06833919882774353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,32,8,power_law_1.01,0.06382079720497132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,16,power_law_1.01,0.06538239717483521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,16,power_law_1.01,0.0690496027469635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,32,8,power_law_1.01,0.07255039811134338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,16,power_law_1.01,0.07177600264549255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,16,power_law_1.01,0.07356160283088684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,32,8,power_law_1.01,0.057055997848510745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,16,power_law_1.01,0.08001919984817504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,16,power_law_1.01,0.08550400137901307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,32,8,power_law_1.01,0.06215680241584778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,16,power_law_1.01,0.08458880186080933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,16,power_law_1.01,0.09941759705543518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,32,8,power_law_1.01,0.06883199810981751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,16,power_law_1.01,0.10992000102996827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,16,power_law_1.01,0.13147519826889037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,32,8,power_law_1.01,0.07008640170097351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,16,power_law_1.01,0.15973119735717772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,16,power_law_1.01,0.19968639612197875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,32,8,power_law_1.01,0.07592960000038147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,8,power_law_1.01,0.14687999486923217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,16,power_law_1.01,0.2428352117538452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,8,power_law_1.01,0.31025919914245603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,32,8,power_law_1.01,0.07574399709701538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,16,power_law_1.01,0.31685121059417726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,8,power_law_1.01,0.1522495985031128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,16,power_law_1.01,0.4259007930755615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,32,8,power_law_1.01,0.07827839851379395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,8,power_law_1.01,0.13920639753341674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,16,power_law_1.01,0.5852672100067139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,32,8,power_law_1.01,0.08009600043296813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,8,power_law_1.01,0.21064319610595703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,16,power_law_1.01,0.7447679996490478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,8,power_law_1.01,0.25770881175994875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,32,8,power_law_1.01,0.08356480002403259
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,8,power_law_1.01,0.34755840301513674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,16,power_law_1.01,0.9376640319824219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,32,8,power_law_1.01,0.08561279773712158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,8,power_law_1.01,0.3456831932067871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,32,8,power_law_1.01,0.08693119883537292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,8,power_law_1.01,0.35358080863952634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,16,power_law_1.01,1.4205439567565918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,8,power_law_1.01,0.37645440101623534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,32,8,power_law_1.01,0.09569919705390931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,8,power_law_1.01,0.36101760864257815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,32,8,power_law_1.01,0.09658240079879761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,8,power_law_1.01,0.36173439025878906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,16,power_law_1.01,2.9922176361083985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,32,8,power_law_1.01,0.10743680000305175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,8,power_law_1.01,0.37107200622558595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,8,power_law_1.01,0.415334415435791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,32,8,power_law_1.01,0.12128000259399414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,8,power_law_1.01,0.39806079864501953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,32,8,power_law_1.01,0.141484797000885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,8,power_law_1.01,0.40840959548950195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,32,8,power_law_1.01,0.17118719816207886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,8,power_law_1.01,0.46021761894226076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,8,power_law_1.01,0.5027391910552979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,32,8,power_law_1.01,0.17894400358200074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,8,power_law_1.01,0.4881472110748291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,32,8,power_law_1.01,0.23184640407562257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,8,power_law_1.01,0.5826752185821533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,8,power_law_1.01,0.5517439842224121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,32,8,power_law_1.01,0.2942591905593872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,8,power_law_1.01,0.6494592189788818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,32,8,power_law_1.01,0.39915521144866944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,8,power_law_1.01,0.7081984043121338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,32,8,power_law_1.01,0.523583984375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,8,power_law_1.01,0.8758208274841308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,1,8,balanced,0.08298666775226593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,1,8,balanced,0.10533866286277771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,1,8,balanced,0.0831520011027654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,1,8,balanced,0.08400533596674602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,1,8,balanced,0.08416533470153809
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,1,8,balanced,0.08434666196505229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,1,8,balanced,0.08480000495910645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,1,8,balanced,0.08455466230710347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,1,8,balanced,0.08661333719889323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,1,8,balanced,0.08553600311279297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,1,8,balanced,0.0895253320535024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,1,8,balanced,0.08915733297665913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,1,8,balanced,0.08873599767684937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,1,8,balanced,0.09640000263849895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,1,8,balanced,0.10114133358001709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,1,8,balanced,0.1030346651871999
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,1,8,balanced,0.10700800021489461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,1,8,balanced,0.1304693321386973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,1,8,balanced,0.13334932923316956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,1,8,balanced,0.16310399770736694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,1,8,balanced,0.1920479933420817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,32,8,power_law_1.01,0.7305408000946045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,8,power_law_1.01,1.0710080146789551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,1,8,balanced,0.23852266867955527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,1,8,balanced,0.2890026569366455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,1,8,balanced,0.404202659924825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,1,8,balanced,0.5051093498865763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,1,8,balanced,0.7136373519897461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,1,8,balanced,0.9553546905517578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,1,8,balanced,1.1620426972707112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,1,8,balanced,1.8244585990905762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,8,power_law_1.01,1.3598464012145997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,32,8,power_law_1.01,0.9481663703918457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,1,8,balanced,3.637125333150228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,8,power_law_1.01,1.7176767349243165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,32,8,power_law_1.01,1.134438419342041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,8,power_law_1.01,2.1031551361083984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,32,8,power_law_1.01,1.7272512435913085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,8,power_law_1.01,3.197427177429199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,32,8,power_law_1.01,3.5156993865966797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,8,power_law_1.01,5.5156608581542965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,16,power_law_1.2,0.08081279993057251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,16,power_law_1.2,0.07316480278968811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,16,power_law_1.2,0.04906240105628967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,16,power_law_1.2,0.050419199466705325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,16,power_law_1.2,0.053465598821640016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,16,power_law_1.2,0.05848960280418396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,16,power_law_1.2,0.06665599942207337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,16,power_law_1.2,0.06694399714469909
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,16,power_law_1.2,0.06707199811935424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,16,power_law_1.2,0.06911360025405884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,8,power_law_1.01,0.05299199819564819
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,16,power_law_1.2,0.07057920098304749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,32,4,power_law_1.2,0.038252800703048706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,32,4,power_law_1.2,0.035462400317192076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,32,4,power_law_1.2,0.03424000144004822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,32,4,power_law_1.2,0.03516159951686859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,8,power_law_1.01,0.05864319801330566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,16,power_law_1.2,0.07204480171203613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,32,4,power_law_1.2,0.03551360070705414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,32,4,power_law_1.2,0.035872000455856326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,32,4,power_law_1.2,0.03639039993286133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,32,4,power_law_1.2,0.03681280016899109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,8,power_law_1.01,0.05309439897537231
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,16,power_law_1.2,0.07471359968185425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,32,4,power_law_1.2,0.03658879995346069
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,32,4,power_law_1.2,0.03823359906673431
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,32,4,power_law_1.2,0.03935360014438629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,32,4,power_law_1.2,0.04047360122203827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,32,4,power_law_1.2,0.04066559970378876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,16,power_law_1.2,0.07637119889259339
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,8,power_law_1.01,0.05448960065841675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,32,4,power_law_1.2,0.04145280122756958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,32,4,power_law_1.2,0.0469760000705719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,32,4,power_law_1.2,0.04695039987564087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,32,4,power_law_1.2,0.051667201519012454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,16,power_law_1.2,0.08744320273399353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,32,4,power_law_1.2,0.056620800495147706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,8,power_law_1.01,0.05885440111160278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,32,4,power_law_1.2,0.06485120058059693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,32,4,power_law_1.2,0.08339840173721313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,32,4,power_law_1.2,0.1026047945022583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,32,4,power_law_1.2,0.1305343985557556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,16,power_law_1.2,0.09074559807777405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,8,power_law_1.01,0.06595839858055115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,32,4,power_law_1.2,0.15066239833831788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,32,4,power_law_1.2,0.23511040210723877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,32,4,power_law_1.2,0.29367039203643797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,8,power_law_1.01,0.06913920044898987
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,16,power_law_1.2,0.09297919869422913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,32,4,power_law_1.2,0.4236735820770264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,32,4,power_law_1.2,0.6070144176483154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,8,power_law_1.01,0.07214080095291138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,16,power_law_1.2,0.10263680219650269
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,32,4,power_law_1.2,0.7748288154602051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,32,4,power_law_1.2,0.945747184753418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,8,power_law_1.01,0.0741375982761383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,16,power_law_1.2,0.10895359516143799
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,8,power_law_1.01,0.07521920204162598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,32,4,power_law_1.2,2.1815935134887696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,16,power_law_1.2,0.12929279804229737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,8,power_law_1.01,0.0760640025138855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,16,power_law_1.2,0.16591999530792237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,8,power_law_1.01,0.08021759986877441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,16,power_law_1.2,0.1933632016181946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,8,power_law_1.01,0.08240640163421631
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,16,power_law_1.2,0.2200256109237671
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,8,power_law_1.01,0.08270080089569092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,8,power_law_1.01,0.09095680117607116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,16,power_law_1.2,0.3186624050140381
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,8,power_law_1.01,0.09511039853096008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,16,power_law_1.2,0.3936383962631226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,8,power_law_1.01,0.10714240074157715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,16,power_law_1.2,0.5801599979400635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,8,power_law_1.01,0.12051199674606324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,8,power_law_1.01,0.13767679929733276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,16,power_law_1.2,0.7495552062988281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,8,power_law_1.01,0.15897599458694459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,8,power_law_1.01,0.18853119611740113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,16,power_law_1.2,0.8031680107116699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,8,power_law_1.01,0.2285952091217041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,32,balanced,0.06038400034109751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,32,balanced,0.06012799839178721
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,32,balanced,0.059978668888409935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,32,balanced,0.06226666768391927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,32,balanced,0.0729013333717982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,32,balanced,0.09312533338864644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,32,balanced,0.10452799995740254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,32,balanced,0.10653866330782573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,8,power_law_1.01,0.2688960075378418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,32,balanced,0.10679466525713603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,32,balanced,0.10517866412798564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,16,power_law_1.2,1.396454429626465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,32,balanced,0.10711466272672017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,32,balanced,0.10598933696746826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,32,balanced,0.11102933684984843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,32,balanced,0.10495466987291972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,32,balanced,0.11136533816655476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,32,balanced,0.11333333452542622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,32,balanced,0.11979200442632039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,32,balanced,0.1241973340511322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,32,balanced,0.13126933574676514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,32,balanced,0.14432000120480856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,32,balanced,0.15633066495259604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,32,balanced,0.1830986738204956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,32,balanced,0.208624005317688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,32,balanced,0.26124799251556396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,8,power_law_1.01,0.3516992092132568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,32,balanced,0.3394026756286621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,32,balanced,0.44124265511830646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,32,balanced,0.5548906723658243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,32,balanced,0.688431978225708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,32,balanced,1.040218671162923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,8,power_law_1.01,0.4886847972869873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,32,balanced,1.9923466046651204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,16,power_law_1.2,2.748588752746582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,8,power_law_1.01,0.6959487915039062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,8,power_law_1.01,0.8523584365844726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,8,power_law_1.01,1.0615551948547364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,8,power_law_1.01,1.6962303161621093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,8,power_law_1.01,3.3297855377197267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,64,power_law_1.2,0.11983360052108764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,8,power_law_1.01,0.09507840275764465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,64,power_law_1.2,0.11998720169067383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,8,power_law_1.01,0.1908735990524292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,8,power_law_1.01,0.08423680067062378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,64,power_law_1.2,0.11375360488891602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,8,power_law_1.01,0.09984639883041382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,64,power_law_1.2,0.07879679799079894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,8,power_law_1.01,0.12401280403137208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,64,power_law_1.2,0.085452800989151
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,8,power_law_1.01,0.14562560319900514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,8,power_law_1.01,0.18844799995422362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,64,power_law_1.2,0.09259520173072815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,8,power_law_1.01,0.19923839569091797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,64,power_law_1.2,0.09688320159912109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,8,power_law_1.01,0.20262401103973388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,32,power_law_1.01,0.09331200122833253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,64,power_law_1.2,0.0959231972694397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,8,power_law_1.01,0.20602240562438964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,32,power_law_1.01,0.09591040015220642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,32,power_law_1.01,0.05907840132713318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,8,power_law_1.01,0.20435841083526612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,64,power_law_1.2,0.10807679891586304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,32,power_law_1.01,0.0639743983745575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,8,power_law_1.01,0.21709439754486085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,32,power_law_1.01,0.06606079936027527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,64,power_law_1.2,0.10113919973373413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,32,power_law_1.01,0.07611520290374756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,8,power_law_1.01,0.22263679504394532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,32,power_law_1.01,0.07986559867858886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,8,power_law_1.01,0.23434240818023683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,64,power_law_1.2,0.10708479881286621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,32,power_law_1.01,0.07888000011444092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,32,power_law_1.01,0.08082559704780579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,8,power_law_1.01,0.23682560920715331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,64,power_law_1.2,0.10807039737701415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,32,power_law_1.01,0.08065279722213745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,8,power_law_1.01,0.2364351987838745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,32,power_law_1.01,0.08227840065956116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,64,power_law_1.2,0.11008640527725219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,32,power_law_1.01,0.08042880296707153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,8,power_law_1.01,0.26063361167907717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,32,power_law_1.01,0.08149120211601257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,64,power_law_1.2,0.11670399904251098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,8,power_law_1.01,0.2799232006072998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,32,power_law_1.01,0.08371840119361877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,32,power_law_1.01,0.09057919979095459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,8,power_law_1.01,0.2752000093460083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,64,power_law_1.2,0.12034560441970825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,32,power_law_1.01,0.09646720290184022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,8,power_law_1.01,0.3265088081359863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,32,power_law_1.01,0.09873279929161072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,64,power_law_1.2,0.12159359455108643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,32,power_law_1.01,0.11367679834365844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,8,power_law_1.01,0.3354880094528198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,32,power_law_1.01,0.11931519508361817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,64,power_law_1.2,0.12471040487289428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,8,power_law_1.01,0.4026815891265869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,32,power_law_1.01,0.1364159941673279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,32,power_law_1.01,0.14755840301513673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,64,power_law_1.2,0.14161280393600464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,8,power_law_1.01,0.4554304122924805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,32,power_law_1.01,0.1752128005027771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,64,power_law_1.2,0.1546239972114563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,8,power_law_1.01,0.5826047897338867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,32,power_law_1.01,0.21777920722961425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,32,power_law_1.01,0.2989887952804565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,64,power_law_1.2,0.19287680387496947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,8,power_law_1.01,0.6809279918670654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,32,power_law_1.01,0.3615744113922119
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,64,power_law_1.2,0.21987199783325195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,8,power_law_1.01,0.8736960411071777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,32,power_law_1.01,0.5355840206146241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,64,power_law_1.2,0.2846463918685913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,32,power_law_1.01,0.5841919898986816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,8,power_law_1.01,1.2240511894226074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,64,power_law_1.2,0.3409856081008911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,32,power_law_1.01,0.7256127834320069
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,8,power_law_1.01,1.5033408164978028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,64,power_law_1.2,0.48968958854675293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,32,power_law_1.01,1.1822272300720216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,8,power_law_1.01,2.0792383193969726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,64,power_law_1.2,0.6188992023468017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,64,power_law_1.2,0.9415871620178222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,32,power_law_1.01,2.163884735107422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,8,power_law_1.01,3.9189697265625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,64,power_law_1.2,1.1357888221740722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,64,power_law_1.2,1.4182016372680664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,64,power_law_1.2,2.31591682434082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,64,power_law_1.2,4.752288055419922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,8,balanced,0.05392533540725708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,8,balanced,0.05412800113360087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,8,balanced,0.06213866670926412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,8,balanced,0.08342933654785156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,8,balanced,0.11635200182596843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,8,balanced,0.17671465873718262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,8,balanced,0.23604265848795572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,8,balanced,0.2349546750386556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,8,balanced,0.237936000029246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,8,balanced,0.23759466409683228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,8,balanced,0.23897600173950195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,8,balanced,0.2434986631075541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,8,balanced,0.24234666426976523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,8,balanced,0.24579733610153198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,8,balanced,0.2531733314196269
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,8,balanced,0.2550026575724284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,8,balanced,0.26241066058476764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,8,balanced,0.2744800051053365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,8,balanced,0.28365333875020343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,8,balanced,0.30317866802215576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,8,balanced,0.3253119985262553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,8,balanced,0.37144001324971515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,8,balanced,0.4190133412679036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,8,balanced,0.5105280081431071
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,8,balanced,0.6207946538925171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,8,power_law_1.2,0.052934402227401735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,8,balanced,0.7936267058054606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,8,balanced,1.039136012395223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,8,balanced,1.2990506490071614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,8,power_law_1.2,0.058432000875473025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,8,balanced,1.8609813054402669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,8,balanced,3.529616038004557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,8,power_law_1.2,0.05097600221633911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,8,power_law_1.2,0.05447040200233459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,8,power_law_1.2,0.058310401439666745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,8,power_law_1.2,0.06471679806709289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,8,power_law_1.2,0.07015680074691773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,8,power_law_1.2,0.0716863989830017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,1,balanced,0.037962667644023895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,1,balanced,0.040336000422636666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,1,balanced,0.06031466523806254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,8,power_law_1.2,0.07306240200996399
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,1,balanced,0.08763733506202698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,1,balanced,0.13198933005332947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,1,balanced,0.1360586682955424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,1,balanced,0.13949867089589438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,1,balanced,0.13803733388582864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,1,balanced,0.14174933234850565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,1,balanced,0.1483786702156067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,1,balanced,0.14724799990653992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,1,balanced,0.154341330130895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,1,balanced,0.15610667069753012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,1,balanced,0.16029333074887595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,1,balanced,0.16687999169031778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,1,balanced,0.17594132820765176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,8,power_law_1.2,0.07527679800987244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,1,balanced,0.18608532349268594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,1,balanced,0.2155839999516805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,1,balanced,0.23516800006230673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,1,balanced,0.3084533413251241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,1,balanced,0.34149332841237384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,1,balanced,0.4786613384882609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,1,balanced,0.5705866813659668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,1,balanced,0.8130559921264648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,8,power_law_1.2,0.07841280102729797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,1,balanced,1.0380480289459229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,1,balanced,1.5043466885884602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,1,balanced,1.9510399500528972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,8,power_law_1.2,0.07985919713973999
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,1,balanced,2.39737606048584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,8,power_law_1.2,0.08216320276260376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,1,balanced,3.7573493321736655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,8,power_law_1.2,0.08341119885444641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,1,balanced,7.28441047668457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,8,power_law_1.2,0.09170560240745544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,8,power_law_1.2,0.10030720233917237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,1,power_law_1.01,0.09464319944381713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,8,power_law_1.2,0.10559359788894654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,1,power_law_1.01,0.12327040433883667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,8,power_law_1.2,0.12316800355911255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,1,power_law_1.01,0.16129280328750611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,1,power_law_1.01,0.24062719345092773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,8,power_law_1.2,0.13994879722595216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,1,power_law_1.01,0.3296832084655762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,8,power_law_1.2,0.163046395778656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,1,power_law_1.01,0.5479296207427978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,1,power_law_1.01,0.5675007820129394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,8,power_law_1.2,0.18447359800338745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,1,power_law_1.01,0.6037888050079345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,8,power_law_1.2,0.2531008005142212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,1,power_law_1.01,0.6174528121948242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,1,power_law_1.01,0.6425407886505127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,8,power_law_1.2,0.27594239711761476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,1,power_law_1.01,0.6552256107330322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,8,power_law_1.2,0.3658112049102783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,1,power_law_1.01,0.6745471954345703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,1,power_law_1.01,0.6697535991668702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,8,power_law_1.2,0.5039040088653565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,1,power_law_1.01,0.706496000289917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,1,power_law_1.01,0.7172736167907715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,8,power_law_1.2,0.7883584022521972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,1,power_law_1.01,0.7386879920959473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,8,power_law_1.01,0.06872320175170898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,1,power_law_1.01,0.7678592205047607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,8,power_law_1.01,0.09708799719810486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,8,power_law_1.2,0.929196834564209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,1,power_law_1.01,0.832038402557373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,8,power_law_1.01,0.061887997388839724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,1,power_law_1.01,0.9093376159667969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,8,power_law_1.2,1.2056192398071288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,8,power_law_1.01,0.06962559819221496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,1,power_law_1.01,1.0645759582519532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,8,power_law_1.01,0.09016960263252258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,1,power_law_1.01,1.1242560386657714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,8,power_law_1.01,0.10447360277175903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,8,power_law_1.2,1.7407936096191405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,1,power_law_1.01,1.3840831756591796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,8,power_law_1.01,0.1252671957015991
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,8,power_law_1.01,0.13191039562225343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,1,power_law_1.01,1.6406848907470704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,8,power_law_1.01,0.13362560272216797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,1,power_law_1.01,2.172064018249512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,8,power_law_1.01,0.1347839951515198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,8,power_law_1.2,4.064646530151367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,1,power_law_1.01,2.7822336196899413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,8,power_law_1.01,0.1393407940864563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,8,power_law_1.01,0.13822720050811768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,1,power_law_1.01,3.748275375366211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,8,power_law_1.01,0.14366079568862916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,8,power_law_1.01,0.1435263991355896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,8,power_law_1.01,0.148364794254303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,1,power_law_1.01,4.689952087402344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,8,power_law_1.01,0.15601919889450072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,8,power_law_1.01,0.17023999691009523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,1,power_law_1.01,5.649113464355469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,8,power_law_1.01,0.19404799938201905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,8,power_law_1.01,0.1995967984199524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,8,power_law_1.01,0.24785280227661133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,8,power_law_1.01,0.2900799989700317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,1,power_law_1.01,8.525791931152344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,8,power_law_1.01,0.37263998985290525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,8,power_law_1.01,0.39040000438690187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,2,power_law_1.01,0.041606399416923526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,8,power_law_1.01,0.5158783912658691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,2,power_law_1.01,0.04864639937877655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,2,power_law_1.01,0.05121920108795166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,8,power_law_1.01,0.6414783954620361
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,2,power_law_1.01,0.06392319798469544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,2,power_law_1.01,0.07552000284194946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,2,power_law_1.01,0.09502720236778259
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,1,power_law_1.01,16.5162109375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,8,power_law_1.01,0.972332763671875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,2,power_law_1.01,0.09997439980506898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,2,power_law_1.01,0.10405119657516479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,8,power_law_1.01,1.2094847679138183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,2,power_law_1.01,0.1033087968826294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,2,power_law_1.01,0.10654720067977905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,2,power_law_1.01,0.10837759971618652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,8,power_law_1.01,1.45414400100708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,2,power_law_1.01,0.11364480257034301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,2,power_law_1.01,0.1161728024482727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,2,power_law_1.01,0.12117120027542114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,2,power_law_1.01,0.1273151993751526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,2,power_law_1.01,0.12904959917068481
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,8,power_law_1.01,2.229484748840332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,2,power_law_1.01,0.1382848024368286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,2,power_law_1.01,0.1508288025856018
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,2,power_law_1.01,0.17612160444259645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,2,power_law_1.01,0.20790400505065917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,2,power_law_1.01,0.23714559078216552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,8,power_law_1.01,4.295264053344726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,2,power_law_1.01,0.31363840103149415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,2,power_law_1.01,0.389516806602478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,2,power_law_1.01,0.5227200031280518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,2,power_law_1.01,0.6762303829193115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,2,power_law_1.01,0.9577216148376465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,2,power_law_1.01,1.2737152099609375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,2,power_law_1.01,1.6837247848510741
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,16,power_law_1.2,0.049414399266242984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,16,power_law_1.2,0.0934719979763031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,2,power_law_1.01,2.360953521728516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,16,power_law_1.2,0.06867200136184692
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,16,power_law_1.2,0.052172797918319705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,2,power_law_1.01,4.6798656463623045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,16,power_law_1.2,0.05885440111160278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,64,power_law_1.01,0.17426559925079346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,16,power_law_1.2,0.06480000019073487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,64,power_law_1.01,0.17196160554885864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,16,power_law_1.2,0.06894720196723939
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,64,power_law_1.01,0.16828160285949706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,16,power_law_1.2,0.07864959836006165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,64,power_law_1.01,0.10675840377807617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,16,power_law_1.2,0.07888640165328979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,64,power_law_1.01,0.11268479824066162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,16,power_law_1.2,0.08163840174674988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,64,power_law_1.01,0.14232319593429565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,16,power_law_1.2,0.08179839849472045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,64,power_law_1.01,0.14613759517669678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,16,power_law_1.2,0.08710399866104127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,64,power_law_1.01,0.16507519483566285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,16,power_law_1.2,0.0890175998210907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,2,balanced,0.043935999274253845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,2,balanced,0.04029866556326548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,64,power_law_1.01,0.15023360252380372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,2,balanced,0.04414399961630503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,2,balanced,0.051967998345692955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,2,balanced,0.06108266611893972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,2,balanced,0.0627040018637975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,2,balanced,0.06294399996598561
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,2,balanced,0.06296533346176147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,16,power_law_1.2,0.0939136028289795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,2,balanced,0.06459199885527293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,2,balanced,0.06439466774463654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,2,balanced,0.06668266654014587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,2,balanced,0.06897599995136261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,2,balanced,0.06795200208822887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,2,balanced,0.07195200026035309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,2,balanced,0.07459199925263722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,2,balanced,0.07790400087833405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,2,balanced,0.08297599852085114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,2,balanced,0.09963200489679973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,2,balanced,0.10948266585667928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,2,balanced,0.131632000207901
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,2,balanced,0.15456533432006836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,2,balanced,0.21024000644683838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,2,balanced,0.24950400988260904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,64,power_law_1.01,0.16936960220336914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,2,balanced,0.3604533274968465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,2,balanced,0.45977067947387695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,2,balanced,0.6754559675852457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,16,power_law_1.2,0.10341119766235352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,2,balanced,0.8763999938964844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,2,balanced,1.0895946820576985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,2,balanced,1.7088054021199544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,64,power_law_1.01,0.16177279949188234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,16,power_law_1.2,0.10710400342941284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,2,balanced,3.3201173146565757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,64,power_law_1.01,0.16408319473266603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,16,power_law_1.2,0.11400959491729737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,64,power_law_1.01,0.15649919509887694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,16,power_law_1.2,0.1303231954574585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,64,power_law_1.01,0.1513983964920044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,16,power_law_1.2,0.1416383981704712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,64,power_law_1.01,0.16656639575958251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,16,power_law_1.2,0.17604479789733887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,64,power_law_1.01,0.15364480018615723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,16,power_law_1.2,0.20559999942779542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,64,power_law_1.01,0.17710720300674437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,16,power_law_1.2,0.2684927940368652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,64,power_law_1.01,0.17686400413513184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,16,power_law_1.2,0.3129791975021362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,64,power_law_1.01,0.18732160329818726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,16,power_law_1.2,0.410265588760376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,64,power_law_1.01,0.20660479068756105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,16,power_law_1.2,0.5460351943969727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,64,power_law_1.01,0.22663040161132814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,32,power_law_1.2,0.07443199753761291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,64,power_law_1.01,0.29075839519500735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,16,power_law_1.2,0.7611519813537597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,64,power_law_1.01,0.3506880044937134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,32,power_law_1.2,0.06959999799728393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,16,power_law_1.2,1.0541695594787597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,64,power_law_1.01,0.4791168212890625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,32,power_law_1.2,0.05330560207366943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,16,power_law_1.2,1.2680959701538086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,64,power_law_1.01,0.5586368083953858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,32,power_law_1.2,0.05742080211639404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,64,power_law_1.01,0.8007871627807617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,32,power_law_1.2,0.059033602476119995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,16,power_law_1.2,2.0123519897460938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,64,power_law_1.01,0.9734463691711426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,32,power_law_1.2,0.07120000123977661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,32,power_law_1.2,0.07431039810180665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,64,power_law_1.01,1.205951976776123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,16,power_law_1.2,4.180646514892578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,32,power_law_1.2,0.07400959730148315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,64,power_law_1.01,1.8283136367797852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,32,power_law_1.2,0.0754688024520874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,32,power_law_1.2,0.07516160011291503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,32,power_law_1.2,0.07729920148849487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,64,power_law_1.01,3.6706817626953123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,32,power_law_1.2,0.08175359964370728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,32,power_law_1.2,0.08179200291633607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,32,power_law_1.2,0.08551040291786194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,32,power_law_1.2,0.08949120044708252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,32,8,balanced,0.029509333272775013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,32,8,balanced,0.03141866624355316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,32,8,balanced,0.029680001238981884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,32,8,balanced,0.03152533372243246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,32,8,balanced,0.03151999910672506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,32,8,balanced,0.031210665901501972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,32,8,balanced,0.03146666785081228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,32,8,balanced,0.03148266673088074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,32,8,balanced,0.03160000095764796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,32,8,balanced,0.03173866619666418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,32,8,balanced,0.03146666785081228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,32,8,balanced,0.0334346666932106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,32,8,balanced,0.03348266581694285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,32,8,balanced,0.033546666304270424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,32,8,balanced,0.03826666623353958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,32,8,balanced,0.03782399992148081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,32,8,balanced,0.03993066648642222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,32,8,balanced,0.0415786678592364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,32,8,balanced,0.04389866689840952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,32,8,balanced,0.050240000089009605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,32,8,balanced,0.05194133520126343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,32,8,balanced,0.06348800162474315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,32,8,balanced,0.08308266599973042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,32,8,balanced,0.12352533141771953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,32,8,balanced,0.1516320010026296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,32,8,balanced,0.21209599574406943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,32,power_law_1.2,0.09459199905395507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,32,8,balanced,0.25569067398707074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,32,8,balanced,0.3147786657015483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,32,8,balanced,0.4604533513387044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,32,8,balanced,0.8740959962209066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,32,power_law_1.2,0.10501760244369507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,32,power_law_1.2,0.12199039459228515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,1,power_law_1.01,0.050547200441360476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,1,power_law_1.01,0.0801472008228302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,32,power_law_1.2,0.1223296046257019
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,1,power_law_1.01,0.0815168023109436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,32,power_law_1.2,0.14396799802780152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,1,power_law_1.01,0.12762880325317383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,1,power_law_1.01,0.18083839416503905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,32,power_law_1.2,0.17709439992904663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,1,power_law_1.01,0.24182400703430176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,1,power_law_1.01,0.32293119430541994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,32,power_law_1.2,0.22773120403289795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,1,power_law_1.01,0.342630410194397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,1,power_law_1.01,0.36190080642700195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,32,power_law_1.2,0.24131839275360106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,1,power_law_1.01,0.37048959732055664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,1,power_law_1.01,0.3876159906387329
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,64,power_law_1.2,0.03247359991073608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,1,power_law_1.01,0.4033088207244873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,64,power_law_1.2,0.0296640008687973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,32,power_law_1.2,0.3281280040740967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,64,power_law_1.2,0.028198400139808656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,1,power_law_1.01,0.41378560066223147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,64,power_law_1.2,0.0293503999710083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,64,power_law_1.2,0.02988159954547882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,1,power_law_1.01,0.4268223762512207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,64,power_law_1.2,0.03049600124359131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,32,power_law_1.2,0.41365761756896974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,1,power_law_1.01,0.44659199714660647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,64,power_law_1.2,0.03054080009460449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,64,power_law_1.2,0.031020799279212953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,1,power_law_1.01,0.45381760597229004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,64,power_law_1.2,0.031116798520088196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,64,power_law_1.2,0.03160319924354553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,1,power_law_1.01,0.4995840072631836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,64,power_law_1.2,0.032364800572395325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,32,power_law_1.2,0.5757823944091797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,64,power_law_1.2,0.03281919956207276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,1,power_law_1.01,0.5360383987426758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,64,power_law_1.2,0.03372800052165985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,64,power_law_1.2,0.0341376006603241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,1,power_law_1.01,0.5905280113220215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,64,power_law_1.2,0.037222400307655334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,64,power_law_1.2,0.038201600313186646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,1,power_law_1.01,0.6997439861297607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,64,power_law_1.2,0.04012799859046936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,32,power_law_1.2,0.7135488033294678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,64,power_law_1.2,0.044947201013565065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,1,power_law_1.01,0.8083456039428711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,64,power_law_1.2,0.0510591983795166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,64,power_law_1.2,0.06379520297050476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,1,power_law_1.01,1.035097599029541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,64,power_law_1.2,0.06273919939994813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,64,power_law_1.2,0.08167679905891419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,1,power_law_1.01,1.1662400245666504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,4,32,power_law_1.2,0.8249855995178222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,64,power_law_1.2,0.09781119823455811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,64,power_law_1.2,0.14142719507217408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,1,power_law_1.01,1.5422207832336425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,64,power_law_1.2,0.1817471981048584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,1,power_law_1.01,1.9130048751831055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,64,power_law_1.2,0.308953595161438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,64,power_law_1.2,0.3513727903366089
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,1,power_law_1.01,2.636038398742676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,4,32,power_law_1.2,1.310758399963379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,64,power_law_1.2,0.5267903804779053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,2,power_law_1.01,0.04632959961891174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,1,power_law_1.01,3.375987243652344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,64,power_law_1.2,0.7760128021240235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,2,power_law_1.01,0.04961279928684235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,1,power_law_1.01,4.118483352661133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,2,power_law_1.01,0.05231999754905701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,64,power_law_1.2,1.705023956298828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,2,power_law_1.01,0.06229119896888733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,2,power_law_1.01,0.07617279887199402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,1,power_law_1.01,6.189215850830078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,4,32,power_law_1.2,3.7619457244873047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,2,power_law_1.01,0.08849279880523682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,2,power_law_1.01,0.11284480094909669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,2,power_law_1.01,0.11547520160675048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,2,power_law_1.01,0.1195904016494751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,1,power_law_1.01,11.92138900756836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,2,power_law_1.01,0.12016639709472657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,2,power_law_1.01,0.1315135955810547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,2,power_law_1.01,0.13299200534820557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,2,power_law_1.01,0.1388543963432312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,2,power_law_1.2,0.03237760066986084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,2,power_law_1.01,0.14998400211334229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,2,power_law_1.2,0.032076799869537355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,2,power_law_1.2,0.03466239869594574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,2,power_law_1.01,0.1636415958404541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,2,power_law_1.2,0.03575679957866669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,2,power_law_1.2,0.04067200124263763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,2,power_law_1.2,0.05234559774398804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,2,power_law_1.01,0.17815040349960326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,2,power_law_1.2,0.05238400101661682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,32,power_law_1.01,0.05633280277252197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,2,power_law_1.2,0.05437440276145935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,32,power_law_1.01,0.04552960097789764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,2,power_law_1.01,0.17225600481033326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,2,power_law_1.2,0.05671039819717407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,32,power_law_1.01,0.04325760006904602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,2,power_law_1.2,0.06112639904022217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,2,power_law_1.2,0.06063359975814819
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,32,power_law_1.01,0.04394879937171936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,2,power_law_1.01,0.20072319507598876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,2,power_law_1.2,0.06543359756469727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,32,power_law_1.01,0.045132800936698914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,2,power_law_1.2,0.06999679803848266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,32,power_law_1.01,0.0460671991109848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,2,power_law_1.01,0.2222912073135376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,2,power_law_1.2,0.06980479955673217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,32,power_law_1.01,0.0463808000087738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,2,power_law_1.2,0.07580800056457519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,32,power_law_1.01,0.07886080145835876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,32,power_law_1.01,0.04684799909591675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,2,power_law_1.01,0.28170878887176515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,2,power_law_1.2,0.08131200075149536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,32,power_law_1.01,0.04668160080909729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,32,power_law_1.01,0.08359040021896362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,2,power_law_1.2,0.09016960263252258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,32,power_law_1.01,0.047577598690986635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,2,power_law_1.2,0.10721919536590577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,2,power_law_1.01,0.3382719993591309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,32,power_law_1.01,0.053388798236846925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,2,power_law_1.2,0.12653440237045288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,32,power_law_1.01,0.04919680058956146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,2,power_law_1.2,0.1549631953239441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,32,power_law_1.01,0.05190399885177612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,32,power_law_1.01,0.04987519979476929
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,2,power_law_1.01,0.433241605758667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,2,power_law_1.2,0.19137279987335204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,32,power_law_1.01,0.05074560046195984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,32,power_law_1.01,0.05503360033035278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,2,power_law_1.2,0.2474047899246216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,32,power_law_1.01,0.05278720259666443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,32,power_law_1.01,0.06601600050926208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,32,power_law_1.01,0.05802879929542541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,2,power_law_1.2,0.34640638828277587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,2,power_law_1.01,0.5482304096221924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,32,power_law_1.01,0.05888640284538269
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,32,power_law_1.01,0.06741759777069092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,2,power_law_1.2,0.489734411239624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,32,power_law_1.01,0.06313599944114685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,2,power_law_1.01,0.756390380859375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,2,power_law_1.2,0.6050047874450684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,32,power_law_1.01,0.06731520295143127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,32,power_law_1.01,0.07301120162010193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,32,power_law_1.01,0.07975040078163147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,32,power_law_1.01,0.06988160014152527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,2,power_law_1.2,0.9289407730102539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,32,power_law_1.01,0.09618560075759888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,2,power_law_1.01,0.9817215919494628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,32,power_law_1.01,0.0692031979560852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,2,power_law_1.2,1.0878591537475586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,32,power_law_1.01,0.10984959602355956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,32,power_law_1.01,0.07039999961853027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,32,power_law_1.01,0.1275264024734497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,2,power_law_1.01,1.4101823806762694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,2,power_law_1.2,1.3368448257446288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,32,power_law_1.01,0.07120640277862549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,32,power_law_1.01,0.17120640277862548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,32,power_law_1.01,0.07035520076751708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,32,power_law_1.01,0.2083967924118042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,32,power_law_1.01,0.07407360076904297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,2,power_law_1.2,2.130995178222656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,32,power_law_1.01,0.2581183910369873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,2,power_law_1.01,1.7489856719970702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,32,power_law_1.01,0.08227840065956116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,32,power_law_1.01,0.373907208442688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,32,power_law_1.01,0.08173440098762512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,32,power_law_1.01,0.43751678466796873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,2,power_law_1.01,2.1902847290039062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,32,power_law_1.01,0.08421760201454162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,2,power_law_1.2,4.821388626098633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,32,power_law_1.01,0.09537280201911927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,32,power_law_1.01,0.5227456092834473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,32,power_law_1.01,0.10126080513000488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,32,power_law_1.01,0.11487360000610351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,2,power_law_1.01,3.4560897827148436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,32,power_law_1.01,0.9312383651733398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,32,power_law_1.01,0.1332479953765869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,32,power_law_1.01,0.16149760484695436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,32,power_law_1.01,1.6133184432983398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,32,power_law_1.01,0.19454079866409302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,32,power_law_1.01,0.2616127967834473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,2,power_law_1.01,6.629856109619141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,32,power_law_1.01,0.3044800043106079
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,32,power_law_1.01,0.45815677642822267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,32,power_law_1.01,0.5514624118804932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,32,power_law_1.01,0.7196864128112793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,8,power_law_1.01,0.04568960070610047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,32,power_law_1.01,0.9554752349853516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,8,power_law_1.01,0.05430399775505066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,8,power_law_1.01,0.044377601146698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,8,power_law_1.01,0.04579200148582459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,32,1,power_law_1.2,0.034835198521614076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,32,power_law_1.01,2.0929983139038084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,8,power_law_1.01,0.04919680058956146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,32,1,power_law_1.2,0.03959679901599884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,32,1,power_law_1.2,0.03646079897880554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,32,1,power_law_1.2,0.053260797262191774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,32,1,power_law_1.2,0.05175679922103882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,32,1,power_law_1.2,0.05330560207366943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,8,power_law_1.01,0.055283200740814206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,32,1,power_law_1.2,0.0535103976726532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,32,1,power_law_1.2,0.05580800175666809
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,32,1,power_law_1.2,0.05925120115280151
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,32,1,power_law_1.2,0.058713597059249875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,32,1,power_law_1.2,0.05976960062980652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,8,power_law_1.01,0.05966079831123352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,32,1,power_law_1.2,0.06007680296897888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,32,1,power_law_1.2,0.06115840077400207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,32,1,power_law_1.2,0.06593919992446899
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,32,1,power_law_1.2,0.07677440047264099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,32,2,power_law_1.01,0.03758080005645752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,32,1,power_law_1.2,0.07780479788780212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,8,power_law_1.01,0.060729598999023436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,32,1,power_law_1.2,0.07968000173568726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,32,1,power_law_1.2,0.08972160220146179
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,32,1,power_law_1.2,0.10605440139770508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,32,1,power_law_1.2,0.1307711958885193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,8,power_law_1.01,0.06170880198478699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,32,1,power_law_1.2,0.13868160247802735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,32,2,power_law_1.01,0.0331712007522583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,32,1,power_law_1.2,0.18426239490509033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,32,1,power_law_1.2,0.2183295965194702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,32,1,power_law_1.2,0.30120320320129396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,8,power_law_1.01,0.06229119896888733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,32,1,power_law_1.2,0.379967999458313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,32,2,power_law_1.01,0.0350847989320755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,32,1,power_law_1.2,0.5441664218902588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,8,power_law_1.01,0.06325119733810425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,32,1,power_law_1.2,0.6834496021270752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,32,2,power_law_1.01,0.03772160112857818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,32,1,power_law_1.2,0.8466303825378418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,8,power_law_1.01,0.06759039759635925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,32,1,power_law_1.2,1.3096511840820313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,32,2,power_law_1.01,0.040031999349594116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,8,power_law_1.01,0.06970880031585694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,32,2,power_law_1.01,0.049446401000022885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,32,1,power_law_1.2,2.566694450378418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,8,power_law_1.01,0.07600640058517456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,32,2,power_law_1.01,0.05082880258560181
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,8,power_law_1.01,0.08450559973716736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,32,2,power_law_1.01,0.05175039768218994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,8,power_law_1.01,0.08725759983062745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,32,2,power_law_1.01,0.052076798677444455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,8,power_law_1.01,0.09463040232658386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,32,2,power_law_1.01,0.05245440006256104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,4,balanced,0.031712000568707786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,32,2,power_law_1.01,0.05544319748878479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,4,balanced,0.031370667119820915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,8,power_law_1.01,0.10902400016784668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,4,balanced,0.03143466760714849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,4,balanced,0.035402665535608925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,4,balanced,0.04823466638724009
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,4,balanced,0.058362667759259544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,4,balanced,0.05866666634877523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,4,balanced,0.05902933577696482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,4,balanced,0.05875733494758606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,4,balanced,0.05871999760468801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,4,balanced,0.06234666705131531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,4,balanced,0.06080000102519989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,4,balanced,0.06002133091290792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,4,balanced,0.06267733375231425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,4,balanced,0.06769600013891856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,4,balanced,0.07196799914042155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,32,2,power_law_1.01,0.06051200032234192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,4,balanced,0.06981866558392842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,4,balanced,0.07309333483378093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,4,balanced,0.07656000057856242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,8,power_law_1.01,0.12591359615325928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,4,balanced,0.09149866302808125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,4,balanced,0.09658132990201314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,4,balanced,0.11914666493733723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,4,balanced,0.13942399621009827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,4,balanced,0.19074134031931558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,4,balanced,0.20586133003234863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,4,balanced,0.29366934299468994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,4,balanced,0.35488001505533856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,32,2,power_law_1.01,0.060192000865936277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,2,4,balanced,0.43104533354441327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,2,4,balanced,0.6467519998550415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,8,power_law_1.01,0.15273599624633788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,2,4,balanced,1.2160853544871013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,32,2,power_law_1.01,0.06390399932861328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,8,power_law_1.01,0.17778559923171997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,32,2,power_law_1.01,0.0742143988609314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,32,2,power_law_1.01,0.07727360129356384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,8,power_law_1.01,0.2266239881515503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,32,2,power_law_1.01,0.08056319952011108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,8,power_law_1.01,0.28707199096679686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,32,2,power_law_1.01,0.09239680171012879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,8,power_law_1.01,0.4013184070587158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,32,2,power_law_1.01,0.10859520435333252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,32,2,power_law_1.01,0.1325503945350647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,8,power_law_1.01,0.5338751792907714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,32,2,power_law_1.01,0.17144960165023804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,2,power_law_1.01,0.04874880015850067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,32,2,power_law_1.01,0.23020799160003663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,8,power_law_1.01,0.7052735805511474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,2,power_law_1.01,0.061504000425338747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,2,power_law_1.01,0.05807999968528747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,32,2,power_law_1.01,0.2694783926010132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,2,power_law_1.01,0.08554880023002624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,8,power_law_1.01,0.9508607864379883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,2,power_law_1.01,0.11230720281600952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,32,2,power_law_1.01,0.37166080474853513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,2,power_law_1.01,0.1584768056869507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,2,power_law_1.01,0.18206080198287963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,32,2,power_law_1.01,0.4614528179168701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,8,power_law_1.01,1.2709759712219237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,2,power_law_1.01,0.18307199478149414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,2,power_law_1.01,0.18405120372772216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,32,2,power_law_1.01,0.6742720127105712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,2,power_law_1.01,0.18851840496063232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,2,power_law_1.01,0.18708479404449463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,32,1,balanced,0.058186665177345276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,32,1,balanced,0.06204266846179962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,32,1,balanced,0.0705386648575465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,32,1,balanced,0.09427199761072795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,32,1,balanced,0.13668266932169595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,32,1,balanced,0.2109760046005249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,8,power_law_1.01,1.8130880355834962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,32,2,power_law_1.01,0.8360320091247558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,32,1,balanced,0.2821600039800008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,2,power_law_1.01,0.20236799716949463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,32,1,balanced,0.2836266756057739
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,32,1,balanced,0.2849173347155253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,32,1,balanced,0.2903199990590413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,32,1,balanced,0.2927093307177226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,32,1,balanced,0.2998186747233073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,32,1,balanced,0.30006933212280273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,2,power_law_1.01,0.20010879039764404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,32,1,balanced,0.3079093297322591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,32,1,balanced,0.3140160044034322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,32,1,balanced,0.3198346694310506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,32,1,balanced,0.3354880015055339
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,32,1,balanced,0.37515199184417725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,32,1,balanced,0.4019999901453654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,2,power_law_1.01,0.20254719257354736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,32,1,balanced,0.47324268023173016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,32,1,balanced,0.5302026669184366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,32,2,power_law_1.01,1.0453503608703614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,32,1,balanced,0.6604160070419312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,32,1,balanced,0.8636853694915771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,2,power_law_1.01,0.20862081050872802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,32,1,balanced,1.149509350458781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,32,1,balanced,1.491418679555257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,2,power_law_1.01,0.2135551929473877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,32,1,balanced,2.1119786898295083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,2,power_law_1.01,0.22561280727386473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,32,1,balanced,2.8201866149902344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,2,power_law_1.01,0.2530112028121948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,8,power_law_1.01,3.7620864868164063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,32,2,power_law_1.01,1.620140838623047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,32,1,balanced,3.4507840474446616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,2,power_law_1.01,0.25740799903869627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,2,power_law_1.01,0.2888319969177246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,32,1,balanced,5.263925234476726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,2,power_law_1.01,0.3441792011260986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,2,power_law_1.01,0.4116543769836426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,32,1,balanced,10.267791748046875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,32,2,power_law_1.01,3.3998016357421874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,2,power_law_1.01,0.4073472023010254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,2,power_law_1.01,0.5091775894165039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,2,power_law_1.01,0.641267204284668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,2,power_law_1.01,0.8189056396484375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,2,power_law_1.01,0.9032704353332519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,1,2,power_law_1.01,1.138803195953369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,1,2,power_law_1.01,1.6054464340209962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,1,2,power_law_1.01,3.1781312942504885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,256,power_law_1.01,0.08279680013656616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,256,power_law_1.01,0.08379520177841186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,4,balanced,0.03358400116364161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,4,balanced,0.03500800083080927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,4,balanced,0.03694933404525121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,4,balanced,0.05188799897829691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,4,balanced,0.06994666655858357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,4,balanced,0.08286400139331818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,4,balanced,0.08378666639328003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,4,balanced,0.08458667000134786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,4,balanced,0.08455466230710347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,4,balanced,0.0849226713180542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,4,balanced,0.0844533344109853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,256,power_law_1.01,0.08002560138702393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,4,balanced,0.08489599823951721
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,4,balanced,0.08725333213806152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,4,balanced,0.08618666728337605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,4,balanced,0.09197333455085754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,4,balanced,0.09177066882451375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,4,balanced,0.09492266178131104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,4,balanced,0.10462933778762817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,4,balanced,0.10284800330797832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,4,balanced,0.12481600046157837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,4,balanced,0.12205333511034648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,256,power_law_1.01,0.07839360237121581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,4,balanced,0.1614346702893575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,4,balanced,0.16523733735084534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,4,balanced,0.2200160026550293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,4,balanced,0.242682675520579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,4,balanced,0.343727986017863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,4,balanced,0.40377600987752277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,1,4,balanced,0.49142932891845703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,32,power_law_1.01,0.10942720174789429
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,256,power_law_1.01,0.07898880243301391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,1,4,balanced,0.7268266677856445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,1,4,balanced,1.3658560117085774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,256,power_law_1.01,0.08105599880218506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,256,power_law_1.01,0.08149120211601257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,32,power_law_1.01,0.1030784010887146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,256,power_law_1.01,0.07950080037117005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,256,power_law_1.01,0.08046720027923585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,32,power_law_1.01,0.06584960222244263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,256,power_law_1.01,0.07908480167388916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,32,power_law_1.01,0.07033600211143494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,256,power_law_1.01,0.08098559975624084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,256,power_law_1.01,0.07972480058670044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,8,balanced,0.07817600170771281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,8,balanced,0.09494400024414062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,8,balanced,0.13527466853459677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,8,balanced,0.21971732378005981
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,8,balanced,0.3493386507034302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,8,balanced,0.44545066356658936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,8,balanced,0.4461973508199056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,32,power_law_1.01,0.08156800270080566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,8,balanced,0.4500053326288859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,8,balanced,0.4535253445307414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,256,power_law_1.01,0.07894399762153625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,8,balanced,0.45096532503763836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,8,balanced,0.45438400904337567
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,8,balanced,0.45600001017252606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,8,balanced,0.45897066593170166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,8,balanced,0.4597546656926473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,8,balanced,0.4667626619338989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,8,balanced,0.4693919817606608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,256,power_law_1.01,0.07728639841079712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,8,balanced,0.4769066572189331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,8,balanced,0.5696426630020142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,8,balanced,0.5110453367233276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,8,balanced,0.6293919881184896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,32,power_law_1.01,0.09688959717750549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,8,balanced,0.5761546691258749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,8,balanced,0.9644373257954916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,256,power_law_1.01,0.081004798412323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,8,balanced,0.6970667044321696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,8,balanced,1.0844799677530925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,8,balanced,1.0107146898905437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,256,power_law_1.01,0.07991679906845092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,8,balanced,1.632431983947754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,8,balanced,1.5186079343159993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,32,power_law_1.01,0.1007040023803711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,1,8,balanced,1.9831786155700684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,256,power_law_1.01,0.08238720297813415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,1,8,balanced,2.832207997639974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,256,power_law_1.01,0.09925119876861573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,32,power_law_1.01,0.09932159781455993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,1,8,balanced,5.3578236897786455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,256,power_law_1.01,0.09978240132331848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,256,power_law_1.01,0.1195520043373108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,32,power_law_1.01,0.1028223991394043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,256,power_law_1.01,0.12993919849395752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,16,power_law_1.01,0.08060799837112427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,32,power_law_1.01,0.10041600465774536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,256,power_law_1.01,0.18353279829025268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,256,power_law_1.01,0.20774400234222412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,32,power_law_1.01,0.10019199848175049
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,16,power_law_1.01,0.074726402759552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,256,power_law_1.01,0.28682239055633546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,32,power_law_1.01,0.10677759647369385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,256,power_law_1.01,0.3448832035064697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,16,power_law_1.01,0.08792960047721862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,32,power_law_1.01,0.10657919645309448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,256,power_law_1.01,0.5080639839172363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,32,power_law_1.01,0.11103359460830689
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,256,power_law_1.01,0.6610879898071289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,16,power_law_1.01,0.05367680191993714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,32,power_law_1.01,0.11808639764785767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,256,power_law_1.01,0.8524031639099121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,16,power_law_1.01,0.056543999910354616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,32,4,balanced,0.03431999931732813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,32,4,balanced,0.03373866776625315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,32,4,balanced,0.03395200024048487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,32,4,balanced,0.033887999753157295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,32,4,balanced,0.033439998825391136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,32,4,balanced,0.033914667864640556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,32,4,balanced,0.03580799947182337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,32,4,balanced,0.03403199960788091
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,32,4,balanced,0.03610666592915853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,32,4,balanced,0.03540800015131632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,32,4,balanced,0.03643200049797694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,32,4,balanced,0.03772799919048945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,32,power_law_1.01,0.12121599912643433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,32,4,balanced,0.037861332297325134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,32,4,balanced,0.038005332152048744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,32,4,balanced,0.04285866518815359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,32,4,balanced,0.04417066772778829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,32,4,balanced,0.044213334719340004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,32,4,balanced,0.051370665431022644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,32,4,balanced,0.05554133156935374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,32,4,balanced,0.06490666667620341
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,32,4,balanced,0.08009600142637889
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,32,4,balanced,0.11642666657765706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,32,4,balanced,0.1432266632715861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,32,4,balanced,0.19249600172042847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,256,power_law_1.01,1.2397824287414552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,32,4,balanced,0.23834667603174844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,32,4,balanced,0.3375733296076457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,32,4,balanced,0.4200479984283447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,32,4,balanced,0.5177866617838541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,16,power_law_1.01,0.06689280271530151
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,32,4,balanced,0.7905759811401367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,32,power_law_1.01,0.13043199777603148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,32,4,balanced,1.5260319709777832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,16,power_law_1.01,0.06889600157737732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,32,power_law_1.01,0.15393279790878295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,256,power_law_1.01,2.616864013671875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,32,power_law_1.01,0.15839999914169312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,16,power_law_1.01,0.07823359966278076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,32,power_law_1.01,0.19790719747543334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,16,power_law_1.01,0.07752320170402527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,32,power_law_1.01,0.23118081092834472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,16,power_law_1.01,0.07831680178642272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,32,power_law_1.01,0.2859839916229248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,16,power_law_1.01,0.08227199912071229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,32,power_law_1.01,0.2881088018417358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,16,power_law_1.01,0.08455039858818054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,16,power_law_1.01,0.08533120155334473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,4,power_law_1.2,0.05060480237007141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,32,power_law_1.01,0.4200704097747803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,4,power_law_1.2,0.047788798809051514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,16,power_law_1.01,0.08861439824104309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,4,power_law_1.2,0.04867199957370758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,4,power_law_1.2,0.05315200090408325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,32,power_law_1.01,0.47829761505126955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,4,power_law_1.2,0.05743359923362732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,16,power_law_1.01,0.0970624029636383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,4,power_law_1.2,0.07933440208435058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,4,power_law_1.01,0.08760960102081299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,4,power_law_1.2,0.08099200129508972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,4,power_law_1.2,0.08427519798278808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,16,power_law_1.01,0.09983360171318054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,32,power_law_1.01,0.6629119873046875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,4,power_law_1.01,0.13418879508972167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,4,power_law_1.2,0.08630399703979492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,4,power_law_1.2,0.08554880023002624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,4,power_law_1.2,0.0895359992980957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,16,power_law_1.01,0.1064255952835083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,4,power_law_1.01,0.08355200290679932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,4,power_law_1.2,0.09133440256118774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,4,power_law_1.2,0.09503999948501587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,32,power_law_1.01,0.7418623924255371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,16,power_law_1.01,0.11442559957504272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,4,power_law_1.01,0.10197759866714477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,4,power_law_1.2,0.09742720127105713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,4,power_law_1.2,0.11003520488739013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,4,power_law_1.2,0.11144959926605225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,16,power_law_1.01,0.1290112018585205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,4,power_law_1.01,0.12579840421676636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,4,power_law_1.2,0.11628799438476563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,4,power_law_1.2,0.1306048035621643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,1,32,power_law_1.01,0.927302360534668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,16,power_law_1.01,0.1480512022972107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,4,power_law_1.01,0.17140480279922485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,4,power_law_1.2,0.14529279470443726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,4,power_law_1.2,0.17689599990844726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,4,power_law_1.2,0.22048640251159668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,4,power_law_1.01,0.17953920364379883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,16,power_law_1.01,0.16030720472335816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,4,power_law_1.2,0.27143039703369143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,4,power_law_1.2,0.327622389793396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,4,power_law_1.01,0.18819839954376222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,16,power_law_1.01,0.20554239749908448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,4,power_law_1.2,0.46167678833007814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,1,32,power_law_1.01,1.428377628326416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,4,power_law_1.01,0.18967679738998414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,4,power_law_1.2,0.5684607982635498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,16,power_law_1.01,0.23732481002807618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,4,power_law_1.2,0.8332927703857422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,4,power_law_1.01,0.18914560079574586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,4,power_law_1.2,1.255769634246826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,16,power_law_1.01,0.30871679782867434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,4,power_law_1.01,0.20149118900299073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,4,power_law_1.2,1.3783167839050292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,4,power_law_1.01,0.21171839237213136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,16,power_law_1.01,0.3964479923248291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,4,power_law_1.2,2.7070655822753906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,4,power_law_1.01,0.2134399890899658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,1,32,power_law_1.01,2.935424041748047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,16,power_law_1.01,0.5263872146606445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,4,power_law_1.01,0.2185983896255493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,16,power_law_1.01,0.6826176166534423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,4,power_law_1.2,4.3924415588378904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,4,power_law_1.01,0.23291521072387694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,4,power_law_1.01,0.24188799858093263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,16,power_law_1.01,0.7964928150177002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,4,power_law_1.01,0.2585664033889771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,16,power_law_1.01,1.2284992218017579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,4,power_law_1.01,0.3020672082901001
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,4,power_law_1.01,0.33958399295806885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,4,power_law_1.01,0.4278463840484619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,16,power_law_1.01,2.412019157409668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,4,power_law_1.01,0.47763838768005373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,2,power_law_1.2,0.032025599479675294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,2,power_law_1.2,0.034995201230049136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,2,power_law_1.2,0.03959679901599884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,2,power_law_1.2,0.04447360038757324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,4,power_law_1.01,0.6299647808074951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,1,2,power_law_1.01,0.0830847978591919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,1,2,power_law_1.01,0.08391680121421814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,2,power_law_1.2,0.05470079779624939
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,1,2,power_law_1.01,0.10187519788742065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,1,2,power_law_1.01,0.13719680309295654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,2,power_law_1.2,0.0777728021144867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,1,2,power_law_1.01,0.13841279745101928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,1,2,power_law_1.01,0.14147839546203614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,2,power_law_1.2,0.0802623987197876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,1,2,power_law_1.01,0.14846080541610718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,1,2,power_law_1.01,0.1526080012321472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,1,2,power_law_1.01,0.15356160402297975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,4,power_law_1.01,0.7287487983703613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,2,power_law_1.2,0.08447359800338745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,1,2,power_law_1.01,0.15475840568542482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,1,2,power_law_1.01,0.1571776032447815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,1,2,power_law_1.01,0.1649791955947876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,2,power_law_1.2,0.08729599714279175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,1,2,power_law_1.01,0.1657472014427185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,1,2,power_law_1.01,0.18394240140914916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,2,power_law_1.2,0.08942720293998718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,1,2,power_law_1.01,0.19265279769897461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,1,2,power_law_1.01,0.20039680004119872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,1,2,power_law_1.01,0.2410048007965088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,2,power_law_1.2,0.09192320108413696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,1,2,power_law_1.01,0.2589119911193848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,1,2,power_law_1.01,0.2905600070953369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,2,power_law_1.2,0.09708799719810486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,4,power_law_1.01,0.9709440231323242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,1,2,power_law_1.01,0.37816319465637205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,1,2,power_law_1.01,0.4731135845184326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,2,power_law_1.2,0.09593600034713745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,1,2,power_law_1.01,0.5899903774261475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,1,2,power_law_1.01,0.7593599796295166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,2,power_law_1.2,0.10117759704589843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,1,2,power_law_1.01,1.237996768951416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,2,power_law_1.2,0.10954240560531617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,1,2,power_law_1.01,1.5097536087036132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,2,power_law_1.2,0.11744639873504639
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,4,power_law_1.01,1.2639167785644532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,2,power_law_1.2,0.12135679721832275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,1,2,power_law_1.01,2.2944704055786134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,2,power_law_1.2,0.14176000356674195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,1,2,power_law_1.01,3.1790847778320312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,2,power_law_1.2,0.1605504035949707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,4,1,power_law_1.01,0.057920002937316896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,2,power_law_1.2,0.20015358924865723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,1,2,power_law_1.01,3.3509567260742186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,4,1,power_law_1.01,0.07230079770088196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,2,power_law_1.2,0.2453887939453125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,4,1,power_law_1.01,0.0857151985168457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,4,power_law_1.01,1.7483200073242187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,4,1,power_law_1.01,0.1307136058807373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,2,power_law_1.2,0.30734078884124755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,4,1,power_law_1.01,0.13112319707870485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,2,power_law_1.2,0.3615744113922119
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,1,2,power_law_1.01,5.829740905761719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,4,1,power_law_1.01,0.1320255994796753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,4,1,power_law_1.01,0.13341439962387086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,4,1,power_law_1.01,0.13665920495986938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,2,power_law_1.2,0.5338240146636963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,4,1,power_law_1.01,0.13914239406585693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,4,1,power_law_1.01,0.14143359661102295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,2,power_law_1.2,0.7075200080871582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,4,1,power_law_1.01,0.1416767954826355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,4,1,power_law_1.01,0.1434880018234253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,4,1,power_law_1.01,0.14778879880905152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,4,power_law_1.01,2.3233280181884766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,1,2,power_law_1.01,9.489900970458985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,2,power_law_1.2,0.9897407531738281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,4,1,power_law_1.01,0.1557760000228882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,4,1,power_law_1.01,0.1657472014427185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,4,1,power_law_1.01,0.17964799404144288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,4,1,power_law_1.01,0.1793984055519104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,2,power_law_1.2,1.223910427093506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,4,1,power_law_1.01,0.2129983901977539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,4,1,power_law_1.01,0.22379519939422607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,4,1,power_law_1.01,0.2675136089324951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,4,1,power_law_1.01,0.315775990486145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,2,power_law_1.2,1.6460351943969727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,4,1,power_law_1.01,0.4103551864624023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,4,1,power_law_1.01,0.5030208110809327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,4,1,power_law_1.01,0.6894336223602295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,2,4,power_law_1.01,2.848089599609375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,4,1,power_law_1.01,0.8743616104125976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,2,power_law_1.2,2.4656511306762696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,4,1,power_law_1.01,1.2680768013000487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,4,1,power_law_1.01,1.6305599212646484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,4,1,power_law_1.01,2.038559913635254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,4,1,power_law_1.01,3.11081600189209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,2,power_law_1.2,5.0977214813232425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,2,4,power_law_1.01,4.301311874389649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,4,1,power_law_1.01,6.0627391815185545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,16,power_law_1.01,0.04069760143756866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,16,power_law_1.01,0.030803200602531434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,16,power_law_1.01,0.02961919903755188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,16,power_law_1.01,0.030579200387001036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,16,power_law_1.01,0.030937600135803222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,16,power_law_1.01,0.03237119913101196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,16,power_law_1.01,0.03229439854621887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,16,power_law_1.01,0.03281919956207276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,16,power_law_1.01,0.033632001280784606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,16,power_law_1.01,0.0342848002910614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,2,4,power_law_1.01,8.206681823730468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,16,power_law_1.01,0.03524479866027832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,16,power_law_1.01,0.03758080005645752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,16,power_law_1.01,0.0384768009185791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,16,power_law_1.01,0.038412800431251524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,2,balanced,0.07012266914049785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,2,balanced,0.12311466534932454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,16,power_law_1.01,0.0434688001871109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,2,balanced,0.14427199959754944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,2,balanced,0.24288000663121542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,2,balanced,0.43489599227905273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,2,balanced,0.8263839880625407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,16,power_law_1.01,0.045126399397850035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,2,balanced,0.8276159763336182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,16,2,balanced,0.03722666700681051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,2,balanced,0.8299787044525146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,16,2,balanced,0.03922666609287262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,16,2,balanced,0.054917335510253906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,2,balanced,0.8283306757609049
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,16,2,balanced,0.05537599821885427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,16,power_law_1.01,0.046438398957252505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,2,balanced,0.8346186478932699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,16,2,balanced,0.05681066711743673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,16,2,balanced,0.05594133337338766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,2,balanced,0.8380640347798666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,16,2,balanced,0.0576853354771932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,16,2,balanced,0.056143999099731445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,2,balanced,0.8396693070729574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,16,2,balanced,0.05606399973233541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,16,2,balanced,0.05754133562246958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,2,balanced,0.8446346918741862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,16,power_law_1.01,0.05100799798965454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,16,2,balanced,0.059077332417170204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,16,2,balanced,0.06172266602516174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,2,balanced,0.8506879806518555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,16,2,balanced,0.05965333183606466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,16,2,balanced,0.061343997716903687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,16,2,balanced,0.06940799951553345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,2,balanced,0.8569280306498209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,16,2,balanced,0.07010666529337566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,16,2,balanced,0.07468266785144806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,2,balanced,0.86516801516215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,16,power_law_1.01,0.055308800935745236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,16,2,balanced,0.08015466729799907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,16,2,balanced,0.08538132905960083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,2,balanced,0.8802560170491537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,16,2,balanced,0.10169066985448201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,16,2,balanced,0.1251359979311625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,2,balanced,0.9014506340026855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,16,2,balanced,0.15651200215021768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,16,2,balanced,0.1744426687558492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,2,balanced,0.9302559693654379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,16,power_law_1.01,0.061587202548980716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,16,2,balanced,0.2334666649500529
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,2,balanced,0.9758346875508627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,16,2,balanced,0.28936533133188885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,16,2,balanced,0.40506664911905926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,2,balanced,1.0197439988454182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,16,2,balanced,0.4931306838989258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,2,balanced,1.107103983561198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,16,power_law_1.01,0.07360640168190002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,16,2,balanced,0.6125706831614176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,2,balanced,1.164410670598348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,16,2,balanced,0.9327733516693115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,2,balanced,1.4503733317057292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,16,power_law_1.01,0.1009600043296814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,16,2,balanced,1.7788267135620117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,2,balanced,1.5100159645080566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,2,balanced,2.231541315714518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,16,power_law_1.01,0.11292159557342529
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,2,balanced,2.3889546394348145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,16,power_law_1.01,0.14999040365219116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,2,balanced,3.349381446838379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,16,power_law_1.01,0.19680639505386352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,2,balanced,4.457621256510417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,16,power_law_1.01,0.27870080471038816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,2,balanced,8.700848261515299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,16,power_law_1.01,0.3624576091766357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,16,power_law_1.01,0.4133312225341797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,2,power_law_1.2,0.04563199877738953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,2,power_law_1.2,0.050918400287628174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,16,power_law_1.01,0.6185279846191406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,2,power_law_1.2,0.05430399775505066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,2,power_law_1.2,0.06159359812736511
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,16,power_law_1.01,1.3057024002075195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,2,power_law_1.2,0.07281919717788696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,2,power_law_1.2,0.08629119992256165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,2,power_law_1.2,0.10732159614562989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,16,power_law_1.01,0.05457280278205871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,2,power_law_1.2,0.11489280462265014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,16,power_law_1.01,0.06062080264091492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,2,power_law_1.2,0.11900160312652588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,2,power_law_1.2,0.12245759963989258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,16,power_law_1.01,0.049132800102233885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,2,power_law_1.2,0.1282304048538208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,16,power_law_1.01,0.050944000482559204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,2,power_law_1.2,0.13481600284576417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,64,balanced,0.055717334151268005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,64,balanced,0.05840000013510386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,64,balanced,0.058101331194241844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,64,balanced,0.05994666616121928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,64,balanced,0.06838933130105336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,64,balanced,0.0676693320274353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,64,balanced,0.067221333583196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,64,balanced,0.06659733255704244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,64,balanced,0.06865600248177846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,64,balanced,0.06727466483910878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,64,balanced,0.0684799998998642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,64,balanced,0.07281066477298737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,64,balanced,0.07097066442171733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,64,balanced,0.07658133407433827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,64,balanced,0.07887466748555501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,2,power_law_1.2,0.14131200313568115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,64,balanced,0.08029866715272267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,64,balanced,0.0888159970442454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,64,balanced,0.09406933188438416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,64,balanced,0.10539733370145161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,64,balanced,0.13354667027791342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,64,balanced,0.151936004559199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,16,power_law_1.01,0.05199360251426697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,64,balanced,0.1895786722501119
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,64,balanced,0.1879733403523763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,64,balanced,0.2406933307647705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,64,balanced,0.2653813362121582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,2,power_law_1.2,0.15211520195007325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,64,balanced,0.35469333330790204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,64,balanced,0.41950400670369464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,1,64,balanced,0.5097440083821615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,1,64,balanced,0.7433119614919027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,16,power_law_1.01,0.061504000425338747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,2,power_law_1.2,0.16858880519866942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,1,64,balanced,1.5871680577596028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,2,power_law_1.2,0.1765887975692749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,16,power_law_1.01,0.06180480122566223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,2,power_law_1.2,0.1717120051383972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,16,power_law_1.01,0.062080001831054686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,2,power_law_1.2,0.2061824083328247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,1,balanced,0.1301653285821279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,1,balanced,0.19074134031931558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,1,balanced,0.32870399951934814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,1,balanced,0.5967413187026978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,16,power_law_1.01,0.06160640120506287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,1,balanced,1.121285359064738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,2,power_law_1.2,0.23503360748291016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,1,balanced,1.625696023305257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,1,balanced,1.633530616760254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,16,power_law_1.01,0.06275839805603027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,1,balanced,1.6439627011617024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,1,balanced,1.6453653971354167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,2,power_law_1.2,0.28695681095123293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,1,balanced,1.6521546045939128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,1,balanced,1.6660159428914387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,16,power_law_1.01,0.06281599998474122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,1,balanced,1.6721812884012859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,2,power_law_1.2,0.33887999057769774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,1,balanced,1.6854186058044434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,1,balanced,1.700544039408366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,16,power_law_1.01,0.06674559712409973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,1,balanced,1.714255968729655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,2,power_law_1.2,0.4486015796661377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,1,balanced,1.7291679382324219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,1,balanced,1.7641919453938801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,16,power_law_1.01,0.0664192020893097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,1,balanced,2.08351469039917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,2,power_law_1.2,0.5571392059326172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,1,balanced,1.8729813893636067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,16,power_law_1.01,0.07335039973258972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,1,balanced,2.362885316212972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,1,balanced,2.1039199829101562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,2,power_law_1.2,0.7518655776977539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,16,power_law_1.01,0.07797120213508606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,1,balanced,3.3731091817220054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,1,balanced,2.5506614049275718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,16,power_law_1.01,0.0810368001461029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,2,power_law_1.2,0.987059211730957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,1,balanced,3.9444265365600586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,16,power_law_1.01,0.08251519799232483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,1,balanced,3.8878825505574546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,2,power_law_1.2,1.3902912139892578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,4,8,power_law_1.01,0.033792001008987424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,16,power_law_1.01,0.09939839839935302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,4,8,power_law_1.01,0.034796801209449765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,1,balanced,6.03874142964681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,4,8,power_law_1.01,0.03309440016746521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,4,8,power_law_1.01,0.03365119993686676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,4,8,power_law_1.01,0.033766400814056394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,16,power_law_1.01,0.09347839951515198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,4,8,power_law_1.01,0.035257598757743834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,4,8,power_law_1.01,0.0395904004573822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,1,balanced,6.389386494954427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,4,8,power_law_1.01,0.041407999396324155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,2,power_law_1.2,1.7982336044311524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,4,8,power_law_1.01,0.045407998561859134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,4,8,power_law_1.01,0.045875200629234315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,16,power_law_1.01,0.11006720066070556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,4,8,power_law_1.01,0.043987199664115906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,4,8,power_law_1.01,0.0469760000705719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,4,8,power_law_1.01,0.0530239999294281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,4,8,power_law_1.01,0.05626239776611328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,4,8,power_law_1.01,0.06513280272483826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,1,1,balanced,8.073621114095053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,16,power_law_1.01,0.1335935950279236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,4,8,power_law_1.01,0.06715520024299622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,4,8,power_law_1.01,0.08412799835205079
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,2,power_law_1.2,2.1480512619018555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,4,8,power_law_1.01,0.09878399968147278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,4,8,power_law_1.01,0.10472320318222046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,4,8,power_law_1.01,0.1330623984336853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,16,power_law_1.01,0.15873279571533203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,4,8,power_law_1.01,0.1699072003364563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,4,8,power_law_1.01,0.23669760227203368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,4,8,power_law_1.01,0.28570239543914794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,4,8,power_law_1.01,0.353600001335144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,1,1,balanced,11.746074676513672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,16,power_law_1.01,0.18776960372924806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,4,8,power_law_1.01,0.4989952087402344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,4,8,power_law_1.01,0.6722879886627198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,2,power_law_1.2,3.4550975799560546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,4,8,power_law_1.01,0.9904512405395508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,16,power_law_1.01,0.2800704002380371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,4,8,power_law_1.01,1.1721728324890137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,16,power_law_1.01,0.3312959909439087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,4,8,power_law_1.01,1.8080896377563476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,16,power_law_1.01,0.43584637641906737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,4,8,power_law_1.01,3.4008510589599608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,1,1,balanced,23.196673075358074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,2,power_law_1.2,6.7169921875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,16,power_law_1.01,0.5904575824737549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,16,power_law_1.01,0.687340784072876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,16,power_law_1.01,1.1493247985839843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,16,power_law_1.01,1.871129608154297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,2,balanced,0.062362665931383766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,2,balanced,0.07515733440717061
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,2,balanced,0.09120532870292664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,2,balanced,0.1262453297773997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,2,balanced,0.18634132544199625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,2,balanced,0.31538132826487225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,2,balanced,0.3153066635131836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,2,balanced,0.31946667035420734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,2,balanced,0.31700799862543744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,8,balanced,0.050106664498647056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,2,balanced,0.32156266768773395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,8,balanced,0.0543039987484614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,2,balanced,0.3210879961649577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,2,balanced,0.33026132980982464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,8,balanced,0.05211733281612396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,8,balanced,0.05589333176612854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,2,balanced,0.32993600765864056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,8,balanced,0.062080000837643944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,8,balanced,0.07260266443093617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,2,balanced,0.333626667658488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,8,balanced,0.07343466579914093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,8,balanced,0.07413333157698314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,2,balanced,0.34723198413848877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,8,balanced,0.07445866862932841
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,8,balanced,0.07680533329645793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,2,balanced,0.3517119884490967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,8,balanced,0.07460799813270569
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,8,balanced,0.07855466504891713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,2,balanced,0.36240533987681073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,8,balanced,0.07842133442560832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,8,balanced,0.07871466875076294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,2,balanced,0.3904800017674764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,8,balanced,0.08675199747085571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,8,balanced,0.0869599978129069
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,2,balanced,0.41491198539733887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,8,balanced,0.0909493366877238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,8,balanced,0.10106666882832845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,8,power_law_1.2,0.06895999908447266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,2,balanced,0.468506654103597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,8,balanced,0.10919466614723206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,8,balanced,0.13121599952379862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,2,balanced,0.510474681854248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,8,balanced,0.1543359955151876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,8,balanced,0.1887306571006775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,2,balanced,0.5948746601740519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,8,balanced,0.21698667605717978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,2,balanced,0.6773653030395508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,8,balanced,0.30343466997146606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,8,balanced,0.3678773244222005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,2,balanced,0.9332160154978434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,8,balanced,0.5236533482869467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,2,balanced,1.0929653644561768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,8,balanced,0.6772159735361735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,8,balanced,0.8310986359914144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,2,balanced,1.5694079399108887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,8,balanced,1.2757759888966878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,2,balanced,1.956208070119222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,8,power_law_1.2,0.0632960021495819
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,8,balanced,2.502117315928141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,2,balanced,2.5029919942220054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,2,balanced,3.6806131998697915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,8,power_law_1.2,0.05754240155220032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,256,balanced,0.08056533336639404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,256,balanced,0.08144000172615051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,2,balanced,7.132282892862956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,256,balanced,0.08285333216190338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,256,balanced,0.08385066191355388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,256,balanced,0.08190399905045827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,256,balanced,0.08410132924715678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,256,balanced,0.08341333270072937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,256,balanced,0.08408533533414204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,256,balanced,0.08388800422350566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,256,balanced,0.08542933066685994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,256,balanced,0.08589866757392883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,256,balanced,0.0872373382250468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,8,power_law_1.2,0.0656063973903656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,256,balanced,0.08532800277074178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,256,balanced,0.08583999673525493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,256,balanced,0.08955732981363933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,256,balanced,0.09205333391825359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,256,balanced,0.09316266576449077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,256,balanced,0.09794666369756062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,256,balanced,0.10160533587137859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,256,balanced,0.10980266332626343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,256,balanced,0.11753066380818684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,256,balanced,0.1402666668097178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,256,balanced,0.15552533666292825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,256,balanced,0.19246399402618408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,256,balanced,0.22954134146372476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,256,balanced,0.30719467004140216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,256,balanced,0.3776266574859619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,8,power_law_1.2,0.0763584017753601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,256,balanced,0.464789350827535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,256,balanced,0.684554656346639
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,256,balanced,1.3350507418314617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,8,power_law_1.2,0.0956928014755249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,8,power_law_1.2,0.09799039959907532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,8,power_law_1.2,0.1017024040222168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,8,power_law_1.2,0.10354559421539307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,64,balanced,0.0584799995024999
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,64,balanced,0.04634666442871094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,64,balanced,0.045738667249679565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,64,balanced,0.0458133320013682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,64,balanced,0.047824000318845115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,64,balanced,0.05120000243186951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,64,balanced,0.06863999863465627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,8,power_law_1.2,0.10643839836120605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,64,balanced,0.06660800178845723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,64,balanced,0.06874133149782817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,64,balanced,0.06904533505439758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,64,balanced,0.06901866694291432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,64,balanced,0.07077333331108093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,64,balanced,0.07029333213965099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,64,balanced,0.07247466842333476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,64,balanced,0.07868800063927968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,64,balanced,0.07796266674995422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,64,balanced,0.08273600041866302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,64,balanced,0.08715732892354329
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,64,balanced,0.09097599983215332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,64,balanced,0.10466667016347249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,64,balanced,0.11344533165295918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,64,balanced,0.13551466663678488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,64,balanced,0.15893333156903586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,64,balanced,0.20348799228668213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,64,balanced,0.2441706657409668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,64,balanced,0.34307201703389484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,64,balanced,0.4155786832173665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,128,balanced,0.05892266829808553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,64,balanced,0.5149173339207967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,8,power_law_1.2,0.10754560232162476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,128,balanced,0.06108266611893972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,128,balanced,0.060506666700045265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,128,balanced,0.05995733539263407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,64,balanced,0.7559306621551514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,128,balanced,0.058559998869895935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,128,balanced,0.06638933221499126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,128,balanced,0.07689066727956136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,128,balanced,0.07261333366235097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,128,balanced,0.07150933146476746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,64,balanced,1.470154603322347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,128,balanced,0.07213866710662842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,128,balanced,0.07361066838105519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,128,balanced,0.07606400052706401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,128,balanced,0.07482133309046428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,128,balanced,0.07474133372306824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,128,balanced,0.09262933333714803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,128,balanced,0.08309866487979889
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,128,balanced,0.09169600407282512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,128,balanced,0.09691733121871948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,128,balanced,0.10619733730951945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,128,balanced,0.12251200278600057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,128,balanced,0.13682132959365845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,128,balanced,0.16962667306264242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,128,balanced,0.19542932510375977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,8,power_law_1.2,0.11166080236434936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,128,balanced,0.2555626630783081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,128,balanced,0.32316799958546955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,128,balanced,0.4378453493118286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,128,balanced,0.556384007136027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,128,balanced,0.6808053652445475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,128,balanced,1.0434719721476238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,32,1,power_law_1.2,0.03840000033378601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,128,balanced,2.0424426396687827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,32,1,power_law_1.2,0.038047999143600464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,8,power_law_1.2,0.11568000316619872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,32,1,power_law_1.2,0.037862399220466615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,32,1,power_law_1.2,0.04156799912452698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,32,1,power_law_1.2,0.041254401206970215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,32,1,power_law_1.2,0.042259201407432556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,8,power_law_1.2,0.11843199729919433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,32,1,power_law_1.2,0.04307200014591217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,32,1,power_law_1.2,0.04300160109996796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,32,1,power_law_1.2,0.04411520063877106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,32,1,power_law_1.2,0.043424001336097716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,32,1,power_law_1.2,0.04447360038757324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,32,1,power_law_1.2,0.046316799521446225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,8,power_law_1.2,0.12946560382843017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,32,1,power_law_1.2,0.048518401384353635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,32,1,power_law_1.2,0.048588800430297854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,32,1,power_law_1.2,0.05618559718132019
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,32,1,power_law_1.2,0.05921919941902161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,32,1,power_law_1.2,0.06503679752349853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,32,1,power_law_1.2,0.07544320225715637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,8,power_law_1.2,0.13311359882354737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,32,1,power_law_1.2,0.09119359850883484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,32,1,power_law_1.2,0.10841599702835084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,32,1,power_law_1.2,0.12090239524841309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,32,1,power_law_1.2,0.16206719875335693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,32,1,power_law_1.2,0.19342080354690552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,32,1,power_law_1.2,0.27017600536346437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,8,power_law_1.2,0.1397055983543396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,32,1,power_law_1.2,0.3433664083480835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,32,1,power_law_1.2,0.49685120582580566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,32,1,power_law_1.2,0.6425280094146728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,32,1,power_law_1.2,0.7904640197753906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,8,power_law_1.2,0.15881600379943847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,32,1,power_law_1.2,1.226854419708252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,1,power_law_1.2,0.07761920094490052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,8,power_law_1.2,0.17739520072937012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,32,1,power_law_1.2,2.381747245788574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,1,power_law_1.2,0.08346239924430847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,32,2,power_law_1.2,0.03425920009613037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,32,2,power_law_1.2,0.034944000840187076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,8,power_law_1.2,0.22041599750518798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,32,2,power_law_1.2,0.0336896002292633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,1,power_law_1.2,0.09141759872436524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,32,2,power_law_1.2,0.033292800188064575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,32,2,power_law_1.2,0.034272000193595886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,32,2,power_law_1.2,0.03438720107078552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,32,2,power_law_1.2,0.03503359854221344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,32,2,power_law_1.2,0.03653759956359863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,1,power_law_1.2,0.10810240507125854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,32,2,power_law_1.2,0.036620798707008365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,32,2,power_law_1.2,0.03694080114364624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,32,2,power_law_1.2,0.038099199533462524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,32,2,power_law_1.2,0.03847039937973022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,8,power_law_1.2,0.2610431909561157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,32,2,power_law_1.2,0.03906559944152832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,1,power_law_1.2,0.13295359611511232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,32,2,power_law_1.2,0.04081920087337494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,32,2,power_law_1.2,0.04579200148582459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,32,2,power_law_1.2,0.047513601183891294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,32,2,power_law_1.2,0.05013759732246399
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,32,2,power_law_1.2,0.057183998823165896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,1,power_law_1.2,0.16327680349349977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,32,2,power_law_1.2,0.06293759942054748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,32,2,power_law_1.2,0.07294719815254211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,32,2,power_law_1.2,0.08535680174827576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,32,2,power_law_1.2,0.11310080289840699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,8,power_law_1.2,0.3356735944747925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,1,power_law_1.2,0.20855040550231935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,32,2,power_law_1.2,0.13453439474105836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,32,2,power_law_1.2,0.2006592035293579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,32,2,power_law_1.2,0.24106879234313966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,32,2,power_law_1.2,0.36499199867248533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,1,power_law_1.2,0.22387199401855468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,32,2,power_law_1.2,0.4755648136138916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,32,2,power_law_1.2,0.5397056102752685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,8,power_law_1.2,0.37354240417480467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,1,power_law_1.2,0.22696321010589598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,32,2,power_law_1.2,0.8658687591552734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,1,power_law_1.2,0.2304447889328003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,32,2,power_law_1.2,1.6465984344482423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,1,power_law_1.2,0.24368638992309571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,8,power_law_1.2,0.5111999988555909
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,1,power_law_1.2,0.2533567905426025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,1,power_law_1.2,0.26201601028442384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,8,power_law_1.2,0.6479296207427978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,8,4,power_law_1.2,0.039001598954200745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,1,power_law_1.2,0.26334080696105955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,8,4,power_law_1.2,0.040780800580978396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,8,4,power_law_1.2,0.0364544004201889
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,1,power_law_1.2,0.28443520069122313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,8,4,power_law_1.2,0.03565439879894257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,8,4,power_law_1.2,0.035820800065994265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,8,4,power_law_1.2,0.03902080059051514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,1,power_law_1.2,0.29569919109344484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,8,4,power_law_1.2,0.04261119961738587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,8,4,power_law_1.2,0.04366079866886139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,8,power_law_1.2,0.8915264129638671
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,1,power_law_1.2,0.3163840055465698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,8,4,power_law_1.2,0.045926401019096376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,8,4,power_law_1.2,0.044563201069831845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,8,4,power_law_1.2,0.04490880072116852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,1,power_law_1.2,0.3639359951019287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,8,4,power_law_1.2,0.04837760031223297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,8,4,power_law_1.2,0.05198079943656921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,8,4,power_law_1.2,0.05246719717979431
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,8,4,power_law_1.2,0.0657920002937317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,1,power_law_1.2,0.400710391998291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,8,4,power_law_1.2,0.07463039755821228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,8,4,power_law_1.2,0.07379199862480164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,8,power_law_1.2,1.1433343887329102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,8,4,power_law_1.2,0.08696960210800171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,1,power_law_1.2,0.4994624137878418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,8,4,power_law_1.2,0.10546560287475586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,8,4,power_law_1.2,0.1295807957649231
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,8,4,power_law_1.2,0.16123520135879515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,1,power_law_1.2,0.5758336067199707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,8,4,power_law_1.2,0.2094208002090454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,8,4,power_law_1.2,0.2563584089279175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,8,4,power_law_1.2,0.3646591901779175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,1,power_law_1.2,0.7526591777801513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,16,balanced,0.03152533372243246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,16,balanced,0.03268799930810928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,16,balanced,0.033359999457995095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,16,balanced,0.035162667433420815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,16,balanced,0.056799997886021934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,8,4,power_law_1.2,0.506828784942627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,16,balanced,0.0580266664425532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,16,balanced,0.05967999994754791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,16,balanced,0.059450666109720864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,16,balanced,0.05860800047715505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,16,balanced,0.06003733476003011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,16,balanced,0.059114664793014526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,16,balanced,0.06078400214513143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,16,balanced,0.06218666831652323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,16,balanced,0.06200533111890157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,16,balanced,0.06713599960009257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,8,4,power_law_1.2,0.6868991851806641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,16,balanced,0.06690133114655812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,16,balanced,0.069882666071256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,16,balanced,0.07450133562088013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,16,balanced,0.07963733375072479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,16,balanced,0.08974933624267578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,16,balanced,0.09709866841634114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,4,8,power_law_1.2,1.4142463684082032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,1,power_law_1.2,0.9202624320983886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,16,balanced,0.12013866504033406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,16,balanced,0.13620799779891968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,16,balanced,0.17613333463668823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,8,4,power_law_1.2,0.7419648170471191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,16,balanced,0.21608533461888632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,16,balanced,0.29711467027664185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,16,balanced,0.3712426821390788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,16,balanced,0.4498666524887085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,16,balanced,0.6754773457845052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,8,4,power_law_1.2,1.123904037475586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,16,balanced,1.2789279619852703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,1,power_law_1.2,1.261945629119873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,8,4,power_law_1.2,1.6022720336914062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,1,power_law_1.2,1.6076416015625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,8,4,power_law_1.2,3.9049854278564453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,1,power_law_1.2,2.287545585632324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,4,8,power_law_1.2,2.1810752868652346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,1,power_law_1.2,3.0086912155151366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,1,power_law_1.2,3.7010112762451173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,1,power_law_1.2,5.708172988891602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,4,8,power_law_1.2,4.871398544311523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,32,2,balanced,0.0480320006608963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,32,2,balanced,0.05193066596984863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,32,2,balanced,0.053930665055910744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,32,2,balanced,0.06243200103441874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,32,2,balanced,0.09070932865142822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,32,2,balanced,0.13351999719937643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,32,2,balanced,0.137882669766744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,32,2,balanced,0.132042666276296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,32,2,balanced,0.13286933302879333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,32,2,balanced,0.13286933302879333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,32,2,balanced,0.13636799653371176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,32,2,balanced,0.13863999644915262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,32,2,balanced,0.14087999860445657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,32,2,balanced,0.14282133181889853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,32,2,balanced,0.15064000089963278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,32,2,balanced,0.1530506710211436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,32,2,balanced,0.16113066673278809
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,32,2,balanced,0.1800640026728312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,32,2,balanced,0.1962453325589498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,32,2,balanced,0.23375999927520752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,32,2,balanced,0.26971733570098877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,32,2,balanced,0.3749973376592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,32,2,balanced,0.4292373259862264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,32,2,balanced,0.6201440095901489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,32,2,balanced,0.7618666489919027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,2,2,balanced,0.029781334102153778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,2,2,balanced,0.03219733387231827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,32,2,balanced,1.1106133460998535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,2,2,balanced,0.0358240008354187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,2,2,balanced,0.05129600067933401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,2,2,balanced,0.08060800035794576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,2,2,balanced,0.07914666831493378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,32,2,balanced,1.403242588043213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,2,2,balanced,0.07878399888674419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,2,2,balanced,0.08029333253701527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,2,2,balanced,0.08098666866620381
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,2,2,balanced,0.07998399933179219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,2,2,balanced,0.08167999982833862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,32,2,balanced,1.7391146024068196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,2,2,balanced,0.08333333333333333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,2,2,balanced,0.08443199594815572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,2,2,balanced,0.08777599533398946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,2,2,balanced,0.09313066800435384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,2,2,balanced,0.09294399619102478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,2,2,balanced,0.10060800115267436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,2,2,balanced,0.1069599986076355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,32,2,balanced,2.712805430094401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,2,balanced,0.11098133524258931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,2,balanced,0.13005866607030234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,2,balanced,0.13991467157999674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,2,balanced,0.17708800236384073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,1,power_law_1.2,11.165177917480468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,2,balanced,0.19678932428359985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,2,balanced,0.27988266944885254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,2,balanced,0.347327987353007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,2,balanced,0.48002668221791583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,32,2,balanced,5.331290562947591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,2,balanced,0.6143733263015747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,2,balanced,0.7505653699239095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,2,balanced,1.1665600140889485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,2,balanced,2.2513599395751953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,8,1,balanced,0.040021332601706185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,8,1,balanced,0.04381866753101349
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,8,1,balanced,0.06533866624037425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,8,1,balanced,0.06418666740258534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,8,1,balanced,0.0656160016854604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,8,1,balanced,0.06673599779605865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,8,1,balanced,0.06678933401902516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,8,1,balanced,0.0658240020275116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,8,1,balanced,0.06770666440327962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,8,1,balanced,0.06800533334414165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,8,1,balanced,0.07032000025113423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,8,1,balanced,0.0698880006869634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,8,1,balanced,0.07181333502133687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,8,1,balanced,0.07936533292134602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,8,1,balanced,0.085125337044398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,8,1,balanced,0.08620267113049825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,8,1,balanced,0.09054399530092876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,8,1,balanced,0.10394133130709331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,8,1,balanced,0.11514133214950562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,8,1,balanced,0.14629333217938742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,8,1,balanced,0.16337066888809204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,8,1,balanced,0.21118932962417603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,8,1,balanced,0.2626346747080485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,8,1,balanced,0.36982933680216473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,8,1,balanced,0.47154664993286133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,8,1,balanced,0.6750986576080322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,8,1,balanced,0.8770453135172526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,8,1,balanced,1.0772266387939453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,8,1,balanced,1.6706026395161946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,8,1,balanced,3.266293207804362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,8,4,balanced,0.05223466455936432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,8,4,balanced,0.05397333204746246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,8,4,balanced,0.056159997979799904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,8,4,balanced,0.06213866670926412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,8,4,balanced,0.081386665503184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,8,4,balanced,0.09278399745623271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,1,4,balanced,0.13364799817403158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,8,4,balanced,0.09100799759229024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,8,4,balanced,0.09464533130327861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,1,4,balanced,0.18173867464065552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,8,4,balanced,0.0946613351504008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,1,4,balanced,0.21994666258494058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,8,4,balanced,0.09299733241399129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,8,4,balanced,0.09511466821034749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,1,4,balanced,0.21491199731826782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,8,4,balanced,0.09628267089525859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,1,4,balanced,0.2144213318824768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,8,4,balanced,0.09908800323804219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,1,4,balanced,0.2178879976272583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,8,4,balanced,0.10100799798965454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,1,4,balanced,0.21724265813827515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,8,4,balanced,0.10864532987276714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,1,4,balanced,0.21817066272099814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,8,4,balanced,0.10956266522407532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,1,4,balanced,0.2195733388264974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,8,4,balanced,0.11553066968917847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,1,4,balanced,0.21890666087468466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,8,4,balanced,0.12617599964141846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,1,4,balanced,0.2317919929822286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,8,4,balanced,0.13583999872207642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,1,4,balanced,0.22013866901397705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,16,power_law_1.2,0.0401856005191803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,8,4,balanced,0.16029866536458334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,1,4,balanced,0.22161600987116495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,8,4,balanced,0.1774186690648397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,1,4,balanced,0.2328959902127584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,8,4,balanced,0.24131733179092407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,1,4,balanced,0.2604106664657593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,8,4,balanced,0.27880533536275226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,1,4,balanced,0.26095465819040936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,1,4,balanced,0.2462666630744934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,8,4,balanced,0.40174933274586994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,1,4,balanced,0.3199733297030131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,8,4,balanced,0.5011680126190186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,1,4,balanced,0.2928853432337443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,1,4,balanced,0.41494933764139813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,8,4,balanced,0.7192533016204834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,1,4,balanced,0.40879468123118085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,1,4,balanced,0.5500320196151733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,8,4,balanced,0.9170400301615397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,16,power_law_1.2,0.03384959995746613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,1,4,balanced,0.6784586906433105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,8,4,balanced,1.1061866283416748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,1,4,balanced,0.9487466812133789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,1,4,balanced,1.2662933667500813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,8,4,balanced,1.6968533198038738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,1,4,balanced,1.8181440035502117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,16,power_law_1.2,0.03308799862861633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,8,4,balanced,3.311802546183268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,1,4,balanced,2.424784024556478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,16,power_law_1.2,0.03198719918727875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,1,4,balanced,3.0125226974487305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,16,power_law_1.2,0.03326080143451691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,1,4,balanced,4.851114590962728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,16,power_law_1.2,0.035283198952674864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,16,power_law_1.2,0.036268800497055054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,1,4,balanced,9.75162124633789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,16,power_law_1.2,0.035641598701477054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,16,power_law_1.2,0.03565439879894257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,16,power_law_1.2,0.036627200245857236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,16,power_law_1.2,0.03792639970779419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,16,power_law_1.2,0.03953920006752014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,16,power_law_1.2,0.041459199786186215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,16,power_law_1.2,0.04455040097236633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,8,power_law_1.01,0.08797439932823181
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,16,power_law_1.2,0.04818559885025024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,16,power_law_1.2,0.049516800045967105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,8,power_law_1.01,0.11495039463043213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,16,power_law_1.2,0.05118079781532288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,16,power_law_1.2,0.056601601839065555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,8,power_law_1.01,0.17148159742355346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,16,power_law_1.2,0.06444799900054932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,8,power_law_1.01,0.08181759715080261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,16,power_law_1.2,0.08092799782752991
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,1,1,power_law_1.2,0.08859519958496094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,1,1,power_law_1.2,0.14080640077590942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,1,1,power_law_1.2,0.15241600275039674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,16,power_law_1.2,0.09695360064506531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,1,1,power_law_1.2,0.22734079360961915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,1,1,power_law_1.2,0.23644158840179444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,1,1,power_law_1.2,0.24384639263153077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,8,power_law_1.01,0.10998400449752807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,1,1,power_law_1.2,0.2471872091293335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,16,power_law_1.2,0.1633792042732239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,1,1,power_law_1.2,0.24410879611968994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,1,1,power_law_1.2,0.245414400100708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,1,1,power_law_1.2,0.25667200088500974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,1,1,power_law_1.2,0.2529792070388794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,16,power_law_1.2,0.17503360509872437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,1,1,power_law_1.2,0.2724735975265503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,8,power_law_1.01,0.1356287956237793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,1,1,power_law_1.2,0.27888638973236085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,1,1,power_law_1.2,0.28523519039154055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,1,1,power_law_1.2,0.3090559959411621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,2,1,balanced,0.0827946662902832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,2,1,balanced,0.17786665757497153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,2,1,balanced,0.23541865746180216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,1,1,power_law_1.2,0.31458559036254885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,16,power_law_1.2,0.25148160457611085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,2,1,balanced,0.23477333784103394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,2,1,balanced,0.23624533414840698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,2,1,balanced,0.23907732963562012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,1,1,power_law_1.2,0.3292608022689819
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,2,1,balanced,0.2367146611213684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,2,1,balanced,0.23720532655715942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,2,1,balanced,0.23924267292022705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,1,1,power_law_1.2,0.3869312047958374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,2,1,balanced,0.23875200748443604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,8,power_law_1.01,0.14228479862213134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,2,1,balanced,0.24431999524434408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,2,1,balanced,0.24336000283559164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,2,1,balanced,0.24613332748413086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,1,1,power_law_1.2,0.44510722160339355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,2,1,balanced,0.2558346589406331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,2,1,balanced,0.2641013264656067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,2,1,balanced,0.2662453254063924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,1,1,power_law_1.2,0.550816011428833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,2,1,balanced,0.2932479977607727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,16,power_law_1.2,0.30064001083374026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,2,1,balanced,0.31574400266011554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,2,1,balanced,0.34144000212351483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,2,1,balanced,0.42955199877421063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,1,1,power_law_1.2,0.6534272193908691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,2,1,balanced,0.4784640073776245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,2,1,balanced,0.654581348101298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,1,1,power_law_1.2,0.8728384017944336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,2,1,balanced,0.7918720245361328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,8,power_law_1.01,0.18095999956130981
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,2,1,balanced,1.1627306938171387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,1,1,power_law_1.2,1.0928832054138184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,16,power_law_1.2,0.474073600769043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,2,1,balanced,1.5167039235432942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,1,1,power_law_1.2,1.495622444152832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,2,1,balanced,2.228922684987386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,2,1,balanced,2.9267040888468423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,8,power_law_1.01,0.1895807981491089
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,1,1,power_law_1.2,1.945631980895996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,16,power_law_1.2,0.6759488105773925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,2,1,balanced,3.6419893900553384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,1,1,power_law_1.2,2.8273408889770506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,8,power_law_1.01,0.1833024024963379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,2,1,balanced,5.823093414306641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,16,power_law_1.2,0.7569407939910888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,1,1,power_law_1.2,3.7149631500244142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,8,power_law_1.01,0.18624639511108398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,1,1,power_law_1.2,4.591468811035156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,2,1,balanced,11.511156717936197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,16,power_law_1.2,1.4213055610656737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,8,power_law_1.01,0.19611519575119019
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,1,1,power_law_1.2,7.400128173828125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,8,power_law_1.01,0.1894592046737671
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,8,power_law_1.01,0.20074241161346434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,16,power_law_1.2,2.881888008117676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,8,power_law_1.01,0.22159359455108643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,1,1,power_law_1.2,14.308447265625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,8,power_law_1.01,0.2232127904891968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,8,power_law_1.01,0.22595200538635254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,8,power_law_1.01,0.2511359930038452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,8,power_law_1.01,0.25552639961242674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,8,power_law_1.01,0.2932735919952393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,8,32,power_law_1.01,0.054604798555374146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,8,power_law_1.01,0.3409728050231934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,8,32,power_law_1.01,0.04863359928131104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,8,32,power_law_1.01,0.04762240052223206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,8,power_law_1.01,0.4246848106384277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,8,32,power_law_1.01,0.04775680005550385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,32,balanced,0.04347200194994608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,32,balanced,0.03338133295377096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,32,balanced,0.03349866718053818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,32,balanced,0.03270933280388514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,8,32,power_law_1.01,0.04836480021476745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,32,balanced,0.03311466674009959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,32,balanced,0.033359999457995095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,32,balanced,0.03374933451414108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,32,balanced,0.03499733408292135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,32,balanced,0.034688000877698265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,8,power_law_1.01,0.4832831859588623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,32,balanced,0.03422400106986364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,32,balanced,0.03477333237727483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,8,32,power_law_1.01,0.04993920028209686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,32,balanced,0.034741332133611046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,32,balanced,0.03509333233038584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,32,balanced,0.03676266719897588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,32,balanced,0.03802666564782461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,32,balanced,0.03850133220354716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,32,balanced,0.03932799895604452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,32,balanced,0.0408746674656868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,32,balanced,0.04081599911053976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,32,balanced,0.044682666659355164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,8,32,power_law_1.01,0.050323200225830075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,32,balanced,0.04540266593297323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,32,balanced,0.05267733335494995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,32,balanced,0.053077335158983864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,32,balanced,0.06637333333492279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,32,balanced,0.07619733115037282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,32,balanced,0.09269332885742188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,8,32,power_law_1.01,0.051609599590301515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,32,balanced,0.09621866544087727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,8,power_law_1.01,0.6245183944702148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,2,32,balanced,0.10938666264216106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,2,32,balanced,0.1532746652762095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,2,32,balanced,0.24461867411931357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,128,power_law_1.01,0.03255679905414581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,8,32,power_law_1.01,0.05156480073928833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,8,32,power_law_1.01,0.05195519924163818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,128,power_law_1.01,0.031276801228523256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,8,power_law_1.01,0.7341184139251709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,8,32,power_law_1.01,0.052697598934173584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,128,power_law_1.01,0.031865599751472476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,8,32,power_law_1.01,0.05428479909896851
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,128,power_law_1.01,0.03158400058746338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,8,32,power_law_1.01,0.055340802669525145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,128,power_law_1.01,0.03274880051612854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,8,power_law_1.01,1.0514304161071777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,128,power_law_1.01,0.033030399680137636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,8,32,power_law_1.01,0.05724160075187683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,128,power_law_1.01,0.033471998572349546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,8,32,power_law_1.01,0.06020479798316956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,128,power_law_1.01,0.03375999927520752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,8,32,power_law_1.01,0.06214399933815003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,128,power_law_1.01,0.034041601419448855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,8,power_law_1.01,1.2553536415100097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,8,32,power_law_1.01,0.06490880250930786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,128,power_law_1.01,0.033990401029586795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,128,power_law_1.01,0.03424000144004822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,8,32,power_law_1.01,0.07169280052185059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,128,power_law_1.01,0.03505919873714447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,8,32,power_law_1.01,0.08167679905891419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,128,power_law_1.01,0.03514240086078644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,8,power_law_1.01,1.4926719665527344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,8,32,power_law_1.01,0.09952639937400817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,128,power_law_1.01,0.036032000184059144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,8,32,power_law_1.01,0.11377919912338257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,128,power_law_1.01,0.03852800130844116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,128,power_law_1.01,0.03874559998512268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,8,32,power_law_1.01,0.1427072048187256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,128,power_law_1.01,0.03971199989318848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,8,32,power_law_1.01,0.16809600591659546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,128,power_law_1.01,0.041657599806785586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,8,power_law_1.01,2.267193603515625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,128,power_law_1.01,0.04412800073623657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,8,32,power_law_1.01,0.22830719947814943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,128,power_law_1.01,0.049132800102233885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,8,32,power_law_1.01,0.2903872013092041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,128,power_law_1.01,0.05219200253486633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,128,power_law_1.01,0.06110720038414001
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,8,32,power_law_1.01,0.39911038875579835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,128,power_law_1.01,0.06817920207977295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,8,power_law_1.2,0.07864320278167725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,128,power_law_1.01,0.10302720069885254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,8,32,power_law_1.01,0.528000020980835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,8,power_law_1.01,4.183526229858399
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,128,power_law_1.01,0.12932480573654176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,8,power_law_1.2,0.054073601961135864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,8,32,power_law_1.01,0.6459648132324218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,128,power_law_1.01,0.17916799783706666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,8,power_law_1.2,0.061766397953033444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,128,power_law_1.01,0.21941759586334228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,8,32,power_law_1.01,0.9702719688415528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,8,power_law_1.2,0.07552000284194946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,128,power_law_1.01,0.2763200044631958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,8,power_law_1.2,0.07850239872932434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,128,power_law_1.01,0.41753602027893066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,8,32,power_law_1.01,1.8267648696899415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,8,power_law_1.2,0.10093439817428589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,8,power_law_1.2,0.10634880065917969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,128,power_law_1.01,0.9023167610168457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,8,power_law_1.2,0.10955519676208496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,8,power_law_1.2,0.11063040494918823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,8,power_law_1.2,0.11334400177001953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,8,power_law_1.2,0.11200640201568604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,8,power_law_1.2,0.11696640253067017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,8,power_law_1.2,0.11330560445785523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,2,balanced,0.0995093286037445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,8,power_law_1.2,0.12021119594573974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,2,balanced,0.1925440033276876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,2,balanced,0.24157333374023438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,2,balanced,0.43613334496816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,2,balanced,0.8234133720397949
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,2,balanced,1.603941281636556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,8,power_law_1.2,0.12460160255432129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,2,balanced,2.377621332804362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,2,balanced,2.370954672495524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,8,power_law_1.2,0.131769597530365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,2,balanced,2.3781439463297525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,2,balanced,2.3813759485880532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,8,power_law_1.2,0.13624320030212403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,2,balanced,2.382453282674154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,2,balanced,2.388373374938965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,8,power_law_1.2,0.15037440061569213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,2,balanced,2.390757401784261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,2,balanced,2.4139572779337564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,8,power_law_1.2,0.1696895956993103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,2,balanced,2.4249866803487143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,2,balanced,2.437023957570394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,8,power_law_1.2,0.21710081100463868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,2,balanced,2.444906711578369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,1,2,power_law_1.01,0.09608960151672363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,2,balanced,2.480069319407145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,8,power_law_1.2,0.2184448003768921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,1,2,power_law_1.01,0.12766720056533815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,1,2,power_law_1.01,0.12540160417556762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,2,balanced,2.509402592976888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,1,2,power_law_1.01,0.14610559940338136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,1,2,power_law_1.01,0.1462656021118164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,1,2,power_law_1.01,0.14851839542388917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,2,balanced,2.5686240196228027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,8,power_law_1.2,0.3000960111618042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,1,2,power_law_1.01,0.14929920434951782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,1,2,power_law_1.01,0.15044480562210083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,2,balanced,2.629968007405599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,1,2,power_law_1.01,0.15296640396118164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,1,2,power_law_1.01,0.1524672031402588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,2,balanced,2.758821487426758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,1,2,power_law_1.01,0.15705599784851074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,8,power_law_1.2,0.39555840492248534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,1,2,power_law_1.01,0.16094720363616943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,1,2,power_law_1.01,0.1669376015663147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,2,balanced,2.8541332880655923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,1,2,power_law_1.01,0.17011200189590453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,1,2,power_law_1.01,0.18763519525527955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,2,balanced,3.0370826721191406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,1,2,power_law_1.01,0.19560960531234742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,8,power_law_1.2,0.4651968002319336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,1,2,power_law_1.01,0.21262080669403077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,1,2,power_law_1.01,0.26150400638580323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,2,balanced,3.3198719024658203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,1,2,power_law_1.01,0.28993279933929444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,1,2,power_law_1.01,0.37059841156005857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,1,2,power_law_1.01,0.4607679843902588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,2,balanced,3.6751734415690103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,8,power_law_1.2,0.5842944145202636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,1,2,power_law_1.01,0.6432127952575684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,1,2,power_law_1.01,0.9142016410827637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,2,balanced,4.944954554239909
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,1,2,power_law_1.01,1.2169599533081055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,8,power_law_1.2,0.7191487789154053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,1,2,power_law_1.01,1.4584896087646484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,2,balanced,6.299274444580078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,1,2,power_law_1.01,2.2312448501586912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,8,power_law_1.2,0.9703359603881836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,1,2,power_law_1.01,2.5834815979003904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,2,balanced,7.652453104654948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,1,2,power_law_1.01,3.7757247924804687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,8,power_law_1.2,1.3085056304931642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,1,2,power_law_1.01,4.248479843139648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,2,balanced,14.902581532796225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,8,power_law_1.2,2.09486083984375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,1,2,power_law_1.01,10.678054046630859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,8,power_law_1.2,3.6133121490478515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,128,power_law_1.2,0.040006399154663086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,128,power_law_1.2,0.03919999897480011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,128,power_law_1.2,0.03880319893360138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,128,power_law_1.2,0.039878401160240176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,128,power_law_1.01,0.045535999536514285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,128,power_law_1.2,0.039136001467704774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,1,8,power_law_1.2,0.0816703975200653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,128,power_law_1.01,0.046540799736976626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,128,power_law_1.2,0.04044800102710724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,1,8,power_law_1.2,0.08481280207633972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,1,8,power_law_1.2,0.08222079873085023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,128,power_law_1.01,0.046387198567390445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,128,power_law_1.2,0.041094401478767396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,1,8,power_law_1.2,0.08321920037269592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,1,8,power_law_1.2,0.08204799890518188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,128,power_law_1.01,0.046163201332092285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,128,power_law_1.2,0.040633600950241086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,1,8,power_law_1.2,0.08920959830284118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,128,power_law_1.2,0.040729600191116336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,1,8,power_law_1.2,0.09562879800796509
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,128,power_law_1.01,0.04556800127029419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,1,8,power_law_1.2,0.09620479941368103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,128,power_law_1.2,0.04123519957065582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,128,power_law_1.01,0.047679999470710756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,1,8,power_law_1.2,0.09871360063552856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,128,power_law_1.2,0.04181120097637177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,128,power_law_1.01,0.04670720100402832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,1,8,power_law_1.2,0.10242559909820556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,128,power_law_1.2,0.04185599982738495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,128,power_law_1.01,0.045926401019096376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,1,8,power_law_1.2,0.10062079429626465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,128,power_law_1.2,0.04297600090503693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,128,power_law_1.01,0.04597119987010956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,1,8,power_law_1.2,0.10447360277175903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,1,8,power_law_1.2,0.10720000267028809
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,128,power_law_1.01,0.044889599084854126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,128,power_law_1.2,0.04340479969978332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,1,8,power_law_1.2,0.14083199501037597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,128,power_law_1.01,0.04469119906425476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,128,power_law_1.2,0.04734080135822296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,1,8,power_law_1.2,0.17519999742507936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,128,power_law_1.01,0.04384639859199524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,128,power_law_1.2,0.04794879853725433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,1,8,power_law_1.2,0.17753599882125853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,1,8,power_law_1.2,0.15648640394210817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,128,power_law_1.01,0.04350079894065857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,128,power_law_1.2,0.050406402349472045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,1,8,power_law_1.2,0.19328000545501708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,128,power_law_1.01,0.0434112012386322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,128,power_law_1.2,0.053600001335144046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,1,8,power_law_1.2,0.26610560417175294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,128,power_law_1.01,0.045407998561859134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,1,8,power_law_1.2,0.3320512056350708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,128,power_law_1.2,0.06276479959487916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,128,power_law_1.01,0.04922240078449249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,1,8,power_law_1.2,0.3781248092651367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,128,power_law_1.2,0.07223680019378662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,1,8,power_law_1.2,0.5887104034423828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,128,power_law_1.01,0.048307201266288756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,128,power_law_1.2,0.08183040022850037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,1,8,power_law_1.2,0.7888895988464355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,128,power_law_1.01,0.052799999713897705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,128,power_law_1.2,0.11226240396499634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,1,8,power_law_1.2,1.1860735893249512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,128,power_law_1.01,0.05751039981842041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,128,power_law_1.01,0.06809599995613098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,1,8,power_law_1.2,1.4397824287414551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,128,power_law_1.2,0.13133440017700196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,128,power_law_1.01,0.07867519855499268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,128,power_law_1.2,0.1994495987892151
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,1,8,power_law_1.2,2.240678405761719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,128,power_law_1.01,0.10585600137710571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,128,power_law_1.2,0.24194560050964356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,128,power_law_1.01,0.1217087984085083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,1,8,power_law_1.2,2.840716743469238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,128,power_law_1.2,0.3541248083114624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,128,power_law_1.01,0.1689087986946106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,1,8,power_law_1.2,3.4332542419433594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,128,power_law_1.01,0.19838720560073853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,128,power_law_1.2,0.49582719802856445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,128,power_law_1.01,0.3177023887634277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,128,power_law_1.2,0.5609663963317871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,1,8,power_law_1.2,6.0321601867675785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,128,power_law_1.01,0.39521920680999756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,128,power_law_1.2,0.963584041595459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,128,power_law_1.01,0.5112576007843017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,128,power_law_1.01,0.8335935592651367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,1,8,power_law_1.2,11.969049835205078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,128,power_law_1.2,1.6928512573242187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,128,power_law_1.01,1.6836223602294922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,16,power_law_1.01,0.08972799777984619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,8,power_law_1.01,0.07071999907493591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,16,power_law_1.01,0.07265920042991639
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,8,1,power_law_1.2,0.052083200216293334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,16,power_law_1.01,0.03397760093212128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,8,1,power_law_1.2,0.07388160228729249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,8,1,power_law_1.2,0.07619199752807618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,8,1,power_law_1.2,0.11138559579849243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,16,power_law_1.01,0.03797119855880737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,8,1,power_law_1.2,0.11722240447998047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,8,1,power_law_1.2,0.11866879463195801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,8,1,power_law_1.2,0.11951999664306641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,16,power_law_1.01,0.04460160136222839
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,8,1,power_law_1.2,0.11971839666366577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,8,power_law_1.01,0.09832320213317872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,8,1,power_law_1.2,0.12341760396957398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,8,1,power_law_1.2,0.12114559412002564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,16,power_law_1.01,0.05383679866790771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,8,1,power_law_1.2,0.12558720111846924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,8,1,power_law_1.2,0.129203200340271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,8,1,power_law_1.2,0.1276927947998047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,16,power_law_1.01,0.05192959904670715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,8,1,power_law_1.2,0.13895039558410643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,8,1,power_law_1.2,0.15056639909744263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,8,1,power_law_1.2,0.1549056053161621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,16,power_law_1.01,0.05437440276145935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,8,1,power_law_1.2,0.1562432050704956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,8,1,power_law_1.2,0.1760319948196411
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,16,power_law_1.01,0.05392640233039856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,8,1,power_law_1.2,0.19600640535354613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,8,1,power_law_1.2,0.24257280826568603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,8,1,power_law_1.2,0.2826560020446777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,8,power_law_1.01,0.14167040586471558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,16,power_law_1.01,0.05264000296592712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,8,1,power_law_1.2,0.37528319358825685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,8,1,power_law_1.2,0.4435904026031494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,16,power_law_1.01,0.05316479802131653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,8,1,power_law_1.2,0.6055039882659912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,16,power_law_1.01,0.0541055977344513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,8,1,power_law_1.2,0.7862783908843994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,16,power_law_1.01,0.05455999970436096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,8,1,power_law_1.2,1.135308837890625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,8,power_law_1.01,0.07296640276908875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,16,power_law_1.01,0.0573311984539032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,8,1,power_law_1.2,1.406816005706787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,16,power_law_1.01,0.06335359811782837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,8,1,power_law_1.2,1.7305023193359375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,16,power_law_1.01,0.0639680027961731
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,16,power_law_1.01,0.06629120111465454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,8,1,power_law_1.2,2.681964874267578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,8,power_law_1.01,0.09126399755477906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,16,power_law_1.01,0.07149440050125122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,16,power_law_1.01,0.07200639843940734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,16,power_law_1.01,0.08126720190048217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,8,1,power_law_1.2,5.250188827514648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,16,power_law_1.01,0.09308800101280212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,8,power_law_1.01,0.11649919748306274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,16,power_law_1.01,0.09637119770050048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,16,power_law_1.01,0.12040959596633911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,16,power_law_1.01,0.14438400268554688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,8,power_law_1.01,0.1276352047920227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,16,power_law_1.01,0.1905344009399414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,16,power_law_1.01,0.26883199214935305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,8,power_law_1.01,0.15471999645233153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,16,power_law_1.01,0.29234559535980226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,1,16,power_law_1.01,0.3281599998474121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,8,power_law_1.01,0.1591231942176819
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,1,16,power_law_1.01,0.46599678993225097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,8,power_law_1.01,0.16104320287704468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,1,16,power_law_1.01,0.8089152336120605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,8,power_law_1.01,0.17039999961853028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,8,power_law_1.01,0.16560640335083007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,32,4,power_law_1.2,0.060185599327087405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,8,power_law_1.01,0.1772480010986328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,32,4,power_law_1.2,0.0652351975440979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,32,4,power_law_1.2,0.0679423987865448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,8,power_law_1.01,0.18325120210647583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,32,4,power_law_1.2,0.06528639793395996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,8,power_law_1.01,0.19423999786376953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,32,4,power_law_1.2,0.07154560089111328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,32,4,power_law_1.2,0.08164479732513427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,8,power_law_1.01,0.1997696042060852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,32,4,power_law_1.2,0.08821120262145996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,8,power_law_1.01,0.2085632085800171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,8,power_law_1.01,0.057606399059295654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,32,4,power_law_1.2,0.09904639720916748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,8,power_law_1.01,0.060601598024368285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,8,power_law_1.01,0.23149440288543702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,32,4,power_law_1.2,0.10268800258636475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,8,power_law_1.01,0.05299199819564819
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,8,power_law_1.01,0.05748479962348938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,32,4,power_law_1.2,0.10407040119171143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,8,power_law_1.01,0.24817280769348143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,8,power_law_1.01,0.05950080156326294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,8,power_law_1.01,0.07255039811134338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,32,4,power_law_1.2,0.10794880390167236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,8,power_law_1.01,0.07462400197982788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,8,power_law_1.01,0.3053312063217163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,8,power_law_1.01,0.07565439939498901
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,32,4,power_law_1.2,0.11274240016937256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,8,power_law_1.01,0.07710719704627991
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,8,power_law_1.01,0.07565439939498901
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,32,4,power_law_1.2,0.11574399471282959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,8,power_law_1.01,0.3240000009536743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,8,power_law_1.01,0.08097280263900757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,8,power_law_1.01,0.08126720190048217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,32,4,power_law_1.2,0.1176576018333435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,8,power_law_1.01,0.08170880079269409
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,8,power_law_1.01,0.4056960105895996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,8,power_law_1.01,0.08631680011749268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,32,4,power_law_1.2,0.1297152042388916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,8,power_law_1.01,0.0951807975769043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,8,power_law_1.01,0.095551997423172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,32,4,power_law_1.2,0.12978559732437134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,8,power_law_1.01,0.4701375961303711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,8,power_law_1.01,0.1030784010887146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,32,4,power_law_1.2,0.14302719831466676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,8,power_law_1.01,0.11989120244979859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,8,power_law_1.01,0.1287168025970459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,8,power_law_1.01,0.621452808380127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,32,4,power_law_1.2,0.1579200029373169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,8,power_law_1.01,0.15057920217514037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,8,power_law_1.01,0.17523839473724365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,32,4,power_law_1.2,0.18068480491638184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,8,power_law_1.01,0.7177216053009033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,8,power_law_1.01,0.21702399253845214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,32,4,power_law_1.2,0.2128448009490967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,8,power_law_1.01,0.28166399002075193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,8,power_law_1.01,0.3867136001586914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,32,4,power_law_1.2,0.24940800666809082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,8,power_law_1.01,1.0192576408386231
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,8,power_law_1.01,0.44080638885498047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,32,4,power_law_1.2,0.3241152048110962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,8,power_law_1.01,0.6559167861938476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,8,power_law_1.01,1.2638400077819825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,32,4,power_law_1.2,0.39797120094299315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,8,power_law_1.01,0.8191167831420898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,8,power_law_1.01,1.0286144256591796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,32,4,power_law_1.2,0.5559103965759278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,8,power_law_1.01,1.556499195098877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,32,4,power_law_1.2,0.6936192035675048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,8,power_law_1.01,1.6169536590576172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,32,4,power_law_1.2,1.0240320205688476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,8,power_law_1.01,2.313478469848633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,8,power_law_1.01,3.0506175994873046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,32,4,power_law_1.2,1.336409568786621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,32,4,power_law_1.2,1.6434688568115234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,8,power_law_1.01,4.593292617797852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,32,4,power_law_1.2,2.6005695343017576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,32,4,power_law_1.2,5.021952056884766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,16,power_law_1.2,0.11889920234680176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,16,power_law_1.2,0.09313279986381531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,16,power_law_1.2,0.07351040244102477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,2,power_law_1.01,0.07194240093231201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,16,power_law_1.2,0.08346239924430847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,16,power_law_1.2,0.09434880018234253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,2,power_law_1.01,0.09748479723930359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,16,power_law_1.2,0.12581119537353516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,2,power_law_1.2,0.06902400255203248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,2,power_law_1.01,0.12862720489501953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,2,power_law_1.2,0.09408000111579895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,16,power_law_1.2,0.13267199993133544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,2,power_law_1.2,0.09541760087013244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,16,power_law_1.2,0.14158719778060913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,2,power_law_1.01,0.11137919425964356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,2,power_law_1.2,0.11352959871292115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,16,power_law_1.2,0.1394368052482605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,2,power_law_1.2,0.17408000230789183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,2,power_law_1.01,0.16265599727630614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,16,power_law_1.2,0.1429952025413513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,2,power_law_1.2,0.24158079624176027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,2,power_law_1.2,0.2852224111557007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,16,power_law_1.2,0.1449023962020874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,2,power_law_1.01,0.21811840534210206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,2,power_law_1.2,0.3872895956039429
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,16,power_law_1.2,0.1517632007598877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,2,power_law_1.01,0.24174079895019532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,2,power_law_1.2,0.4094719886779785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,16,power_law_1.2,0.15155839920043945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,2,power_law_1.2,0.41578240394592286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,2,power_law_1.01,0.30634241104125975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,16,power_law_1.2,0.1549504041671753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,2,power_law_1.2,0.46771841049194335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,2,power_law_1.2,0.4768959999084473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,16,power_law_1.2,0.1641343951225281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,2,power_law_1.01,0.3125375986099243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,2,power_law_1.2,0.48691840171813966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,16,power_law_1.2,0.16684160232543946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,2,power_law_1.01,0.32193920612335203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,2,power_law_1.2,0.5150271892547608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,16,power_law_1.2,0.18317439556121826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,2,power_law_1.2,0.5350016117095947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,2,power_law_1.01,0.3354880094528198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,2,power_law_1.2,0.5539840221405029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,16,power_law_1.2,0.21629440784454346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,2,power_law_1.01,0.34683520793914796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,2,power_law_1.2,0.5906239986419678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,16,power_law_1.2,0.2218048095703125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,2,power_law_1.2,0.6342207908630371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,2,power_law_1.01,0.36956799030303955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,16,power_law_1.2,0.2666368007659912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,2,power_law_1.2,0.6941184043884278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,2,power_law_1.01,0.3807807922363281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,2,power_law_1.2,0.8242624282836915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,16,power_law_1.2,0.313100790977478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,2,power_law_1.01,0.3914175987243652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,2,power_law_1.2,0.86561279296875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,16,power_law_1.2,0.3894272089004517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,2,power_law_1.2,1.0237567901611329
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,2,power_law_1.01,0.4074560165405273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,16,power_law_1.2,0.40903677940368655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,2,power_law_1.2,1.1365887641906738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,2,power_law_1.01,0.4292031764984131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,2,power_law_1.2,1.457919979095459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,16,power_law_1.2,0.5011839866638184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,2,power_law_1.01,0.47443199157714844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,2,power_law_1.2,1.6266559600830077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,16,power_law_1.2,0.6998015880584717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,2,power_law_1.01,0.5203328132629395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,2,power_law_1.2,2.1316991806030274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,2,power_law_1.01,0.6015999794006348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,16,power_law_1.2,0.92740478515625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,2,power_law_1.2,2.695078468322754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,2,power_law_1.01,0.6393407821655274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,16,power_law_1.2,1.0375679969787597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,2,power_law_1.2,3.2007423400878907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,2,power_law_1.01,0.7751359939575195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,2,power_law_1.01,0.9206591606140136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,2,power_law_1.2,4.807545471191406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,2,16,power_law_1.2,1.4616064071655273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,2,power_law_1.01,1.1941120147705078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,2,power_law_1.01,1.4599040031433106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,2,power_law_1.2,8.538387298583984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,2,16,power_law_1.2,2.1848575592041017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,2,power_law_1.01,1.954572868347168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,2,power_law_1.01,2.5465215682983398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,2,power_law_1.01,3.0451967239379885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,2,16,power_law_1.2,4.464998245239258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,2,power_law_1.01,4.245587158203125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,32,2,power_law_1.01,0.03527039885520935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,32,2,power_law_1.01,0.034745600819587705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,32,2,power_law_1.01,0.0349375993013382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,32,2,power_law_1.01,0.03600000143051148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,32,2,power_law_1.01,0.036883199214935304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,32,2,power_law_1.01,0.036556801199913024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,32,2,power_law_1.01,0.03781760036945343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,32,2,power_law_1.01,0.03790720105171204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,32,2,power_law_1.01,0.03893760144710541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,32,2,power_law_1.01,0.039750400185585025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,32,2,power_law_1.01,0.03996799886226654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,32,2,power_law_1.01,0.041945600509643556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,32,2,power_law_1.01,0.04169600009918213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,32,2,power_law_1.01,0.04286719858646393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,32,2,power_law_1.01,0.048825600743293764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,2,power_law_1.01,8.379219055175781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,32,2,power_law_1.01,0.04937599897384644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,32,2,power_law_1.01,0.0542464017868042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,32,2,power_law_1.01,0.06440320014953613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,32,2,power_law_1.01,0.06787840127944947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,32,2,power_law_1.01,0.08547840118408204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,32,2,power_law_1.01,0.10146559476852417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,32,2,power_law_1.01,0.1272320032119751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,32,2,power_law_1.01,0.15463039875030518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,32,2,power_law_1.01,0.2150144100189209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,32,2,power_law_1.01,0.27781119346618655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,32,2,power_law_1.01,0.37837440967559816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,32,2,power_law_1.01,0.5138751983642578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,32,2,power_law_1.01,0.7381951808929443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,32,2,power_law_1.01,1.0199040412902831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,4,balanced,0.037802666425704956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,4,balanced,0.039503999054431915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,4,balanced,0.040192000567913055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,4,balanced,0.04763199885686239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,4,balanced,0.06274133423964183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,4,balanced,0.07400000095367432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,4,balanced,0.07446933289368947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,4,balanced,0.07317866881688435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,4,balanced,0.07398400207360585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,4,balanced,0.07438933352629344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,32,2,power_law_1.01,1.8327743530273437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,4,balanced,0.07453866799672444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,4,balanced,0.07554666697978973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,4,balanced,0.07645333309968312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,4,balanced,0.07841599980990092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,4,balanced,0.08149866759777069
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,4,balanced,0.08345066507657369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,4,balanced,0.08585600058237712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,4,balanced,0.09178133805592854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,4,balanced,0.09892266988754272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,4,balanced,0.12284266948699951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,4,balanced,0.12343466281890869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,4,balanced,0.1553600033124288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,4,balanced,0.17434666554133096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,4,balanced,0.2502293388048808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,4,balanced,0.2766559918721517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,4,balanced,0.38813332716623944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,4,balanced,0.4745013316472371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,2688,1856,6,128,1,4,balanced,0.5851733287175497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,2688,1856,6,128,1,4,balanced,0.8894080320994059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,2688,1856,6,128,1,4,balanced,1.7075093587239583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,32,power_law_1.2,0.045772799849510194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,32,power_law_1.2,0.037011200189590455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,32,power_law_1.2,0.03716480135917664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,32,power_law_1.2,0.03934719860553741
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,32,power_law_1.2,0.037536001205444335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,32,power_law_1.2,0.03996799886226654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,32,power_law_1.2,0.038873600959777835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,1,power_law_1.2,0.04917120039463043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,32,power_law_1.2,0.038771200180053714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,1,power_law_1.2,0.0553600013256073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,32,power_law_1.2,0.0389631986618042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,1,power_law_1.2,0.06825600266456604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,4,power_law_1.01,0.06578559875488281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,32,power_law_1.2,0.038815999031066896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,1,power_law_1.2,0.0923200011253357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,4,power_law_1.01,0.0856000006198883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,32,power_law_1.2,0.040454399585723874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,1,power_law_1.2,0.11747839450836181
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,4,power_law_1.01,0.07276800274848938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,32,power_law_1.2,0.04168959856033325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,1,power_law_1.2,0.14997119903564454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,32,power_law_1.2,0.04259200096130371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,4,power_law_1.01,0.08892800211906433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,32,power_law_1.2,0.04585599899291992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,1,power_law_1.2,0.20945920944213867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,4,power_law_1.01,0.10435839891433715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,32,power_law_1.2,0.05102720260620117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,1,power_law_1.2,0.21938560009002686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,4,power_law_1.01,0.11936639547348023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,32,power_law_1.2,0.052108800411224364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,1,power_law_1.2,0.23096320629119874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,4,power_law_1.01,0.14542720317840577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,32,power_law_1.2,0.05573760271072388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,1,power_law_1.2,0.24194560050964356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,4,power_law_1.01,0.1583295941352844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,32,power_law_1.2,0.06159999966621399
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,1,power_law_1.2,0.24669439792633058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,4,power_law_1.01,0.1571903944015503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,32,power_law_1.2,0.06679040193557739
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,1,power_law_1.2,0.2634943962097168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,4,power_law_1.01,0.16374399662017822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,32,power_law_1.2,0.0861952006816864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,1,power_law_1.2,0.278054404258728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,4,power_law_1.01,0.16343040466308595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,32,power_law_1.2,0.10321279764175414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,1,power_law_1.2,0.3001919984817505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,4,power_law_1.01,0.1748800039291382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,32,power_law_1.2,0.11245440244674683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,1,power_law_1.2,0.31351680755615235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,4,power_law_1.01,0.17306239604949952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,32,power_law_1.2,0.15007359981536866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,1,power_law_1.2,0.3357952117919922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,4,power_law_1.01,0.1831936001777649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,32,power_law_1.2,0.2072000026702881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,1,power_law_1.2,0.34513919353485106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,4,power_law_1.01,0.19009920358657836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,32,power_law_1.2,0.25523200035095217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,1,power_law_1.2,0.41425280570983886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,4,power_law_1.01,0.19540480375289918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,32,power_law_1.2,0.44232959747314454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,1,power_law_1.2,0.466431999206543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,4,power_law_1.01,0.20713601112365723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,32,power_law_1.2,0.4692800045013428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,1,power_law_1.2,0.563212776184082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,4,power_law_1.01,0.23030400276184082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,4,power_law_1.01,0.2563904047012329
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,1,power_law_1.2,0.6187776088714599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,32,power_law_1.2,0.7499711990356446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,4,power_law_1.01,0.3048448085784912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,1,power_law_1.2,0.7886464118957519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,4,power_law_1.01,0.32899200916290283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,32,power_law_1.2,1.010591983795166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,1,power_law_1.2,0.9813376426696777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,4,power_law_1.01,0.4069183826446533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,1,power_law_1.2,1.3137855529785156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,4,power_law_1.01,0.48312959671020506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,1,power_law_1.2,1.6510143280029297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,32,power_law_1.2,2.216998481750488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,4,power_law_1.01,0.6349631786346436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,4,power_law_1.01,0.7578239917755127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,1,power_law_1.2,2.3051328659057617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,4,power_law_1.01,1.042240047454834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,1,power_law_1.2,2.993235206604004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,4,power_law_1.01,1.390873622894287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,1,power_law_1.2,3.643199920654297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,4,power_law_1.01,1.7158527374267578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,4,power_law_1.01,2.565945625305176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,1,power_law_1.2,5.607046508789063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,4,power_law_1.01,5.321311950683594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,1,power_law_1.2,0.07274240255355835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,1,power_law_1.2,10.839949035644532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,1,power_law_1.2,0.09466239809989929
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,1,power_law_1.2,0.10668799877166749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,1,power_law_1.2,0.1747648000717163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,1,power_law_1.2,0.25727999210357666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,1,power_law_1.2,0.3516416072845459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,1,power_law_1.2,0.5539135932922363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,1,power_law_1.2,0.5725503921508789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,16,2,balanced,0.03583466758330663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,16,2,balanced,0.03771200031042099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,16,2,balanced,0.0397119993964831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,16,2,balanced,0.03825599948565165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,16,2,balanced,0.03928533444801966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,16,2,balanced,0.03976000100374222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,16,2,balanced,0.039808000127474465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,16,2,balanced,0.04154666761557261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,16,2,balanced,0.03980266551176707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,16,2,balanced,0.03982399900754293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,16,2,balanced,0.041946664452552795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,16,2,balanced,0.04193066557248434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,16,2,balanced,0.043968002001444496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,1,power_law_1.2,0.6009664058685302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,16,2,balanced,0.04390400151411692
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,16,2,balanced,0.050010666251182556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,16,2,balanced,0.05173333485921224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,32,2,balanced,0.06345599889755249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,16,2,balanced,0.05434666574001312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,16,2,balanced,0.06221333146095276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,32,2,balanced,0.0672106643517812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,16,2,balanced,0.06842133402824402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,32,2,balanced,0.07100800176461537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,16,2,balanced,0.0846506655216217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,32,2,balanced,0.08316266536712646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,16,2,balanced,0.09909866253534953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,32,2,balanced,0.10427733262379964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,16,2,balanced,0.1264906624952952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,32,2,balanced,0.14820266763369241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,16,2,balanced,0.14191466569900513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,32,2,balanced,0.1486026644706726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,16,2,balanced,0.1900320053100586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,32,2,balanced,0.1451520025730133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,16,2,balanced,0.2413333257039388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,32,2,balanced,0.14936533570289612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,32,2,balanced,0.1527733306090037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,16,2,balanced,0.3419413169225057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,32,2,balanced,0.15479466319084167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,32,2,balanced,0.15501866738001505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,16,2,balanced,0.4394186735153198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,32,2,balanced,0.1585653324921926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,32,2,balanced,0.16170666615168253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,16,2,balanced,0.5410879850387573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,32,2,balanced,0.17182934284210205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,32,2,balanced,0.17262399196624756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,1,power_law_1.2,0.652672004699707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,32,2,balanced,0.1800373395284017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,16,2,balanced,0.8265653451283773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,32,2,balanced,0.2002293268839518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,32,2,balanced,0.22054400046666464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,32,2,balanced,0.2586560050646464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,32,2,balanced,0.2921280066172282
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,16,2,balanced,1.5938293139139812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,32,2,balanced,0.3832319974899292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,32,2,balanced,0.46299731731414795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,32,2,balanced,0.6775893370310465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,32,2,balanced,0.8657279809316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,32,2,balanced,1.2531039714813232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,1,power_law_1.2,0.6724095821380616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,32,2,balanced,1.661834716796875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,32,2,balanced,2.0525919596354165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,1,power_law_1.2,0.7204736232757568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,32,2,balanced,3.1596800486246743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,32,2,balanced,6.152048110961914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,1,power_law_1.2,0.7225215911865235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,1,power_law_1.2,0.7598271846771241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,1,power_law_1.2,0.7926464080810547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,1,power_law_1.2,0.8002367973327636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,1,power_law_1.2,0.8743103981018067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,1,power_law_1.2,0.9229567527770997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,1,power_law_1.2,1.0051520347595215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,1,power_law_1.2,1.116870403289795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,1,power_law_1.01,0.03130879998207092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,1,power_law_1.01,0.03516800105571747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,1,power_law_1.01,0.0448063999414444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,1,power_law_1.2,1.232198429107666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,1,power_law_1.01,0.05987200140953064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,1,power_law_1.01,0.07430400252342224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,1,power_law_1.01,0.10252799987792968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,1,power_law_1.2,1.4606143951416015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,1,power_law_1.01,0.10913920402526855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,32,4,power_law_1.2,0.045542401075363156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,1,power_law_1.01,0.11248639822006226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,1,power_law_1.01,0.11724799871444702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,1,power_law_1.2,1.5602175712585449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,32,4,power_law_1.2,0.050470399856567386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,1,power_law_1.01,0.11873279809951783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,32,4,power_law_1.2,0.0461760014295578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,1,power_law_1.01,0.1191167950630188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,1,power_law_1.01,0.1272704005241394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,1,power_law_1.2,1.9687679290771485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,32,4,power_law_1.2,0.04919039905071258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,1,power_law_1.01,0.12924799919128419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,1,power_law_1.01,0.1307968020439148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,32,4,power_law_1.2,0.051948797702789304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,1,power_law_1.01,0.13761279582977295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,32,4,power_law_1.2,0.058796799182891844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,1,power_law_1.2,2.346598434448242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,1,power_law_1.01,0.14018559455871582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,1,power_law_1.01,0.14655359983444213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,32,4,power_law_1.2,0.06675840020179749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,1,power_law_1.01,0.16550400257110595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,32,4,power_law_1.2,0.07207679748535156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,1,power_law_1.01,0.18099839687347413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,1,power_law_1.2,3.1462400436401365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,1,power_law_1.01,0.21629440784454346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,32,4,power_law_1.2,0.07217280268669128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,1,power_law_1.01,0.23543040752410888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,32,4,power_law_1.2,0.07573760151863099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,1,power_law_1.01,0.2843264102935791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,1,power_law_1.01,0.3392575979232788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,32,4,power_law_1.2,0.0791167974472046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,1,power_law_1.2,3.9632640838623048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,1,power_law_1.01,0.4463935852050781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,32,4,power_law_1.2,0.08559359908103943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,1,power_law_1.01,0.5496640205383301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,32,4,power_law_1.2,0.08908799886703492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,1,power_law_1.01,0.7597760200500489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,32,4,power_law_1.2,0.09633920192718506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,1,power_law_1.2,4.834214401245117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,1,power_law_1.01,0.9785792350769043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,32,4,power_law_1.2,0.10671360492706299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,32,4,power_law_1.2,0.11079039573669433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,2,1,power_law_1.01,1.1780415534973145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,32,4,power_law_1.2,0.11392639875411988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,2,1,power_law_1.01,1.808915138244629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,32,4,power_law_1.2,0.13148159980773927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,1,power_law_1.2,6.786688232421875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,32,4,power_law_1.2,0.15891200304031372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,32,4,power_law_1.2,0.19345279932022094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,2,1,power_law_1.01,3.504390335083008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,32,4,power_law_1.2,0.22423040866851807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,32,4,power_law_1.2,0.2980031967163086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,32,4,power_law_1.2,0.3783616065979004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,1,power_law_1.2,12.863334655761719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,32,4,power_law_1.2,0.5324416160583496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,32,4,power_law_1.2,0.6768320083618165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,32,4,power_law_1.2,1.0469951629638672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,32,4,power_law_1.2,1.2938559532165528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,32,4,power_law_1.2,1.600057601928711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,32,4,power_law_1.2,2.4066240310668947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,1,power_law_1.01,0.1248128056526184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,32,4,power_law_1.2,4.764992141723633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,1,power_law_1.01,0.15179519653320311
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,32,power_law_1.01,0.051020801067352295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,1,power_law_1.01,0.16609280109405516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,32,power_law_1.01,0.057030397653579715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,1,power_law_1.01,0.2411263942718506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,32,power_law_1.01,0.04733439981937408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,1,power_law_1.01,0.31329920291900637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,32,power_law_1.01,0.04508799910545349
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,1,power_law_1.01,0.49669761657714845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,32,power_law_1.01,0.04757120013237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,1,power_law_1.01,0.542464017868042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,32,power_law_1.01,0.04929920136928558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,1,power_law_1.01,0.5507775783538819
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,32,power_law_1.01,0.04951040148735046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,1,4,power_law_1.2,0.08161919713020324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,1,4,power_law_1.2,0.10553599596023559
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,1,power_law_1.01,0.5784383773803711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,1,4,power_law_1.2,0.09687039852142335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,1,4,power_law_1.2,0.09585919976234436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,32,power_law_1.01,0.05061119794845581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,1,4,power_law_1.2,0.09749760031700135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,1,4,power_law_1.2,0.09722880125045777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,1,4,power_law_1.2,0.09780480265617371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,1,4,power_law_1.2,0.09887999892234803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,1,4,power_law_1.2,0.10101759433746338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,1,4,power_law_1.2,0.10165760517120362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,1,4,power_law_1.2,0.10341119766235352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,1,power_law_1.01,0.592851209640503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,32,power_law_1.01,0.05379199981689453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,1,4,power_law_1.2,0.11032960414886475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,1,4,power_law_1.2,0.11051520109176635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,1,4,power_law_1.2,0.1282688021659851
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,1,4,power_law_1.2,0.16003199815750122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,1,4,power_law_1.2,0.17466880083084108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,1,4,power_law_1.2,0.16051199436187744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,1,4,power_law_1.2,0.16436480283737182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,32,power_law_1.01,0.05130239725112915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,1,4,power_law_1.2,0.22987520694732666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,1,power_law_1.01,0.6200575828552246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,1,4,power_law_1.2,0.3315711975097656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,1,4,power_law_1.2,0.38934400081634524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,1,4,power_law_1.2,0.6652031898498535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,32,power_law_1.01,0.05316479802131653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,1,4,power_law_1.2,0.7460224151611328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,1,4,power_law_1.2,1.3368127822875977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,1,power_law_1.01,0.6358208179473877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,1,4,power_law_1.2,1.3160320281982423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,32,power_law_1.01,0.054476797580718994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,1,4,power_law_1.2,2.0539392471313476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,1,power_law_1.01,0.65414400100708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,32,power_law_1.01,0.05569919943809509
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,1,4,power_law_1.2,2.2383808135986327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,1,4,power_law_1.2,3.4035518646240233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,32,power_law_1.01,0.05888640284538269
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,1,power_law_1.01,0.6764480113983155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,32,power_law_1.01,0.06317440271377564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,1,4,power_law_1.2,4.995596694946289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,1,power_law_1.01,0.7097472190856934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,32,power_law_1.01,0.0655232012271881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,1,power_law_1.01,0.7540544033050537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,32,power_law_1.01,0.07141119837760926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,1,4,power_law_1.2,10.441580963134765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,32,power_law_1.01,0.08091520071029663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,1,power_law_1.01,0.8155072212219239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,32,power_law_1.01,0.09322239756584168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,1,power_law_1.01,0.9342016220092774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,32,power_law_1.01,0.11799039840698242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,32,power_law_1.01,0.13838080167770386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,1,power_law_1.01,1.0460415840148927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,32,power_law_1.01,0.18584320545196534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,1,power_law_1.01,1.2972543716430665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,32,power_law_1.01,0.21841280460357665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,1,power_law_1.01,1.4645824432373047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,32,power_law_1.01,0.2996608018875122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,32,power_law_1.01,0.3916800022125244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,1,power_law_1.01,1.9165184020996093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,16,balanced,0.04344533383846283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,16,balanced,0.04457066456476847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,16,balanced,0.04382933179537455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,16,balanced,0.045925334095954895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,16,balanced,0.04794133206208547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,16,balanced,0.05436799923578898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,16,balanced,0.05579733351866404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,16,balanced,0.05603733162085215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,16,balanced,0.057717333237330117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,16,balanced,0.0561653325955073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,16,balanced,0.058687999844551086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,16,balanced,0.060458665092786155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,16,balanced,0.06032533446947733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,32,power_law_1.01,0.5521279811859131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,16,balanced,0.06414933502674103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,16,balanced,0.06874666611353557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,16,balanced,0.07239466905593872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,16,balanced,0.07879466811815898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,16,balanced,0.09307199716567993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,16,balanced,0.10088533163070679
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,16,balanced,0.11987200379371643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,16,balanced,0.1462506651878357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,16,balanced,0.18116267522176108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,16,balanced,0.21234132846196493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,16,balanced,0.28666667143503827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,16,balanced,0.360426664352417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,16,balanced,0.5206240018208822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,16,balanced,0.6522133350372314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,16,balanced,0.8039893309275309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,1,power_law_1.01,2.322534370422363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,16,balanced,1.2239786783854167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,32,power_law_1.01,0.7058944225311279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,16,balanced,2.409759998321533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,32,power_law_1.01,0.841055965423584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,1,power_law_1.01,3.1671167373657227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,32,power_law_1.01,1.3461824417114259
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,1,power_law_1.01,4.058950424194336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,32,power_law_1.01,2.610233688354492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,1,power_law_1.01,5.819731140136719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,2,power_law_1.01,0.03992320001125336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,2,power_law_1.01,0.045535999536514285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,2,power_law_1.01,0.0480320006608963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,1,power_law_1.01,7.228607940673828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,2,power_law_1.01,0.0562175989151001
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,2,power_law_1.01,0.06792960166931153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,2,power_law_1.01,0.09103360176086425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,2,power_law_1.01,0.09278079867362976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,2,power_law_1.01,0.0948032021522522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,2,power_law_1.01,0.09547520279884339
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,2,power_law_1.01,0.09950079917907714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,2,power_law_1.01,0.10301439762115479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,2,power_law_1.01,0.10333440303802491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,2,power_law_1.01,0.10697599649429321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,2,1,power_law_1.01,8.854329681396484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,2,power_law_1.01,0.11300480365753174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,2,balanced,0.0738613357146581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,2,balanced,0.07655466596285503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,2,power_law_1.01,0.11947519779205322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,2,balanced,0.08057066798210144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,2,balanced,0.09246399998664856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,2,balanced,0.12083199620246887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,2,balanced,0.1733120083808899
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,2,power_law_1.01,0.12337919473648071
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,2,balanced,0.2271626591682434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,2,balanced,0.21499733130137125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,2,balanced,0.21475734313329062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,2,balanced,0.21901333332061768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,2,balanced,0.21983999013900757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,2,power_law_1.01,0.13165440559387206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,2,balanced,0.22286933660507202
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,2,balanced,0.22259734074274698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,2,balanced,0.23107733329137167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,2,balanced,0.23404266436894736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,2,power_law_1.01,0.14321919679641723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,2,balanced,0.2421226700146993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,2,balanced,0.25101866324742633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,2,balanced,0.27155200640360516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,2,balanced,0.29470932483673096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,2,power_law_1.01,0.15619839429855348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,2,balanced,0.33859201272328693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,2,balanced,0.37964268525441486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,2,balanced,0.4712640047073364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,2,balanced,0.563647985458374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,2,power_law_1.01,0.18967039585113527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,2,balanced,0.7591040134429932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,2,balanced,1.0211413701375325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,2,power_law_1.01,0.21939198970794677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,2,balanced,1.3964266777038574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,2,balanced,1.8691147168477376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,2,power_law_1.01,0.2870527982711792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,2,balanced,2.3245813051859536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,2,power_law_1.01,0.33267199993133545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,2,balanced,3.5726401011149087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,2,power_law_1.01,0.44263038635253904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,2,1,power_law_1.01,13.934573364257812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,2,power_law_1.01,0.5464064121246338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,2,balanced,6.8830827077229815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,2,power_law_1.01,0.7984831809997559
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,2,power_law_1.01,1.0325632095336914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,2688,1856,6,128,1,2,power_law_1.01,1.1782591819763184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,2688,1856,6,128,1,2,power_law_1.01,1.9335039138793946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,2688,1856,6,128,1,2,power_law_1.01,3.720569610595703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,2,1,power_law_1.01,26.64993896484375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,8,power_law_1.2,0.05694720149040222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,8,power_law_1.2,0.039468801021575926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,8,power_law_1.2,0.04673919975757599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,8,power_law_1.2,0.04826239943504333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,8,power_law_1.2,0.055174398422241214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,8,power_law_1.2,0.07404800057411194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,8,power_law_1.2,0.07474560141563416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,16,2,power_law_1.01,0.031014400720596313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,16,2,power_law_1.01,0.0318336009979248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,16,2,power_law_1.01,0.03221119940280914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,8,power_law_1.2,0.07596799731254578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,16,2,power_law_1.01,0.03343360126018524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,16,2,power_law_1.01,0.03313280045986176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,16,2,power_law_1.01,0.03514240086078644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,16,2,power_law_1.01,0.03702400028705597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,8,power_law_1.2,0.0785152018070221
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,16,2,power_law_1.01,0.03873920142650604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,16,2,power_law_1.01,0.040582400560379026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,16,2,power_law_1.01,0.04214400053024292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,16,2,power_law_1.01,0.044537600874900815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,16,2,power_law_1.01,0.04935680031776428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,8,power_law_1.2,0.07767680287361145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,16,2,power_law_1.01,0.04791040122509003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,16,2,power_law_1.01,0.0523967981338501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,16,2,power_law_1.01,0.05966079831123352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,16,2,power_law_1.01,0.06452479958534241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,8,power_law_1.2,0.08047999739646912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,16,2,power_law_1.01,0.06323199868202209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,16,2,power_law_1.01,0.0852288007736206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,16,2,power_law_1.01,0.09991040229797363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,16,2,power_law_1.01,0.12264319658279418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,8,power_law_1.2,0.08206080198287964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,16,2,power_law_1.01,0.14956799745559693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,16,2,power_law_1.01,0.18005119562149047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,16,2,power_law_1.01,0.25058560371398925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,8,power_law_1.2,0.08721280097961426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,16,2,power_law_1.01,0.319488000869751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,4,balanced,0.05815466741720835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,4,balanced,0.06222933530807495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,4,balanced,0.06449066599210103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,4,balanced,0.06845333178838094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,4,balanced,0.07878399888674419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,16,2,power_law_1.01,0.43695998191833496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,4,balanced,0.1060693363348643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,4,balanced,0.1373546620210012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,4,balanced,0.13825600345929465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,4,balanced,0.1385599970817566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,4,balanced,0.14197333653767905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,16,2,power_law_1.01,0.6342591762542724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,4,balanced,0.1357866624991099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,8,power_law_1.2,0.08633599877357483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,4,balanced,0.13828266660372415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,4,balanced,0.1407360037167867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,4,balanced,0.14298133055369058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,4,balanced,0.1504693329334259
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,4,balanced,0.15212266643842062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,4,balanced,0.16029866536458334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,16,2,power_law_1.01,0.7109439849853516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,4,balanced,0.17364799976348877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,4,balanced,0.185754656791687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,4,balanced,0.21406400203704834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,4,balanced,0.2397759954134623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,4,balanced,0.2959786653518677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,8,power_law_1.2,0.09583359956741333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,4,balanced,0.35154132048288983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,16,2,power_law_1.01,0.9965503692626954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,4,balanced,0.4745279947916667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,4,balanced,0.6346400181452433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,4,balanced,0.866757313410441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,8,power_law_1.2,0.10098559856414795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,4,balanced,1.1571786403656006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,16,2,power_law_1.01,1.5431488037109375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,4,balanced,1.4317439397176106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,4,balanced,2.203807989756266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,8,power_law_1.2,0.10362880229949951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,16,2,power_law_1.01,2.8481855392456055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,4,balanced,4.249039967854817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,8,power_law_1.2,0.11719679832458496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,8,power_law_1.2,0.14126720428466796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,8,power_law_1.2,0.16764160394668579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,8,power_law_1.2,0.17633919715881347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,8,power_law_1.2,0.23532159328460694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,8,power_law_1.2,0.29270401000976565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,8,power_law_1.2,0.36574079990386965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,8,power_law_1.2,0.4602560043334961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,8,power_law_1.2,0.6594240188598632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,8,power_law_1.2,0.8723648071289063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,8,power_law_1.2,1.3071423530578614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,8,power_law_1.2,2.271571159362793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,32,1,balanced,0.09706667065620422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,32,1,balanced,0.10103999574979146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,32,1,balanced,0.11183999975522359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,32,1,balanced,0.13368533054987589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,32,1,balanced,0.1816533406575521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,32,1,balanced,0.2619839906692505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,32,1,balanced,0.34669331709543866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,32,1,balanced,0.34681065877278644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,1,1,power_law_1.2,0.1260480046272278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,32,1,balanced,0.34786665439605713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,32,1,balanced,0.3476533492406209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,1,1,power_law_1.2,0.17935359477996826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,32,1,balanced,0.35525333881378174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,32,1,balanced,0.36028798421223956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,32,1,balanced,0.3649439811706543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,1,1,power_law_1.2,0.19347200393676758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,32,1,balanced,0.3703626791636149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,32,1,balanced,0.37982932726542157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,32,1,balanced,0.3852800130844116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,1,1,power_law_1.2,0.3835007905960083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,32,1,balanced,0.40123732884724933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,32,1,balanced,0.44884800910949707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,1,1,power_law_1.2,0.36062719821929934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,32,1,balanced,0.4851093292236328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,32,1,balanced,0.5623733202616373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,1,1,power_law_1.2,0.38489599227905275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,32,1,balanced,0.6323253313700358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,32,1,balanced,0.7899200121561686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,1,1,power_law_1.2,0.3907583951950073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,8,power_law_1.2,3.3501697540283204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,32,1,balanced,0.9430027008056641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,1,1,power_law_1.2,0.39945600032806394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,32,1,balanced,1.283029317855835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,1,1,power_law_1.2,0.40151681900024416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,1,1,power_law_1.2,0.41130242347717283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,32,1,balanced,1.757322629292806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,1,1,power_law_1.2,0.4012671947479248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,32,1,balanced,2.4363733927408853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,1,1,power_law_1.2,0.41805438995361327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,1,1,power_law_1.2,0.42558717727661133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,1,1,power_law_1.2,0.4908607959747314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,32,1,balanced,3.2488425572713218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,1,1,power_law_1.2,0.5223040103912353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,1,1,power_law_1.2,0.5536255836486816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,32,1,balanced,4.082085291544597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,1,1,power_law_1.2,0.5684671878814698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,1,1,power_law_1.2,0.6646848201751709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,1,1,power_law_1.2,0.6761023998260498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,1,1,power_law_1.2,0.854150390625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,32,1,balanced,6.311840057373047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,1,1,power_law_1.2,0.9492223739624024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,1,1,power_law_1.2,1.315443229675293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,1,1,power_law_1.2,1.6187583923339843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,1,1,power_law_1.2,2.2998271942138673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,32,1,balanced,12.02578608194987
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,1,1,power_law_1.2,3.0284351348876952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,1,1,power_law_1.2,4.45660171508789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,1,1,power_law_1.2,5.733868789672852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,1,1,power_law_1.2,7.209951782226563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,1,1,power_law_1.2,11.605561828613281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,32,8,power_law_1.2,0.05889279842376709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,32,8,power_law_1.2,0.06352639794349671
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,8,balanced,0.04420800010363261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,8,balanced,0.046682665745417275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,8,balanced,0.0484799991051356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,8,balanced,0.05236800014972687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,8,balanced,0.07474133372306824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,8,balanced,0.10744532942771912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,8,balanced,0.1302186648050944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,8,balanced,0.13054933150609335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,8,balanced,0.13337066769599915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,8,balanced,0.13310399651527405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,8,balanced,0.13306666413942972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,8,balanced,0.13754666845003763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,8,balanced,0.13937600453694662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,8,balanced,0.14134400089581808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,8,balanced,0.15041599671045938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,8,balanced,0.15122133493423462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,8,balanced,0.16081066926320395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,8,balanced,0.1731626590092977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,8,balanced,0.18561599651972452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,8,balanced,0.2108479936917623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,8,balanced,0.23866132895151773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,32,8,power_law_1.2,0.0626688003540039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,8,balanced,0.29078932603200275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,1,1,power_law_1.2,22.225241088867186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,8,balanced,0.34772801399230957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,8,balanced,0.4542773167292277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,8,balanced,0.5613386631011963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,8,balanced,0.7714719772338867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,8,balanced,0.987663984298706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,8,balanced,1.2173706690470378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,8,balanced,1.8519840240478516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,32,8,power_law_1.2,0.058457601070404056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,8,balanced,3.5926507314046225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,32,8,power_law_1.2,0.06410880088806152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,32,8,power_law_1.2,0.06893439888954163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,32,8,power_law_1.2,0.07004160284996033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,32,8,power_law_1.2,0.0748799979686737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,32,8,power_law_1.2,0.07608960270881653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,32,8,power_law_1.2,0.07702400088310242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,32,power_law_1.2,0.08926079869270324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,32,8,power_law_1.2,0.08298239707946778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,8,power_law_1.01,0.03374719917774201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,32,power_law_1.2,0.12002559900283813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,8,power_law_1.01,0.030585598945617676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,8,power_law_1.01,0.029971200227737426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,32,8,power_law_1.2,0.08432639837265014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,8,power_law_1.01,0.03150720000267029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,32,power_law_1.2,0.10670720338821411
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,8,power_law_1.01,0.033318400382995605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,8,power_law_1.01,0.037536001205444335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,32,8,power_law_1.2,0.08595200181007386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,32,power_law_1.2,0.06821119785308838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,8,power_law_1.01,0.03768320083618164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,8,power_law_1.01,0.03925119936466217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,8,power_law_1.01,0.04032639861106872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,32,power_law_1.2,0.07480319738388061
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,32,8,power_law_1.2,0.08784000277519226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,8,power_law_1.01,0.04156799912452698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,8,power_law_1.01,0.04408319890499115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,8,power_law_1.01,0.04717440009117126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,32,power_law_1.2,0.08271999955177307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,8,power_law_1.01,0.048588800430297854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,32,8,power_law_1.2,0.09676799774169922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,8,power_law_1.01,0.04971520006656647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,32,power_law_1.2,0.08118399977684021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,8,power_law_1.01,0.05454080104827881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,8,power_law_1.01,0.05735679864883423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,32,8,power_law_1.2,0.09881600141525268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,8,power_law_1.01,0.05842559933662415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,32,power_law_1.2,0.09128959774971009
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,8,power_law_1.01,0.0697983980178833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,8,power_law_1.01,0.08760319948196411
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,32,power_law_1.2,0.09443839788436889
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,32,8,power_law_1.2,0.10953600406646728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,8,power_law_1.01,0.10076800584793091
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,8,power_law_1.01,0.11202559471130372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,32,power_law_1.2,0.08498560190200806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,8,power_law_1.01,0.1560639977455139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,32,8,power_law_1.2,0.12535680532455445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,8,power_law_1.01,0.19347200393676758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,32,power_law_1.2,0.09291520118713378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,8,power_law_1.01,0.24893438816070557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,32,8,power_law_1.2,0.1435647964477539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,8,power_law_1.01,0.35326080322265624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,32,power_law_1.2,0.0927295982837677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,8,power_law_1.01,0.47711358070373533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,32,8,power_law_1.2,0.1714303970336914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,32,power_law_1.2,0.09866880178451538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,8,power_law_1.01,0.5882559776306152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,32,power_law_1.2,0.10065920352935791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,8,power_law_1.01,0.7193920135498046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,32,8,power_law_1.2,0.1864832043647766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,32,power_law_1.2,0.10343680381774903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,8,power_law_1.01,1.1590144157409668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,32,8,power_law_1.2,0.23897600173950195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,32,power_law_1.2,0.10899840593338013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,32,power_law_1.2,0.11641600131988525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,32,8,power_law_1.2,0.31736319065093993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,8,power_law_1.01,2.2444543838500977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,32,power_law_1.2,0.1246783971786499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,32,8,power_law_1.2,0.43636479377746584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,32,power_law_1.2,0.13715840578079225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,32,power_law_1.2,0.15777280330657958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,32,8,power_law_1.2,0.5522111892700196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,32,power_law_1.2,0.18698240518569947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,32,8,power_law_1.2,0.7816192150115967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,32,power_law_1.2,0.22707839012145997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,8,power_law_1.01,0.06968960165977478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,32,power_law_1.2,0.2932607889175415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,32,8,power_law_1.2,1.0209280014038087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,8,power_law_1.01,0.10337280035018921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,32,power_law_1.2,0.3840895891189575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,8,power_law_1.01,0.056934398412704465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,32,8,power_law_1.2,1.1715711593627929
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,32,power_law_1.2,0.5168191909790039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,4,power_law_1.01,0.037062400579452516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,8,power_law_1.01,0.0680895984172821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,32,power_law_1.2,0.6691904067993164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,4,power_law_1.01,0.0429504007101059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,4,power_law_1.01,0.039103999733924866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,4,power_law_1.01,0.048870399594306946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,8,power_law_1.01,0.07511680126190186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,32,8,power_law_1.2,1.783897590637207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,4,power_law_1.01,0.05566080212593079
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,32,power_law_1.2,0.8865663528442382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,4,power_law_1.01,0.07204480171203613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,4,power_law_1.01,0.07490559816360473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,8,power_law_1.01,0.09303680062294006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,4,power_law_1.01,0.07668480277061462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,4,power_law_1.01,0.07798399925231933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,32,power_law_1.2,1.1827520370483398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,4,power_law_1.01,0.07890560030937195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,8,power_law_1.01,0.09985920190811157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,4,power_law_1.01,0.08332160115242004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,4,power_law_1.01,0.08566399812698364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,32,8,power_law_1.2,4.119206237792969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,4,power_law_1.01,0.08711680173873901
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,8,power_law_1.01,0.10046720504760742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,4,power_law_1.01,0.09151359796524047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,32,power_law_1.2,1.8660736083984375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,4,power_law_1.01,0.0940671980381012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,8,power_law_1.01,0.10464639663696289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,4,power_law_1.01,0.1023360013961792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,4,power_law_1.01,0.10588159561157226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,4,power_law_1.01,0.12006399631500245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,8,power_law_1.01,0.10501760244369507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,4,power_law_1.01,0.1352512001991272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,4,power_law_1.01,0.16845439672470092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,4,power_law_1.01,0.1910272002220154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,32,power_law_1.2,3.3534656524658204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,8,power_law_1.01,0.10515199899673462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,4,power_law_1.01,0.23998079299926758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,4,power_law_1.01,0.2972543954849243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,8,power_law_1.01,0.10787839889526367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,4,power_law_1.01,0.41235198974609377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,4,power_law_1.01,0.5408063888549804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,8,power_law_1.01,0.10905599594116211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,4,power_law_1.01,0.753056001663208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,8,power_law_1.01,0.11123839616775513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,4,power_law_1.01,0.9714816093444825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,4,power_law_1.01,1.127782440185547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,8,power_law_1.01,0.12148480415344239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,32,2,power_law_1.01,0.032307198643684386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,32,2,power_law_1.01,0.03123840093612671
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,8,power_law_1.01,0.12222080230712891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,4,power_law_1.01,1.7918720245361328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,32,2,power_law_1.01,0.030457600951194763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,32,2,power_law_1.01,0.03200640082359314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,32,2,power_law_1.01,0.03264000117778778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,32,2,power_law_1.01,0.03306879997253418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,8,power_law_1.01,0.1305408000946045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,32,2,power_law_1.01,0.03298560082912445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,32,2,power_law_1.01,0.03358719944953918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,32,2,power_law_1.01,0.03425920009613037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,32,2,power_law_1.01,0.03390080034732819
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,32,2,power_law_1.01,0.03527680039405823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,4,power_law_1.01,3.2885696411132814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,32,2,power_law_1.01,0.0383679986000061
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,8,power_law_1.01,0.16444799900054932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,32,2,power_law_1.01,0.037273600697517395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,32,2,power_law_1.01,0.03978880047798157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,32,2,power_law_1.01,0.04357759952545166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,32,2,power_law_1.01,0.04694400131702423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,32,2,power_law_1.01,0.05004159808158874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,8,power_law_1.01,0.16809600591659546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,32,2,power_law_1.01,0.05930240154266357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,32,2,power_law_1.01,0.06412799954414368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,32,2,power_law_1.01,0.07516160011291503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,32,2,power_law_1.01,0.098854398727417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,32,2,power_law_1.01,0.11653759479522705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,32,2,power_law_1.01,0.12930560111999512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,8,power_law_1.01,0.20314879417419435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,32,2,power_law_1.01,0.16332800388336183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,32,1,power_law_1.2,0.03468160033226013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,32,1,power_law_1.2,0.034227201342582704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,32,2,power_law_1.01,0.21087360382080078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,32,1,power_law_1.2,0.034431999921798705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,32,1,power_law_1.2,0.03692159950733185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,32,2,power_law_1.01,0.30231680870056155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,32,1,power_law_1.2,0.03749760091304779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,32,1,power_law_1.2,0.0380160003900528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,32,2,power_law_1.01,0.41093759536743163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,32,1,power_law_1.2,0.03845759928226471
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,8,power_law_1.01,0.24617600440979004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,32,1,power_law_1.2,0.03975679874420166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,32,1,power_law_1.2,0.03958399891853333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,32,2,power_law_1.01,0.5506879806518554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,32,1,power_law_1.2,0.04014720022678375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,32,1,power_law_1.2,0.041407999396324155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,32,1,power_law_1.2,0.042777600884437564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,32,2,power_law_1.01,0.7867392063140869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,32,1,power_law_1.2,0.044435200095176694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,32,1,power_law_1.2,0.04410879909992218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,32,1,power_law_1.2,0.051020801067352295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,32,1,power_law_1.2,0.05311359763145447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,32,1,power_law_1.2,0.05692160129547119
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,8,power_law_1.01,0.3102272033691406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,32,1,power_law_1.2,0.06511359810829162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,32,2,power_law_1.01,1.5024448394775392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,32,1,power_law_1.2,0.07382400035858154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,32,1,power_law_1.2,0.09128320217132568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,32,1,power_law_1.2,0.10483839511871337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,32,1,power_law_1.2,0.13923200368881225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,32,1,power_law_1.2,0.16766079664230346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,32,1,power_law_1.2,0.23152000904083253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,32,1,power_law_1.2,0.2951807975769043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,8,power_law_1.01,0.34902400970458985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,32,1,power_law_1.2,0.42184958457946775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,32,1,power_law_1.2,0.5542912006378173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,32,1,power_law_1.2,0.6776832103729248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,32,1,power_law_1.2,1.0484416007995605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,8,power_law_1.01,0.4714176177978516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,32,1,power_law_1.2,2.0263040542602537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,64,balanced,0.05657066901524862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,64,balanced,0.05640000104904175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,64,balanced,0.0708000014225642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,64,balanced,0.09197333455085754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,64,balanced,0.10481066505114238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,64,balanced,0.10487467050552368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,64,balanced,0.10645332932472229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,64,balanced,0.10569600264231364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,64,balanced,0.10538666447003682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,64,balanced,0.10611200332641602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,64,balanced,0.10777067144711812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,64,balanced,0.10875200231870015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,64,balanced,0.11010133226712544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,64,balanced,0.1111306647459666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,64,balanced,0.11432533462842305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,64,balanced,0.11692266662915547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,64,balanced,0.1197813351949056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,64,balanced,0.1482080022493998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,64,balanced,0.13430933157602945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,64,balanced,0.1637493371963501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,64,balanced,0.1685439944267273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,64,balanced,0.2384693423906962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,64,balanced,0.20639467239379883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,8,power_law_1.01,0.5688767910003663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,64,balanced,0.276256004969279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,64,balanced,0.2773973345756531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,64,balanced,0.3810346523920695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,64,balanced,0.43225598335266113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,1,64,balanced,0.5178133249282837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,1,64,balanced,0.7305066585540771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,1,64,balanced,1.5340639750162761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,8,power_law_1.01,0.7912831783294678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,8,power_law_1.01,1.0522496223449707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,4,4,power_law_1.01,0.028492799401283263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,4,4,power_law_1.01,0.02568320035934448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,4,4,power_law_1.01,0.026188799738883974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,8,8,power_law_1.01,1.2000896453857421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,4,4,power_law_1.01,0.025119999051094057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,4,balanced,0.03815466662247976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,4,balanced,0.03636800001064936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,4,balanced,0.04019733270009359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,4,balanced,0.04390933116277059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,4,balanced,0.05213333169619242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,4,balanced,0.05093333125114441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,4,balanced,0.05277866621812185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,4,balanced,0.05410666763782501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,4,4,power_law_1.01,0.02715519964694977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,4,balanced,0.05194133520126343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,4,balanced,0.0528053343296051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,4,balanced,0.05400000015894572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,4,balanced,0.05522133409976959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,4,balanced,0.05648533503214518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,4,balanced,0.05454400181770325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,4,balanced,0.06005866825580597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,4,balanced,0.06076266864935557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,4,balanced,0.0643146683772405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,4,balanced,0.06850666801134746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,4,balanced,0.07863999903202057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,4,balanced,0.09226666887601216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,4,balanced,0.105295995871226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,4,balanced,0.1337493360042572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,4,balanced,0.15265599886576334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,4,4,power_law_1.01,0.03036159873008728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,4,balanced,0.2079520026842753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,4,balanced,0.2618880073229472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,4,balanced,0.3747893174489339
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,4,balanced,0.4742666482925415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,4,balanced,0.5813226699829102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,4,balanced,0.8934613068898519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,4,4,power_law_1.01,0.030969598889350893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,4,balanced,1.7330400149027507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,4,4,power_law_1.01,0.03288320004940033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,8,8,power_law_1.01,1.8476991653442383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,8,power_law_1.01,0.03511039912700653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,4,4,power_law_1.01,0.036185601353645326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,8,power_law_1.01,0.031276801228523256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,4,4,power_law_1.01,0.03496319949626923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,8,power_law_1.01,0.029631999135017396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,8,power_law_1.01,0.03059839904308319
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,4,4,power_law_1.01,0.035980799794197084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,8,power_law_1.01,0.031763198971748355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,4,4,power_law_1.01,0.03770880103111267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,8,power_law_1.01,0.03738240003585815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,8,power_law_1.01,0.036499199271202085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,4,4,power_law_1.01,0.03785600066184998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,8,power_law_1.01,0.03797119855880737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,4,4,power_law_1.01,0.04307839870452881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,8,power_law_1.01,0.037836799025535585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,8,power_law_1.01,0.03899520039558411
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,4,4,power_law_1.01,0.050387197732925416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,8,power_law_1.01,0.03873920142650604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,4,4,power_law_1.01,0.05233280062675476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,8,power_law_1.01,0.03979519903659821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,8,power_law_1.01,0.043635201454162595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,4,4,power_law_1.01,0.053337597846984865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,8,power_law_1.01,0.044915199279785156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,4,4,power_law_1.01,0.05987200140953064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,8,8,power_law_1.01,3.6873855590820312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,8,power_law_1.01,0.05185920000076294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,8,power_law_1.01,0.05444480180740356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,4,power_law_1.01,0.07311999797821045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,8,power_law_1.01,0.054150402545928955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,4,power_law_1.01,0.08473600149154663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,8,power_law_1.01,0.061894398927688596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,4,power_law_1.01,0.089683198928833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,8,power_law_1.01,0.06829439997673034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,8,power_law_1.01,0.08736000061035157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,1,balanced,0.07403733332951863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,4,power_law_1.01,0.10366079807281495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,1,balanced,0.10947199662526448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,1,balanced,0.12607999642690024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,1,balanced,0.1983840068181356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,1,balanced,0.33321066697438556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,8,power_law_1.01,0.09742720127105713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,1,balanced,0.466048002243042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,1,balanced,0.46942933400472003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,1,balanced,0.47142934799194336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,1,balanced,0.47430400053660077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,4,power_law_1.01,0.13685760498046876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,1,balanced,0.4771733283996582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,8,power_law_1.01,0.11880960464477539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,1,balanced,0.482474684715271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,1,balanced,0.48950401941935223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,1,balanced,0.4954293171564738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,1,balanced,0.5032906532287598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,1,balanced,0.5166186491648356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,8,power_law_1.01,0.13605120182037353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,1,balanced,0.5270826816558838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,4,power_law_1.01,0.1788480043411255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,1,balanced,0.549509326616923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,1,balanced,0.6184106667836508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,1,balanced,0.6436479886372884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,1,balanced,0.8812906742095947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,8,power_law_1.01,0.2076159954071045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,1,balanced,0.8846293290456136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,4,power_law_1.01,0.2191551923751831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,1,balanced,1.4183947245279949
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,1,balanced,1.4587680498758953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,8,power_law_1.01,0.23928320407867432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,1,balanced,2.2863732973734536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,4,power_law_1.01,0.3122944116592407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,1,balanced,2.440282662709554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,8,power_law_1.01,0.3441215991973877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,1,balanced,3.720757484436035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,8,power_law_1.01,0.4651328086853027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,4,power_law_1.01,0.40953598022460935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,1,balanced,4.6462507247924805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,8,power_law_1.01,0.5468160152435303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,4,power_law_1.01,0.4571072101593018
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,8,1,balanced,5.980725606282552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,8,power_law_1.01,0.8532223701477051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,8,1,balanced,8.779989242553711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,4,power_law_1.01,0.7973120212554932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,8,1,balanced,0.037402667105197906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,8,1,balanced,0.039877332746982574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,8,1,balanced,0.04548799991607666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,8,1,balanced,0.04660800099372864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,8,1,balanced,0.048101335763931274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,8,1,balanced,0.048101335763931274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,8,1,balanced,0.04801600178082784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,8,1,balanced,0.04990399877230326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,8,1,balanced,0.04993066688378652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,8,1,balanced,0.05111999809741974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,8,1,balanced,0.051114668448766075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,8,1,balanced,0.05383466680844625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,8,1,balanced,0.05677866439024607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,8,1,balanced,0.06222933530807495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,8,1,balanced,0.0665226678053538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,8,1,balanced,0.0703306645154953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,8,1,balanced,0.07124266525109609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,8,1,balanced,0.08495466907819112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,8,1,balanced,0.09530666470527649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,8,1,balanced,0.11733333269755046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,8,1,balanced,0.13215999801953635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,8,1,balanced,0.1759679913520813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,8,1,balanced,0.20784533023834229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,8,1,balanced,0.29417065779368085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,8,1,balanced,0.3715360164642334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,8,1,balanced,0.5385226806004842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,8,1,balanced,0.6918026606241862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,8,1,balanced,0.8540799617767334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,8,power_law_1.01,1.9751935958862306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,8,1,balanced,1.3133920033772786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,4,power_law_1.01,1.488428783416748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,8,1,balanced,16.92540740966797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,8,1,balanced,2.602469285329183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,8,balanced,0.033439998825391136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,32,4,power_law_1.01,0.046291199326515195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,8,balanced,0.03349866718053818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,8,balanced,0.03346133232116699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,8,balanced,0.05184000233809153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,8,balanced,0.0699893335501353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,8,balanced,0.08080000181992848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,8,balanced,0.08230400085449219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,8,balanced,0.0764160007238388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,16,power_law_1.01,0.07251840233802795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,8,balanced,0.07684266567230225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,8,balanced,0.07748800019423167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,8,balanced,0.07810133198897044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,8,balanced,0.0775733341773351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,8,balanced,0.0796853353579839
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,8,balanced,0.0798880010843277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,8,balanced,0.08277333279450734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,8,balanced,0.08266133566697438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,8,balanced,0.0828959991534551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,8,balanced,0.0969493289788564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,8,balanced,0.09028266867001851
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,16,power_law_1.2,0.055776000022888184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,8,balanced,0.11741333206494649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,8,balanced,0.10177600383758545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,8,balanced,0.12869333227475485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,8,balanced,0.12221333384513855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,8,balanced,0.1648373305797577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,8,balanced,0.18241600195566812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,16,power_law_1.01,0.08511360287666321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,8,balanced,0.25672000646591187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,8,balanced,0.25286932786305744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,1,8,balanced,0.302074670791626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,32,4,power_law_1.01,0.049132800102233885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,1,8,balanced,0.4254719813664754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,1,8,balanced,0.7834719816843668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,16,power_law_1.2,0.05730559825897217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,16,power_law_1.01,0.05069440007209778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,16,power_law_1.2,0.04384639859199524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,16,power_law_1.01,0.04987519979476929
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,32,4,power_law_1.01,0.05785599946975708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,16,power_law_1.2,0.047628799080848695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,16,power_law_1.01,0.05311359763145447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,16,power_law_1.2,0.0471231997013092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,16,power_law_1.01,0.05994240045547485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,32,4,power_law_1.01,0.05273600220680237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,16,power_law_1.2,0.05299839973449707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,16,power_law_1.01,0.06563839912414551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,16,power_law_1.2,0.05427200198173523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,16,power_law_1.01,0.06970880031585694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,16,power_law_1.01,0.06862720251083373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,16,power_law_1.2,0.05479679703712463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,32,4,power_law_1.01,0.06095359921455383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,16,power_law_1.01,0.06686080098152161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,16,power_law_1.2,0.05631999969482422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,16,power_law_1.01,0.07167999744415283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,16,power_law_1.2,0.05707520246505737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,32,4,power_law_1.01,0.06746879816055298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,16,power_law_1.01,0.07388799786567687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,16,power_law_1.2,0.05743359923362732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,16,power_law_1.01,0.07443199753761291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,16,power_law_1.2,0.061427199840545656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,16,power_law_1.01,0.07534719705581665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,16,power_law_1.2,0.06453120112419128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,32,4,power_law_1.01,0.07311360239982605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,16,power_law_1.01,0.08354560136795045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,16,power_law_1.2,0.07145599722862243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,16,power_law_1.01,0.0831488013267517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,16,power_law_1.2,0.050944000482559204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,16,power_law_1.2,0.07896320223808288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,16,power_law_1.2,0.04392960071563721
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,32,4,power_law_1.01,0.08195199966430664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,16,power_law_1.01,0.0908352017402649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,16,power_law_1.2,0.08243839740753174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,16,power_law_1.2,0.04645760059356689
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,16,power_law_1.01,0.0966592013835907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,16,power_law_1.2,0.0473471999168396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,16,power_law_1.2,0.08625280261039733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,16,power_law_1.2,0.05008000135421753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,16,power_law_1.01,0.1044160008430481
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,32,4,power_law_1.01,0.08344320058822632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,16,power_law_1.2,0.09997439980506898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,16,power_law_1.2,0.05576320290565491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,16,power_law_1.01,0.12810879945755005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,16,power_law_1.2,0.0599295973777771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,16,power_law_1.2,0.11247999668121338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,16,power_law_1.01,0.15237120389938355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,16,power_law_1.2,0.061452800035476686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,32,4,power_law_1.01,0.08642560243606567
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,16,power_law_1.2,0.06346240043640136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,16,power_law_1.2,0.13831679821014403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,16,power_law_1.01,0.18002560138702392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,16,power_law_1.2,0.05980160236358643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,16,power_law_1.2,0.17224960327148436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,16,power_law_1.2,0.06627839803695679
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,16,power_law_1.01,0.20478720664978028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,32,4,power_law_1.01,0.08983039855957031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,16,power_law_1.2,0.0673471987247467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,16,power_law_1.2,0.2198335886001587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,16,power_law_1.01,0.2785088062286377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,16,power_law_1.2,0.07182719707489013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,16,power_law_1.2,0.07194240093231201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,16,power_law_1.2,0.2618943929672241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,32,4,power_law_1.01,0.09763839840888977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,16,power_law_1.01,0.3473344087600708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,16,power_law_1.2,0.07960960268974304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,16,power_law_1.2,0.38355839252471924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,16,power_law_1.2,0.08184319734573364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,16,power_law_1.01,0.4902463912963867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,16,power_law_1.2,0.08380799889564514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,32,4,power_law_1.01,0.09868800044059753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,16,power_law_1.2,0.5097599983215332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,16,power_law_1.2,0.0992576003074646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,16,power_law_1.01,0.6180416107177734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,16,power_law_1.2,0.09466879963874816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,16,power_law_1.2,0.6731200218200684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,32,4,power_law_1.01,0.10421760082244873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,16,power_law_1.2,0.11887359619140625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,16,power_law_1.2,0.13153280019760133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,16,power_law_1.01,0.7376704216003418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,16,power_law_1.2,0.9229887962341309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,16,power_law_1.2,0.18815360069274903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,32,4,power_law_1.01,0.11656320095062256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,16,power_law_1.2,0.23299839496612548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,16,power_law_1.01,1.1435775756835938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,16,power_law_1.2,0.31061758995056155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,16,power_law_1.2,1.0944640159606933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,32,4,power_law_1.01,0.12057600021362305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,16,power_law_1.2,0.3470720052719116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,16,power_law_1.2,0.5543424129486084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,32,4,power_law_1.01,0.1361791968345642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,16,power_law_1.2,1.7394111633300782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,16,power_law_1.01,2.0783424377441406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,16,power_law_1.2,0.7199999809265136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,32,4,power_law_1.01,0.161190402507782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,16,power_law_1.2,1.0325247764587402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,32,4,power_law_1.01,0.16202239990234374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,16,power_law_1.2,3.7731582641601564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,16,power_law_1.2,1.452608013153076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,32,4,power_law_1.01,0.20390400886535645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,32,4,power_law_1.01,0.24302079677581787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,16,power_law_1.2,2.8767040252685545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,32,4,power_law_1.01,0.3220736026763916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,32,4,power_law_1.01,0.39831678867340087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,32,4,power_law_1.01,0.5621376037597656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,32,4,power_law_1.01,0.6983359813690185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,128,power_law_1.01,0.08636159896850586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,32,4,power_law_1.01,1.0435263633728027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,128,power_law_1.01,0.09749760031700135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,32,balanced,0.041984001795450844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,32,balanced,0.04387733340263367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,32,balanced,0.04387733340263367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,32,balanced,0.04422399898370107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,32,balanced,0.04653333127498627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,32,balanced,0.048954665660858154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,32,balanced,0.04987200101216634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,32,balanced,0.050069332122802734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,32,balanced,0.0521066685517629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,32,balanced,0.05187733471393585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,128,power_law_1.01,0.08455680012702942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,32,balanced,0.05236800014972687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,32,balanced,0.05409599840641022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,32,balanced,0.05399466554323832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,32,balanced,0.05643733342488607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,32,balanced,0.06042666733264923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,32,balanced,0.062122667829195656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,32,balanced,0.06605333089828491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,32,balanced,0.07097066442171733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,32,balanced,0.07841599980990092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,32,balanced,0.10076266527175903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,32,balanced,0.12180800239245097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,32,balanced,0.15679466724395752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,32,4,power_law_1.01,1.3435392379760742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,32,balanced,0.18602667252222696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,32,balanced,0.25331199169158936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,32,balanced,0.3162613312403361
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,32,balanced,0.44330668449401855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,128,power_law_1.01,0.07817599773406983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,32,balanced,0.5687893231709799
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,32,balanced,0.698527971903483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,32,balanced,1.067029317220052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,32,balanced,2.0903679529825845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,128,power_law_1.01,0.07591040134429931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,16,power_law_1.01,0.05336959958076477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,32,4,power_law_1.01,1.6254528045654297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,16,power_law_1.01,0.0379584014415741
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,128,power_law_1.01,0.0810368001461029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,16,power_law_1.01,0.031206399202346802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,16,power_law_1.01,0.031167998909950256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,128,power_law_1.01,0.08479999899864196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,16,power_law_1.01,0.032576000690460204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,16,power_law_1.01,0.0346560001373291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,32,4,power_law_1.01,2.526035118103027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,128,power_law_1.01,0.08669440150260925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,16,power_law_1.01,0.034764799475669864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,16,power_law_1.01,0.0363072007894516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,16,power_law_1.01,0.03527680039405823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,128,power_law_1.01,0.08751999735832214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,16,power_law_1.01,0.038252800703048706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,16,power_law_1.01,0.03877759873867035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,128,power_law_1.01,0.0843392014503479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,16,power_law_1.01,0.040729600191116336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,16,power_law_1.01,0.04017919898033142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,16,power_law_1.01,0.04399999976158142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,128,power_law_1.01,0.08829439878463745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,16,power_law_1.01,0.05109120011329651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,32,4,power_law_1.01,4.913062286376953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,16,power_law_1.01,0.05248000025749207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,128,power_law_1.01,0.0883903980255127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,16,power_law_1.01,0.05550720095634461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,16,power_law_1.01,0.06190720200538635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,128,power_law_1.01,0.08828799724578858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,16,power_law_1.01,0.06050559878349304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,16,power_law_1.01,0.07029119729995728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,16,power_law_1.01,0.07943040132522583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,128,power_law_1.01,0.08894720077514648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,4,power_law_1.2,0.03565439879894257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,16,power_law_1.01,0.09357439875602722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,4,power_law_1.2,0.038022398948669434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,16,power_law_1.01,0.1133952021598816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,128,power_law_1.01,0.09422720074653626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,4,power_law_1.2,0.03805440068244934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,16,power_law_1.01,0.14817919731140136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,4,power_law_1.2,0.043263998627662656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,128,power_law_1.01,0.0961408019065857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,16,power_law_1.01,0.17222399711608888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,4,power_law_1.2,0.05356799960136414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,4,power_law_1.2,0.07227519750595093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,16,power_law_1.01,0.2563647985458374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,128,power_law_1.01,0.09846400022506714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,4,power_law_1.2,0.07240960001945496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,4,power_law_1.2,0.072326397895813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,16,power_law_1.01,0.29006080627441405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,4,power_law_1.2,0.07363200187683105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,128,power_law_1.01,0.10275839567184449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,1,16,power_law_1.01,0.3646464109420776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,4,power_law_1.2,0.0766975998878479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,4,power_law_1.2,0.07985919713973999
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,128,power_law_1.01,0.11265280246734619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,4,power_law_1.2,0.07941120266914367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,1,16,power_law_1.01,0.5367487907409668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,4,power_law_1.2,0.08021119832992554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,128,power_law_1.01,0.12694400548934937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,4,power_law_1.2,0.08399360179901123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,4,power_law_1.2,0.0907584011554718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,4,power_law_1.2,0.09286400079727172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,16,2,power_law_1.2,0.03357439935207367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,1,16,power_law_1.01,1.0923968315124513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,128,power_law_1.01,0.1462847948074341
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,16,2,power_law_1.2,0.033990401029586795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,4,power_law_1.2,0.10087039470672607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,16,2,power_law_1.2,0.03324800133705139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,4,power_law_1.2,0.11366399526596069
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,16,2,power_law_1.2,0.03537920117378235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,128,power_law_1.01,0.18592000007629395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,16,2,power_law_1.2,0.03509120047092438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,4,power_law_1.2,0.11407359838485717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,16,2,power_law_1.2,0.0352512001991272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,4,power_law_1.2,0.13731199502944946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,16,2,power_law_1.2,0.03780480027198792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,16,2,power_law_1.2,0.0382207989692688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,128,power_law_1.01,0.22777600288391114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,4,power_law_1.2,0.13941760063171388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,16,2,power_law_1.2,0.03806720077991486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,16,2,power_law_1.2,0.04307839870452881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,4,power_law_1.2,0.17074559926986693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,16,2,power_law_1.2,0.041843199729919435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,16,2,power_law_1.2,0.04539520144462585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,4,power_law_1.2,0.19587839841842652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,128,power_law_1.01,0.29002881050109863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,16,2,power_law_1.2,0.04852479994297028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,16,2,power_law_1.2,0.050271999835968015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,16,2,power_law_1.2,0.05756160020828247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,4,power_law_1.2,0.2502912044525146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,16,2,power_law_1.2,0.060147202014923094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,16,2,power_law_1.2,0.06108800172805786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,4,power_law_1.2,0.3150592088699341
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,16,2,power_law_1.2,0.0756991982460022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,128,power_law_1.01,0.3605184078216553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,16,2,power_law_1.2,0.08224639892578126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,16,2,power_law_1.2,0.10181119441986083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,4,power_law_1.2,0.425600004196167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,16,2,power_law_1.2,0.11864960193634033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,16,2,power_law_1.2,0.14179840087890624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,16,2,power_law_1.2,0.17004799842834473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,128,power_law_1.01,0.5331711769104004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,4,power_law_1.2,0.5192639827728271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,16,2,power_law_1.2,0.23912959098815917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,16,2,power_law_1.2,0.3083199977874756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,1,4,power_law_1.2,0.6791999816894532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,16,2,power_law_1.2,0.42218241691589353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,16,2,power_law_1.2,0.5544256210327149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,128,power_law_1.01,0.6363647937774658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,16,2,power_law_1.2,0.5266240119934082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,1,4,power_law_1.2,0.9646143913269043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,16,2,power_law_1.2,1.072505569458008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,128,power_law_1.01,0.7921152114868164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,16,2,power_law_1.2,1.9178047180175781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,1,4,power_law_1.2,2.111955261230469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,2,1,power_law_1.2,0.032416000962257385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,2,1,power_law_1.2,0.04323199987411499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,2,1,power_law_1.2,0.04387840032577515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,128,power_law_1.01,1.2579903602600098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,2,1,power_law_1.2,0.055430400371551516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,2,1,power_law_1.2,0.06158080101013184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,2,1,power_law_1.2,0.10846079587936401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,2,1,power_law_1.2,0.11089919805526734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,2,1,power_law_1.2,0.11663999557495117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,2,1,power_law_1.2,0.1200063943862915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,128,power_law_1.01,2.4204032897949217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,2,1,power_law_1.2,0.12010879516601562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,2,1,power_law_1.2,0.12510720491409302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,2,1,power_law_1.2,0.12855039834976195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,2,1,power_law_1.2,0.13006720542907715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,2,1,power_law_1.2,0.13633279800415038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,2,1,power_law_1.2,0.1400320053100586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,2,1,power_law_1.2,0.14531199932098388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,2,1,power_law_1.2,0.15427839756011963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,2,1,power_law_1.2,0.17173759937286376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,1,power_law_1.2,0.18812799453735352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,1,power_law_1.2,0.22931840419769287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,1,power_law_1.2,0.24736640453338624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,1,power_law_1.2,0.31792640686035156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,1,power_law_1.2,0.38663039207458494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,32,power_law_1.01,0.09608960151672363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,1,power_law_1.2,0.5136576175689698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,1,power_law_1.2,0.09648640155792236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,1,power_law_1.2,0.6393983840942383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,32,power_law_1.01,0.09946240186691284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,1,power_law_1.2,0.8882816314697266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,32,power_law_1.01,0.07411839962005615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,1,power_law_1.2,1.1774975776672363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,32,power_law_1.01,0.06478719711303711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,32,1,balanced,0.03618666778008143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,1,power_law_1.2,0.14702080488204955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,32,1,balanced,0.035989334185918175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,32,1,balanced,0.04085866610209147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,32,1,balanced,0.053818667928377785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,32,1,balanced,0.08222933113574982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,32,1,balanced,0.08306133250395457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,32,1,balanced,0.0865280032157898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,32,1,balanced,0.08621866504351298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,32,1,balanced,0.08741866548856099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,32,1,balanced,0.08934932947158813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,32,1,balanced,0.08904533584912618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,32,1,balanced,0.09281599521636963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,32,power_law_1.01,0.0804095983505249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,32,1,balanced,0.0940053363641103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,32,1,balanced,0.09685867031415303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,32,1,balanced,0.10243733723958333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,32,1,balanced,0.10746133327484131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,32,1,balanced,0.11866666873296101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,32,1,balanced,0.13433067003885904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,32,1,balanced,0.16481066743532816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,1,power_law_1.2,1.4354623794555663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,32,1,balanced,0.19261332352956137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,32,1,balanced,0.23548267285029092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,32,1,balanced,0.35091733932495117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,32,1,balanced,0.40460264682769775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,32,1,balanced,0.575760006904602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,32,power_law_1.01,0.078847998380661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,32,1,balanced,0.7257493336995443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,32,1,balanced,1.083621342976888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,32,1,balanced,1.3993706703186035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,1,power_law_1.2,0.1949952006340027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,32,power_law_1.01,0.09118080139160156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,32,1,balanced,1.752687931060791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,1,power_law_1.2,2.1503103256225584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,32,1,balanced,2.7080586751302085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,32,power_law_1.01,0.08996480107307434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,32,1,balanced,5.2437334060668945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,32,power_law_1.01,0.09648000001907349
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,1,power_law_1.2,0.32855679988861086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,32,power_law_1.01,0.09779199957847595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,1,power_law_1.2,4.222099304199219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,32,power_law_1.01,0.09710720181465149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,1,power_law_1.2,0.5401023864746094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,32,power_law_1.01,0.09800959825515747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,32,power_law_1.01,0.1001855969429016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,32,power_law_1.01,0.10750080347061157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,1,power_law_1.2,0.7883647918701172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,32,power_law_1.01,0.1087615966796875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,32,power_law_1.01,0.11590399742126464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,1,power_law_1.2,0.9610624313354492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,32,power_law_1.01,0.1186560034751892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,32,power_law_1.01,0.140339195728302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,32,power_law_1.01,0.14954240322113038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,1,power_law_1.2,1.35731201171875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,32,power_law_1.01,0.17624319791793824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,32,power_law_1.01,0.18802560567855836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,1,power_law_1.2,1.4638912200927734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,32,power_law_1.01,0.24826879501342775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,128,power_law_1.2,0.0662015974521637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,32,power_law_1.01,0.30028159618377687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,1,power_law_1.2,1.5414079666137694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,32,power_law_1.01,0.3850111961364746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,1,balanced,0.09963732957839966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,128,power_law_1.2,0.06448000073432922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,1,balanced,0.14844266573588052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,1,balanced,0.24673600991566977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,1,balanced,0.44323734442392987
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,1,balanced,0.8311200141906738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,1,balanced,1.6149600346883137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,1,balanced,1.6207680702209473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,1,power_law_1.2,1.66943359375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,32,power_law_1.01,0.5039231777191162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,128,power_law_1.2,0.06380159854888916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,1,balanced,1.6205919583638508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,1,balanced,1.6230932871500652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,1,balanced,1.6267840067545574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,1,balanced,1.64629332224528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,32,power_law_1.01,0.7096255779266357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,128,power_law_1.2,0.06147199869155884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,1,balanced,1.6578346888224285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,1,balanced,1.6636266708374023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,1,balanced,1.667093276977539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,1,power_law_1.2,1.6895360946655273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,1,balanced,1.6793759663899739
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,1,balanced,1.6897865931193035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,128,power_law_1.2,0.05987840294837952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,32,power_law_1.01,0.8672320365905761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,1,balanced,1.712410608927409
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,1,balanced,1.7629173596700032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,1,balanced,1.8162719408671062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,128,power_law_1.2,0.06130560040473938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,1,balanced,1.899903933207194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,1,power_law_1.2,1.7598527908325194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,1,balanced,2.0109333992004395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,32,power_law_1.01,1.1208383560180664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,1,balanced,2.1793972651163735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,1,balanced,2.504138628641764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,128,power_law_1.2,0.061612802743911746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,1,balanced,3.005119959513346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,1,balanced,3.0272159576416016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,1,power_law_1.2,1.8309440612792969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,128,power_law_1.2,0.06269440054893494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,32,power_law_1.01,1.6008512496948242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,1,balanced,4.685920079549153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,128,power_law_1.2,0.0636352002620697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,1,balanced,4.9847412109375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,1,power_law_1.2,1.9069183349609375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,128,power_law_1.2,0.06336640119552613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,1,balanced,7.036069234212239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,32,power_law_1.01,3.2603519439697264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,128,power_law_1.2,0.06451839804649354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,1,power_law_1.2,1.969887924194336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,1,balanced,9.334298451741537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,128,power_law_1.2,0.06714239716529846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,1,power_law_1.2,2.0816320419311523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,128,power_law_1.2,0.06713600158691406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,1,balanced,18.187402089436848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,128,power_law_1.2,0.06706560254096985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,1,power_law_1.2,2.309971237182617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,128,power_law_1.2,0.07215359807014465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,128,power_law_1.2,0.07429760098457336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,1,power_law_1.2,2.469913673400879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,128,power_law_1.2,0.07849599719047547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,1,power_law_1.2,2.6710784912109373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,128,power_law_1.2,0.08720639944076539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,128,power_law_1.2,0.09849600195884704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,1,power_law_1.2,2.7782079696655275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,128,power_law_1.2,0.11771520376205444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,1,power_law_1.01,0.035046398639678955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,1,power_law_1.01,0.037484800815582274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,128,power_law_1.2,0.1377344012260437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,1,power_law_1.01,0.04074879884719849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,1,power_law_1.2,3.108492851257324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,1,power_law_1.01,0.04785920083522797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,128,power_law_1.2,0.1792896032333374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,1,power_law_1.01,0.05761280059814453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,1,power_law_1.01,0.07802240252494812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,1,power_law_1.01,0.08147839903831482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,128,power_law_1.2,0.21932799816131593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,8,power_law_1.01,0.04806399941444397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,1,power_law_1.01,0.0824895977973938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,1,power_law_1.2,3.3356990814208984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,1,power_law_1.01,0.08695679903030396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,8,power_law_1.01,0.04995200037956238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,128,power_law_1.2,0.30448639392852783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,1,power_law_1.01,0.08721920251846313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,8,power_law_1.01,0.046214398741722104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,1,power_law_1.01,0.09049599766731262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,1,power_law_1.01,0.09467520117759705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,8,power_law_1.01,0.050425601005554196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,128,power_law_1.2,0.39728000164031985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,1,power_law_1.01,0.09815040230751038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,1,power_law_1.2,3.9349056243896485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,8,power_law_1.01,0.05235199928283692
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,1,power_law_1.01,0.10611840486526489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,8,power_law_1.01,0.056934398412704465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,1,power_law_1.01,0.1148800015449524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,8,power_law_1.01,0.059648001194000246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,1,power_law_1.01,0.11971839666366577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,128,power_law_1.2,0.5789440155029297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,1,power_law_1.01,0.13010560274124144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,8,power_law_1.01,0.061900800466537474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,1,power_law_1.01,0.15022079944610595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,8,power_law_1.01,0.06414719820022582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,1,power_law_1.2,4.384102249145508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,1,power_law_1.01,0.1670016050338745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,8,power_law_1.01,0.06344959735870362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,128,power_law_1.2,0.7407423973083496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,1,power_law_1.01,0.20706560611724853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,8,power_law_1.01,0.06569600105285645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,1,power_law_1.01,0.2451200008392334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,8,power_law_1.01,0.06885120272636414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,1,power_law_1.01,0.33397760391235354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,8,power_law_1.01,0.07140480279922486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,128,power_law_1.2,0.9407615661621094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,8,power_law_1.01,0.07118719816207886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,1,power_law_1.01,0.4148223876953125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,1,power_law_1.2,5.494527816772461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,8,power_law_1.01,0.07735679745674133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,1,power_law_1.01,0.5931392192840577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,8,power_law_1.01,0.08372480273246766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,1,power_law_1.01,0.7614975929260254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,8,power_law_1.01,0.08663679957389832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,128,power_law_1.2,1.485587215423584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,8,power_law_1.01,0.09493119716644287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,1,power_law_1.01,1.0984448432922362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,8,power_law_1.01,0.1056704044342041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,8,power_law_1.01,0.12403839826583862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,1,power_law_1.01,1.4318911552429199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,1,power_law_1.2,7.219481658935547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,8,power_law_1.01,0.14911999702453613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,8,power_law_1.01,0.1895359992980957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,1,power_law_1.01,1.7671167373657226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,128,power_law_1.2,2.9742080688476564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,8,power_law_1.01,0.2510080099105835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,8,power_law_1.01,0.3207616090774536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,1,power_law_1.01,2.7540735244750976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,8,power_law_1.01,0.3633343935012817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,1,power_law_1.2,8.36135025024414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,8,power_law_1.01,0.5114943981170654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,8,power_law_1.01,0.790771198272705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,1,power_law_1.01,5.391398239135742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,8,power_law_1.01,0.7956543922424316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,1,power_law_1.2,10.686016082763672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,8,power_law_1.01,1.3612223625183106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,8,power_law_1.01,2.589740753173828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,2,32,power_law_1.01,0.02369280010461807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,2,32,power_law_1.01,0.0220223993062973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,2,32,power_law_1.01,0.02263039946556091
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,2,32,power_law_1.01,0.02300799936056137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,1,power_law_1.2,19.92310333251953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,2,32,power_law_1.01,0.02250880002975464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,2,32,power_law_1.01,0.02481919974088669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,2,32,power_law_1.01,0.025798401236534117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,2,32,power_law_1.01,0.02457599937915802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,2,32,power_law_1.01,0.027929601073265076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,32,power_law_1.2,0.08639360070228577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,2,32,power_law_1.01,0.028204798698425293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,2,32,power_law_1.01,0.029868799448013305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,32,power_law_1.2,0.05420799851417542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,2,32,power_law_1.01,0.030643200874328612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,32,power_law_1.2,0.04775039851665497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,2,32,power_law_1.01,0.030630400776863097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,32,power_law_1.2,0.04880639910697937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,2,32,power_law_1.01,0.03089280128479004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,32,power_law_1.2,0.04955520033836365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,2,32,power_law_1.01,0.03392640054225922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,2,32,power_law_1.01,0.03429119884967804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,32,power_law_1.2,0.056428802013397214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,2,32,power_law_1.01,0.03591679930686951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,32,power_law_1.2,0.057708799839019775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,2,32,power_law_1.01,0.0392192006111145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,32,power_law_1.2,0.05672320127487183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,32,power_law_1.01,0.04353919923305512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,32,power_law_1.2,0.06214399933815003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,32,power_law_1.01,0.04679040014743805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,32,power_law_1.01,0.07467520236968994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,32,power_law_1.2,0.06245120167732239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,32,power_law_1.01,0.05057280063629151
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,32,power_law_1.2,0.063372802734375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,32,power_law_1.01,0.060249602794647215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,32,power_law_1.01,0.09696639776229858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,32,power_law_1.2,0.06385279893875122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,32,power_law_1.01,0.06690559983253479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,32,power_law_1.2,0.06447359919548035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,32,power_law_1.01,0.09422079920768738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,32,power_law_1.01,0.08163840174674988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,32,power_law_1.2,0.07248640060424805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,32,power_law_1.01,0.11448320150375366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,32,power_law_1.2,0.07770879864692688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,2,power_law_1.2,0.032627201080322264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,32,power_law_1.01,0.13500159978866577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,32,power_law_1.01,0.06169599890708923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,32,power_law_1.2,0.07832319736480713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,2,power_law_1.2,0.0381632000207901
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,32,power_law_1.2,0.08145920038223267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,32,power_law_1.01,0.1883520007133484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,32,power_law_1.01,0.06578559875488281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,2,power_law_1.2,0.04043520092964172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,32,power_law_1.2,0.08903040289878845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,32,power_law_1.2,0.09924479722976684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,32,power_law_1.01,0.2153343915939331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,2,power_law_1.2,0.04533120095729828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,32,power_law_1.01,0.07252479791641235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,32,power_law_1.2,0.11430399417877198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,2,power_law_1.2,0.05420799851417542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,32,power_law_1.2,0.1249343991279602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,32,power_law_1.01,0.324070405960083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,2,power_law_1.2,0.0744704008102417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,32,power_law_1.01,0.07200639843940734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,32,power_law_1.2,0.16134400367736818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,2,power_law_1.2,0.08011519908905029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,32,power_law_1.2,0.2082304000854492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,32,power_law_1.01,0.07685760259628296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,2,power_law_1.2,0.08228480219841003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,32,power_law_1.2,0.2705343961715698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,32,power_law_1.01,0.6132736206054688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,2,power_law_1.2,0.08392320275306701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,32,power_law_1.01,0.07864959836006165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,32,power_law_1.2,0.36694400310516356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,2,power_law_1.2,0.08589439988136291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,32,power_law_1.01,0.0796992003917694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,2,power_law_1.2,0.08958719968795777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,32,power_law_1.2,0.49740161895751955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,2,power_law_1.2,0.09190400242805481
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,32,power_law_1.01,0.08263040184974671
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,32,power_law_1.2,0.7480063915252686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,2,power_law_1.2,0.09203839898109437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,2,power_law_1.2,0.09363840222358703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,32,power_law_1.01,0.0813696026802063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,32,power_law_1.2,0.9542400360107421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,2,power_law_1.2,0.10071680545806885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,32,power_law_1.01,0.08097919821739197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,2,power_law_1.2,0.10440959930419921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,2,power_law_1.2,0.12629760503768922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,32,power_law_1.2,1.5544639587402345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,32,power_law_1.01,0.0856000006198883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,2,power_law_1.2,0.14885120391845702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,2,power_law_1.2,0.142739200592041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,32,power_law_1.01,0.091430401802063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,2,power_law_1.2,0.17328640222549438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,32,power_law_1.01,0.0948095977306366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,2,power_law_1.2,0.19032319784164428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,32,power_law_1.2,2.8740415573120117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,2,power_law_1.2,0.26201601028442384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,32,power_law_1.01,0.10013439655303955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,2,power_law_1.2,0.320467209815979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,32,power_law_1.01,0.11294080018997192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,2,power_law_1.2,0.4385024070739746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,8,power_law_1.2,0.047679999470710756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,32,power_law_1.01,0.12149759531021118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,2,power_law_1.2,0.5597184181213379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,8,power_law_1.2,0.04230400025844574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,8,power_law_1.2,0.041247999668121337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,2,power_law_1.2,0.7764224052429199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,32,power_law_1.01,0.1495743989944458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,8,power_law_1.2,0.045926401019096376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,8,power_law_1.2,0.04907520115375519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,32,power_law_1.01,0.17020800113677978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,2,power_law_1.2,1.0747648239135743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,8,power_law_1.2,0.06056320071220398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,8,power_law_1.2,0.0633791983127594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,8,power_law_1.2,0.06476160287857055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,32,power_law_1.01,0.22565760612487792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,2,power_law_1.2,1.331065559387207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,8,power_law_1.2,0.06439039707183838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,8,power_law_1.2,0.0661184012889862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,32,power_law_1.01,0.27441279888153075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,8,power_law_1.2,0.06784639954566955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,8,power_law_1.2,0.0695360004901886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,2,power_law_1.2,2.1448192596435547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,8,power_law_1.2,0.06871680021286011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,32,power_law_1.01,0.36944639682769775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,8,power_law_1.2,0.07242239713668823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,8,power_law_1.2,0.0781440019607544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,32,balanced,0.06382399797439575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,32,balanced,0.0649599979321162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,32,balanced,0.06363200147946675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,32,balanced,0.07768533130486806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,32,balanced,0.09218133489290874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,32,balanced,0.12365866700808208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,32,power_law_1.01,0.4685823917388916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,32,balanced,0.16988267501195273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,8,power_law_1.2,0.08432000279426574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,32,balanced,0.1539253294467926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,32,balanced,0.15339199701944986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,32,balanced,0.15473600228627524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,32,balanced,0.1625759998957316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,32,balanced,0.1648373305797577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,32,balanced,0.16481600205103555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,32,balanced,0.16717867056528726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,8,power_law_1.2,0.08926079869270324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,32,balanced,0.17062934239705405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,32,balanced,0.17327467600504556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,32,balanced,0.18176533778508505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,32,balanced,0.18222934007644653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,32,balanced,0.19473065932591757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,32,balanced,0.2119093338648478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,8,power_law_1.2,0.09662719964981079
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,32,balanced,0.23466134071350098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,32,balanced,0.2744053403536479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,2,power_law_1.2,3.8217601776123047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,32,balanced,0.3134346604347229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,32,balanced,0.39498666922251385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,32,balanced,0.5315413475036621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,32,power_law_1.01,0.6595136165618897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,8,power_law_1.2,0.11608320474624634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,32,balanced,0.672714630762736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,32,balanced,0.8734400272369385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,32,balanced,1.1222453117370605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,8,power_law_1.2,0.13608959913253785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,32,balanced,1.694757302602132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,8,power_law_1.2,0.15983359813690184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,32,balanced,3.1864585876464844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,32,power_law_1.01,0.8742591857910156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,8,power_law_1.2,0.2099071979522705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,8,power_law_1.2,0.24019200801849366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,8,power_law_1.2,0.35395839214324953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,32,power_law_1.01,1.0644736289978027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,8,power_law_1.2,0.40436477661132814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,8,power_law_1.2,0.7118207931518554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,32,power_law_1.01,1.6548864364624023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,8,power_law_1.2,0.8536831855773925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,8,power_law_1.2,1.1006976127624513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,8,power_law_1.2,1.6013439178466797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,32,power_law_1.01,3.2630336761474608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,32,balanced,0.03620799879233042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,32,balanced,0.031845333675543465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,32,balanced,0.03183466692765554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,32,balanced,0.032245332996050514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,32,balanced,0.03370666752258936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,32,balanced,0.03608533243338267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,32,balanced,0.03417066733042399
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,32,balanced,0.03457066665093104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,32,balanced,0.033946665624777474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,32,balanced,0.03603733330965042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,32,balanced,0.03605866680542628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,32,balanced,0.03577066709597906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,32,balanced,0.03616533428430557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,32,balanced,0.03788266579310099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,32,balanced,0.04070399949947993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,32,balanced,0.040405333042144775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,32,balanced,0.040250666439533234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,32,balanced,0.04637333254019419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,32,balanced,0.042693331837654114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,8,power_law_1.2,3.2148609161376953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,32,balanced,0.052042668064435325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,32,balanced,0.050111999114354454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,32,balanced,0.07066666583220164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,32,balanced,0.07690666615962982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,32,balanced,0.10549333691596985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,32,balanced,0.1167626678943634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,32,balanced,0.15476266543070474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,32,balanced,0.18405866622924805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,1,4,balanced,0.08482133348782857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,1,32,balanced,0.22230400641759238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,1,4,balanced,0.10337066650390625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,1,4,balanced,0.13031466801961264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,1,32,balanced,0.32206400235493976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,1,4,balanced,0.12802132964134216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,1,4,balanced,0.12773866454760233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,1,4,balanced,0.12797333796819052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,1,32,balanced,0.5781386693318685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,1,4,balanced,0.12980266412099203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,1,4,balanced,0.12991467118263245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,1,4,balanced,0.13702399532000223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,1,4,balanced,0.13687466581662497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,1,4,balanced,0.1344586710135142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,1,4,balanced,0.14131733775138855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,1,4,balanced,0.142277330160141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,1,4,balanced,0.14216533303260803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,1,4,balanced,0.15455466508865356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,1,4,balanced,0.15641599893569946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,1,4,balanced,0.15716800093650818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,1,4,balanced,0.20115200678507486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,1,4,balanced,0.19846399625142416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,1,4,balanced,0.26582932472229004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,1,4,balanced,0.2774133284886678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,1,4,balanced,0.37274666627248126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,1,4,balanced,0.45958932240804035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,1,4,balanced,0.6410826841990153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,1,4,balanced,0.8456479708353678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,1,4,balanced,1.2497973442077637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,1,4,balanced,1.5972480773925781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,1,4,balanced,1.9702773094177246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,1,4,balanced,3.123173395792643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,1,4,balanced,6.37388292948405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,4,4,power_law_1.2,0.07047039866447449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,4,4,power_law_1.2,0.09864320158958435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,4,4,power_law_1.2,0.07990400195121765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,4,4,power_law_1.2,0.07567359805107117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,4,4,power_law_1.2,0.0756991982460022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,4,4,power_law_1.2,0.07696639895439147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,4,4,power_law_1.2,0.07953280210494995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,4,4,power_law_1.2,0.07966079711914062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,4,4,power_law_1.2,0.0954688012599945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,4,4,power_law_1.2,0.09329919815063477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,4,4,power_law_1.2,0.08124160170555114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,4,4,power_law_1.2,0.09388800263404846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,4,4,power_law_1.2,0.09949439764022827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,4,4,power_law_1.2,0.09841920137405395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,2,4,power_law_1.2,0.0605567991733551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,4,4,power_law_1.2,0.10676480531692505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,4,4,power_law_1.2,0.108787202835083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,2,4,power_law_1.2,0.05082240104675293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,4,4,power_law_1.2,0.13066879510879517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,4,4,power_law_1.2,0.15852160453796388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,2,4,power_law_1.2,0.04612480103969574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,4,4,power_law_1.2,0.20434560775756835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,2,4,power_law_1.2,0.04834559857845307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,4,4,power_law_1.2,0.2315903902053833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,4,4,power_law_1.2,0.2774143934249878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,64,power_law_1.01,0.06529920101165772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,2,4,power_law_1.2,0.04838399887084961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,4,4,power_law_1.2,0.44095358848571775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,2,4,power_law_1.2,0.04967679977416992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,4,4,power_law_1.2,0.5114687919616699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,2,4,power_law_1.2,0.049856001138687135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,4,4,power_law_1.2,0.6881343841552734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,2,4,power_law_1.2,0.050969600677490234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,2,4,power_law_1.2,0.0532800018787384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,4,4,power_law_1.2,0.9814144134521484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,2,4,power_law_1.2,0.054016000032424925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,64,power_law_1.01,0.06330879926681518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,4,4,power_law_1.2,1.5547327995300293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,2,4,power_law_1.2,0.058336001634597776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,2,4,power_law_1.2,0.060140800476074216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,2,4,power_law_1.2,0.061894398927688596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,4,4,power_law_1.2,2.121683120727539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,2,4,power_law_1.2,0.06712960004806519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,8,balanced,0.049029335379600525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,8,balanced,0.054234668612480164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,2,4,power_law_1.2,0.07742720246315002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,8,balanced,0.07474133372306824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,8,balanced,0.10171733299891154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,8,balanced,0.15256533026695251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,64,power_law_1.01,0.05447040200233459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,8,balanced,0.18364266554514566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,8,balanced,0.185808002948761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,2,4,power_law_1.2,0.07987840175628662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,4,4,power_law_1.2,2.3032320022583006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,8,balanced,0.18250133593877158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,8,balanced,0.18362132708231607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,8,balanced,0.1843093236287435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,8,balanced,0.18682666619618735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,2,4,power_law_1.2,0.08479999899864196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,8,balanced,0.1863306760787964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,8,balanced,0.1892426609992981
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,8,balanced,0.19342933098475137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,8,balanced,0.19802133242289224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,2,4,power_law_1.2,0.10584319829940796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,8,balanced,0.2002453406651815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,8,balanced,0.2044373353322347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,8,balanced,0.22855466604232788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,8,balanced,0.22614399592081705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,2,4,power_law_1.2,0.12258559465408325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,8,balanced,0.27033066749572754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,8,balanced,0.26816533009211224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,4,4,power_law_1.2,3.4732734680175783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,8,balanced,0.39267198244730633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,2,4,power_law_1.2,0.15014400482177734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,8,balanced,0.3673333326975505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,8,balanced,0.5244853496551514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,64,power_law_1.01,0.05497599840164184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,2,4,power_law_1.2,0.18732800483703613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,8,balanced,0.5676213502883911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,8,balanced,0.8610080083211263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,2,4,power_law_1.2,0.23055999279022216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,8,balanced,1.0047679742177327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,2,4,power_law_1.2,0.321395206451416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,8,balanced,1.1402080059051514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,8,balanced,1.8073760668436687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,2,4,power_law_1.2,0.5156352043151855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,2,4,power_law_1.2,0.5499904155731201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,64,power_law_1.01,0.05619840025901794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,8,balanced,3.438981374104818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,4,4,power_law_1.2,8.016831970214843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,2,4,power_law_1.2,1.0061951637268067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,2,4,power_law_1.2,1.0614015579223632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,64,power_law_1.01,0.056934398412704465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,2,4,power_law_1.2,1.3473919868469237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,2,4,power_law_1.2,2.2091392517089843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,64,power_law_1.01,0.058963197469711306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,2,4,power_law_1.2,4.184108734130859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,64,power_law_1.01,0.05974400043487549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,64,power_law_1.01,0.05957760214805603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,64,power_law_1.01,0.06051200032234192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,64,power_law_1.01,0.05996800065040588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,64,power_law_1.01,0.06267520189285278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,32,2,balanced,0.03573333223660787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,32,2,balanced,0.037578667203585304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,32,2,balanced,0.03959999978542328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,64,power_law_1.01,0.06344959735870362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,32,2,balanced,0.03846399982770284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,32,2,balanced,0.039359999199708305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,32,2,balanced,0.03941333293914795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,32,2,balanced,0.039781334499518074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,32,2,balanced,0.041434665520985924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,32,2,balanced,0.03985599925120672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,32,2,balanced,0.03990933299064636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,32,2,balanced,0.04167999823888143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,32,2,balanced,0.04193066557248434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,32,2,balanced,0.043562665581703186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,32,2,balanced,0.04363200068473816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,32,2,balanced,0.049973333875338234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,32,2,balanced,0.05008533100287119
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,32,2,balanced,0.052015999952952065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,32,2,balanced,0.05810666580994924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,32,2,balanced,0.06422933439413707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,32,2,balanced,0.07462933162848155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,32,2,balanced,0.09101866682370503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,32,2,balanced,0.11012267072995503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,32,2,balanced,0.12998400131861368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,32,2,balanced,0.17631999651590982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,32,1,power_law_1.2,0.05187839865684509
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,32,2,balanced,0.22021333376566568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,32,2,balanced,0.3102933367093404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,32,2,balanced,0.4031253258387248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,32,2,balanced,0.4917706648508708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,64,power_law_1.01,0.06401280164718628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,32,2,balanced,0.7537386417388916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,32,2,balanced,1.4549813270568848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,32,1,power_law_1.2,0.05563520193099976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,64,power_law_1.01,0.07029119729995728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,32,1,power_law_1.2,0.06477439999580384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,128,power_law_1.2,0.03503359854221344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,64,power_law_1.01,0.0722495973110199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,32,1,power_law_1.2,0.08047360181808472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,128,power_law_1.2,0.03498240113258362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,32,1,power_law_1.2,0.10392320156097412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,64,power_law_1.01,0.07955200076103211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,128,power_law_1.2,0.034297600388526917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,32,1,power_law_1.2,0.1343808054924011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,128,power_law_1.2,0.03460479974746704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,64,power_law_1.01,0.09540479779243469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,128,power_law_1.2,0.03516800105571747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,32,1,power_law_1.2,0.15277440547943116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,128,power_law_1.2,0.03585279881954193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,64,power_law_1.01,0.10366079807281495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,32,1,power_law_1.2,0.19363839626312257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,128,power_law_1.2,0.035750401020050046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,128,power_law_1.2,0.03682560026645661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,64,power_law_1.01,0.12993279695510865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,32,1,power_law_1.2,0.2040640115737915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,128,power_law_1.2,0.036345601081848145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,32,1,power_law_1.2,0.2115839958190918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,128,power_law_1.2,0.03622399866580963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,64,power_law_1.01,0.1524415969848633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,8,1,balanced,0.054341331124305725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,8,1,balanced,0.09888000289599101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,8,1,balanced,0.12033599615097046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,8,1,balanced,0.1186293363571167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,8,1,balanced,0.1192586620648702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,8,1,balanced,0.11973333358764648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,128,power_law_1.2,0.03683840036392212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,8,1,balanced,0.12149332960446675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,8,1,balanced,0.12275200088818868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,8,1,balanced,0.12101333340009053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,32,1,power_law_1.2,0.22016639709472657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,8,1,balanced,0.12174399693806966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,8,1,balanced,0.12425067027409871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,8,1,balanced,0.12482666969299316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,8,1,balanced,0.12713600198427835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,8,1,balanced,0.13167466719945273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,8,1,balanced,0.1416106621424357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,8,1,balanced,0.1446453332901001
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,8,1,balanced,0.1458026667435964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,8,1,balanced,0.17354132731755575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,8,1,balanced,0.17429866393407187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,8,1,balanced,0.2334559957186381
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,128,power_law_1.2,0.0371071994304657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,8,1,balanced,0.26545600096384686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,8,1,balanced,0.34564268589019775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,8,1,balanced,0.4023040135701497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,8,1,balanced,0.5725653171539307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,64,power_law_1.01,0.19771519899368287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,8,1,balanced,0.7306666374206543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,32,1,power_law_1.2,0.23264639377593993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,8,1,balanced,1.06059734026591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,128,power_law_1.2,0.03772799968719483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,8,1,balanced,1.375754674275716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,8,1,balanced,1.6974132855733235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,128,power_law_1.2,0.037836799025535585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,8,1,balanced,2.6514026323954263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,32,1,power_law_1.2,0.24003839492797852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,64,power_law_1.01,0.20915839672088624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,128,power_law_1.2,0.04132480025291443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,8,1,balanced,5.212832132975261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,32,1,power_law_1.2,0.2545344114303589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,128,power_law_1.2,0.041484799981117246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,128,power_law_1.2,0.04225279986858368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,32,1,power_law_1.2,0.2668031930923462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,64,power_law_1.01,0.2789695978164673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,128,power_law_1.2,0.04447999894618988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,32,1,power_law_1.2,0.27806720733642576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,128,power_law_1.2,0.04609920084476471
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,128,power_law_1.2,0.05127679705619812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,64,power_law_1.01,0.3384063959121704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,32,1,power_law_1.2,0.29709439277648925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,128,power_law_1.2,0.0571008026599884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,128,power_law_1.2,0.0762943983078003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,32,1,power_law_1.2,0.3486848115921021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,128,power_law_1.2,0.09548799991607666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,64,power_law_1.01,0.4642047882080078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,32,1,power_law_1.2,0.39331200122833254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,128,power_law_1.2,0.13225599527359008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,32,1,power_law_1.2,0.4889855861663818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,128,power_law_1.2,0.1557760000228882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,128,power_law_1.2,0.22644479274749757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,32,1,power_law_1.2,0.5827136039733887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,64,power_law_1.01,0.5540927886962891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,128,power_law_1.2,0.3068608045578003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,32,1,power_law_1.2,0.7487103939056396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,8,balanced,0.043749332427978516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,8,balanced,0.04376000165939331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,8,balanced,0.04588800172011057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,2,128,power_law_1.2,0.3757888078689575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,8,balanced,0.04772266745567322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,8,balanced,0.05589866638183594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,8,balanced,0.06564266482988994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,8,balanced,0.06820266445477803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,8,balanced,0.06711466610431671
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,8,balanced,0.0682239979505539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,8,balanced,0.06845333178838094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,8,balanced,0.07001066704591115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,8,balanced,0.0697813332080841
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,8,balanced,0.07307733098665874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,8,balanced,0.07317333420117696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,8,balanced,0.07747733096281688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,8,balanced,0.08005866905053456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,8,balanced,0.08097066481908162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,32,1,power_law_1.2,0.9399423599243164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,8,balanced,0.09078400333722432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,8,balanced,0.0937013328075409
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,8,balanced,0.11640533804893494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,8,balanced,0.11958400408426921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,8,balanced,0.15890666842460632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,8,balanced,0.17760000626246134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,8,balanced,0.24950933456420898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,8,balanced,0.2898719906806946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,2,64,power_law_1.01,0.6964735984802246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,8,balanced,0.3923146724700928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,2,128,power_law_1.2,0.6107967853546142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,8,balanced,0.5003573497136434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,8,balanced,0.6052373250325521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,8,balanced,0.9110346635182699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,32,1,power_law_1.2,1.2661888122558593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,8,balanced,1.730405330657959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,2,128,power_law_1.2,1.1713983535766601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,32,1,power_law_1.2,1.4873663902282714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,2,64,power_law_1.01,1.0446463584899903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,32,1,power_law_1.2,2.112351989746094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,32,1,power_law_1.2,2.7433984756469725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,32,1,power_law_1.2,3.3348415374755858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,2,64,power_law_1.01,2.181318473815918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,8,32,balanced,0.039605334401130676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,32,1,power_law_1.2,5.3683521270751955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,8,32,balanced,0.03573333223660787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,8,32,balanced,0.03549866626660029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,8,32,balanced,0.035717333356539406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,8,32,balanced,0.037605332831541695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,8,32,balanced,0.037808001041412354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,8,32,balanced,0.03748266647259394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,8,32,balanced,0.038058665891488395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,8,32,balanced,0.03799466788768768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,8,32,balanced,0.03950933367013931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,8,32,balanced,0.03991466760635376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,8,32,balanced,0.03977599988381068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,8,32,balanced,0.03948266555865606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,8,32,balanced,0.04165866722663244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,8,32,balanced,0.04400533437728882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,8,32,balanced,0.04577599962552389
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,8,32,balanced,0.04399466514587402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,8,32,balanced,0.046037331223487854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,8,32,balanced,0.049925332268079124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,8,32,balanced,0.05190399785836538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,8,32,balanced,0.05620799958705902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,8,32,balanced,0.06442133088906606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,8,32,balanced,0.07506666580835979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,8,32,balanced,0.092549333969752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,8,32,balanced,0.11318399508794148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,8,32,balanced,0.14436266819636026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,8,32,balanced,0.17508800824483237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,8,32,balanced,0.20780267318089804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,8,32,balanced,0.3086186647415161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,8,32,balanced,0.5618666807810465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,4,power_law_1.01,0.08435199856758117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,32,1,power_law_1.2,10.140940856933593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,4,power_law_1.01,0.12177920341491699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,4,power_law_1.01,0.07572479844093323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,4,power_law_1.01,0.10195200443267823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,32,power_law_1.2,0.10132479667663574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,4,power_law_1.01,0.13544319868087767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,32,power_law_1.2,0.10005120038986207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,32,1,balanced,0.054048001766204834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,32,1,balanced,0.057904000083605446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,32,1,balanced,0.06237333516279856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,32,1,balanced,0.08368000388145447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,32,1,balanced,0.12226133545239766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,32,1,balanced,0.17641067504882812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,32,power_law_1.2,0.05585920214653015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,32,1,balanced,0.18119466304779053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,32,1,balanced,0.1824000080426534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,32,1,balanced,0.18554667631785074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,32,1,balanced,0.18757333358128866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,32,1,balanced,0.19149333238601685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,32,1,balanced,0.19536532958348593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,32,1,balanced,0.2012373407681783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,32,1,balanced,0.20725866158803305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,32,1,balanced,0.21402132511138916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,32,1,balanced,0.22502400477727255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,32,1,balanced,0.24074133237202963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,32,1,balanced,0.28142933050791424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,32,power_law_1.2,0.055731201171875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,32,1,balanced,0.31353066364924115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,32,1,balanced,0.3859359820683797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,32,1,balanced,0.4440426826477051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,4,power_law_1.01,0.19593600034713746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,32,1,balanced,0.6591039896011353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,32,1,balanced,0.7530399958292643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,32,1,balanced,1.1280426979064941
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,32,power_law_1.2,0.06609280109405517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,32,1,balanced,1.4069280624389648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,32,1,balanced,2.073317368825277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,32,power_law_1.2,0.06625279784202576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,32,1,balanced,2.7133652369181314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,32,1,balanced,3.3524160385131836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,8,power_law_1.2,0.05745919942855835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,4,power_law_1.01,0.20817921161651612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,32,power_law_1.2,0.07760639786720276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,32,1,balanced,5.234693209330241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,32,power_law_1.2,0.0794431984424591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,8,power_law_1.2,0.0704255998134613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,32,power_law_1.2,0.07687039971351624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,4,power_law_1.01,0.21395199298858641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,8,power_law_1.2,0.07127040028572082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,32,1,balanced,10.266842524210611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,32,power_law_1.2,0.08176000118255615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,32,power_law_1.2,0.081632000207901
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,8,power_law_1.2,0.06183040142059326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,32,power_law_1.2,0.0820032000541687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,4,power_law_1.01,0.22367360591888427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,8,power_law_1.2,0.06843519806861878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,32,power_law_1.2,0.08476160168647766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,8,power_law_1.2,0.08342400193214417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,32,power_law_1.2,0.08753920197486878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,4,power_law_1.01,0.22738559246063234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,32,power_law_1.2,0.0948415994644165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,8,power_law_1.2,0.08853759765625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,32,power_law_1.2,0.09489279985427856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,8,power_law_1.2,0.09481599926948547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,4,power_law_1.01,0.2307647943496704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,32,power_law_1.2,0.10224640369415283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,32,power_law_1.2,0.11259520053863525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,8,power_law_1.2,0.10088959932327271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,32,power_law_1.2,0.12653440237045288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,4,power_law_1.01,0.23498239517211914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,8,power_law_1.2,0.1003775954246521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,32,power_law_1.2,0.15163520574569703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,8,power_law_1.2,0.10438400506973267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,32,power_law_1.2,0.1643903970718384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,4,power_law_1.01,0.24187519550323486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,8,power_law_1.2,0.10743680000305175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,32,power_law_1.2,0.21066880226135254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,32,balanced,0.05485333502292633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,32,balanced,0.05249066650867462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,32,balanced,0.05072533090909322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,32,balanced,0.05231999854246775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,32,balanced,0.054383998115857445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,32,balanced,0.05622399846712748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,32,balanced,0.06033066908518473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,32,balanced,0.060421332716941833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,32,balanced,0.060309335589408875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,32,balanced,0.06233066817124685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,32,power_law_1.2,0.25813119411468505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,32,balanced,0.06230400005976359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,32,balanced,0.06474133332570393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,8,power_law_1.2,0.10753920078277587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,32,balanced,0.0646613339583079
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,32,balanced,0.06453866759936015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,32,balanced,0.07086400190989177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,32,balanced,0.07274666428565979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,32,balanced,0.07481599847475688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,4,power_law_1.01,0.24921600818634032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,32,balanced,0.0827946662902832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,32,balanced,0.08906132976214091
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,32,balanced,0.10400000214576721
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,32,balanced,0.12293333808581035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,32,balanced,0.15331733226776123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,32,balanced,0.18052800496419272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,32,balanced,0.23496532440185547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,32,balanced,0.2972426613171895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,32,power_law_1.2,0.3466880083084106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,32,balanced,0.4084906578063965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,32,balanced,0.5210560162862142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,8,power_law_1.2,0.11626240015029907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,32,balanced,0.640176018079122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,32,balanced,0.9687039852142334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,32,balanced,1.8994026184082031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,32,power_law_1.2,0.432755184173584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,4,power_law_1.01,0.25876479148864745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,8,power_law_1.2,0.1236799955368042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,32,power_law_1.2,0.667910385131836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,8,power_law_1.2,0.1250815987586975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,4,power_law_1.01,0.26572160720825194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,32,power_law_1.2,0.80697603225708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,8,power_law_1.2,0.13986560106277465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,8,power_law_1.2,0.1558527946472168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,32,power_law_1.2,0.9958847999572754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,4,power_law_1.01,0.29049599170684814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,8,power_law_1.2,0.16492799520492554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,32,power_law_1.2,1.504742431640625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,8,power_law_1.2,0.19143680334091187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,4,power_law_1.01,0.32882559299468994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,8,power_law_1.2,0.2228991985321045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,4,power_law_1.01,0.3436095952987671
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,8,power_law_1.2,0.2765631914138794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,32,power_law_1.2,3.438617706298828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,8,power_law_1.2,0.3555583953857422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,4,power_law_1.01,0.4015615940093994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,1,power_law_1.01,0.047065600752830505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,8,power_law_1.2,0.46639361381530764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,1,power_law_1.01,0.05774719715118408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,1,power_law_1.01,0.06881279945373535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,4,power_law_1.01,0.45697917938232424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,1,power_law_1.01,0.08806399703025818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,8,power_law_1.2,0.5912320137023925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,1,power_law_1.01,0.10855679512023926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,1,power_law_1.01,0.1510975956916809
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,1,power_law_1.01,0.1579200029373169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,8,power_law_1.2,0.8240192413330079
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,1,power_law_1.01,0.16690560579299926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,4,power_law_1.01,0.5746047973632813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,1,power_law_1.01,0.16778240203857422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,1,power_law_1.01,0.17420799732208253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,8,power_law_1.2,1.0673727989196777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,1,power_law_1.01,0.17992960214614867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,1,power_law_1.01,0.1851583957672119
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,4,power_law_1.01,0.6237120151519775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,1,power_law_1.01,0.19119999408721924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,1,power_law_1.01,0.19186559915542603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,8,power_law_1.2,1.342182445526123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,1,power_law_1.01,0.20222079753875732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,1,power_law_1.01,0.21071999073028563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,1,power_law_1.01,0.22427520751953126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,4,power_law_1.01,0.8059264183044433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,1,power_law_1.01,0.26087040901184083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,2,power_law_1.01,0.04044159948825836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,8,power_law_1.2,2.027667236328125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,1,power_law_1.01,0.28627839088439944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,2,power_law_1.01,0.040089601278305055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,1,power_law_1.01,0.3490943908691406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,2,power_law_1.01,0.042368000745773314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,1,power_law_1.01,0.41182718276977537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,2,power_law_1.01,0.05114240050315857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,1,power_law_1.01,0.5510015964508057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,4,power_law_1.01,1.06048641204834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,2,power_law_1.01,0.05596799850463867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,1,power_law_1.01,0.6753791809082031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,2,power_law_1.01,0.07694720029830933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,2,power_law_1.01,0.07586560249328614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,1,power_law_1.01,0.9295040130615234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,8,power_law_1.2,4.090201568603516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,2,power_law_1.01,0.07772160172462464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,2,power_law_1.01,0.07743359804153442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,1,power_law_1.01,1.176921558380127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,2,power_law_1.01,0.07949439883232116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,4,power_law_1.01,1.4769791603088378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,2,power_law_1.01,0.08038399815559387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,1,power_law_1.01,1.6825279235839843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,2,power_law_1.01,0.08120959997177124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,2,power_law_1.01,0.0824895977973938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,2,power_law_1.01,0.08577920198440551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,1,power_law_1.01,2.1877056121826173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,2,power_law_1.01,0.09385600090026855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,2,power_law_1.01,0.09491199851036072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,2,power_law_1.01,0.10014079809188843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,1,power_law_1.01,2.694816017150879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,4,power_law_1.01,1.713465690612793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,2,power_law_1.01,0.11786880493164062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,2,power_law_1.01,0.13299839496612548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,2,power_law_1.01,0.15429120063781737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,2,power_law_1.01,0.1782464027404785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,1,power_law_1.01,4.157740783691406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,2,power_law_1.01,0.2460416078567505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,2,power_law_1.01,0.2890752077102661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,2,power_law_1.01,0.3852735996246338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,4,4,power_law_1.01,2.137107276916504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,2,power_law_1.01,0.5085055828094482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,2,power_law_1.01,0.6930304050445557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,1,power_law_1.01,8.127731323242188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,1,2,balanced,0.0848479966322581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,1,2,balanced,0.19793067375818887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,1,2,balanced,0.23932266235351562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,1,2,balanced,0.23720000187555948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,1,2,balanced,0.2379306753476461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,2,power_law_1.01,0.9574080467224121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,1,2,balanced,0.23761065800984701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,1,2,balanced,0.23819732666015625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,1,2,balanced,0.23837333917617798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,1,2,balanced,0.24755734205245972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,1,2,balanced,0.2498613397280375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,1,2,balanced,0.24506133794784546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,1,2,balanced,0.2458933393160502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,1,2,balanced,0.24868800242741904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,1,2,balanced,0.26314665873845416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,1,2,balanced,0.2628320058186849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,1,2,balanced,0.26661866903305054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,1,2,balanced,0.2788693308830261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,1,2,balanced,0.34990398089090985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,1,2,balanced,0.34058666229248047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,1,2,balanced,0.42024532953898114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,2,power_law_1.01,1.222156810760498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,1,2,balanced,0.4809226592381795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,1,2,balanced,0.6525173187255859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,1,2,balanced,0.8015893300374349
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,1,2,balanced,1.156890630722046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,1,2,balanced,1.4654560089111328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,1,2,balanced,2.1826559702555337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,4,4,power_law_1.01,3.0921024322509765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,2,power_law_1.01,1.9101119995117188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,1,2,balanced,2.8662665685017905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,1,2,balanced,3.5312372843424478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,1,2,balanced,5.653450647989909
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,2,power_law_1.01,3.7460609436035157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,1,2,balanced,11.303061167399088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,16,power_law_1.01,0.04084480106830597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,4,4,power_law_1.01,6.445465850830078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,16,power_law_1.01,0.03506560027599335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,16,power_law_1.01,0.033292800188064575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,16,power_law_1.01,0.033478400111198424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,16,power_law_1.01,0.034508800506591795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,16,power_law_1.01,0.03485440015792847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,16,power_law_1.01,0.03540480136871338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,16,power_law_1.01,0.03527680039405823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,16,power_law_1.01,0.03543039858341217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,16,power_law_1.01,0.03583999872207642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,16,power_law_1.01,0.036582401394844054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,16,power_law_1.01,0.03733119964599609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,16,power_law_1.01,0.03749760091304779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,16,power_law_1.01,0.03903999924659729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,16,power_law_1.01,0.04266240000724793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,8,1,power_law_1.2,0.038361600041389464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,8,1,power_law_1.2,0.05914239883422852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,16,power_law_1.01,0.04379520118236542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,8,1,power_law_1.2,0.056883198022842404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,8,1,power_law_1.2,0.07448319792747497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,16,power_law_1.01,0.04716159999370575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,8,1,power_law_1.2,0.074508798122406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,16,power_law_1.01,0.04578559994697571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,8,1,power_law_1.2,0.08087679743766785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,8,1,power_law_1.2,0.0828607976436615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,8,1,power_law_1.2,0.0830847978591919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,8,1,power_law_1.2,0.08447359800338745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,16,power_law_1.01,0.05072640180587769
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,8,1,power_law_1.2,0.08543360233306885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,8,1,power_law_1.2,0.08416000008583069
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,16,power_law_1.01,0.0562175989151001
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,8,1,power_law_1.2,0.0889024019241333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,16,power_law_1.01,0.052934402227401735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,8,1,power_law_1.2,0.09029120206832886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,8,1,power_law_1.2,0.09436799883842469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,8,1,power_law_1.2,0.10294400453567505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,16,power_law_1.01,0.05939840078353882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,8,1,power_law_1.2,0.10899200439453124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,8,1,power_law_1.2,0.11091840267181396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,8,1,power_law_1.2,0.13159040212631226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,16,power_law_1.01,0.054067200422286986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,16,power_law_1.01,0.06700800061225891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,8,1,power_law_1.2,0.13575680255889894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,8,1,power_law_1.2,0.16501760482788086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,8,1,power_law_1.2,0.18899199962615967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,16,power_law_1.01,0.07928959727287292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,8,1,power_law_1.2,0.24807679653167725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,8,1,power_law_1.2,0.29282560348510744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,16,power_law_1.01,0.04559360146522522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,8,1,power_law_1.2,0.3946943998336792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,16,power_law_1.01,0.09445760250091553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,8,1,power_law_1.2,0.5028287887573242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,16,power_law_1.2,0.060198402404785155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,8,1,power_law_1.2,0.7132351875305176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,16,power_law_1.01,0.12471040487289428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,8,1,power_law_1.2,0.895257568359375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,16,power_law_1.01,0.04721280038356781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,16,power_law_1.01,0.16308480501174927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,8,1,power_law_1.2,1.0987839698791504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,16,power_law_1.2,0.11860480308532714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,8,1,power_law_1.2,1.6947008132934571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,16,power_law_1.01,0.22620160579681398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,16,power_law_1.01,0.05058559775352478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,16,power_law_1.2,0.05687680244445801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,16,power_law_1.01,0.2879744052886963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,8,1,power_law_1.2,3.3137535095214843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,16,power_law_1.01,0.052339202165603636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,16,power_law_1.2,0.05924479961395264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,16,power_law_1.01,0.35635199546813967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,16,power_law_1.01,0.055232000350952146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,16,power_law_1.2,0.06509439945220948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,16,power_law_1.01,0.5314432144165039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,16,power_law_1.01,0.05617920160293579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,16,power_law_1.2,0.06873599886894226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,16,power_law_1.01,0.056441599130630495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,16,power_law_1.2,0.07982079982757569
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,16,power_law_1.01,0.9658304214477539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,16,power_law_1.01,0.059571200609207155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,16,power_law_1.2,0.08376320004463196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,16,power_law_1.01,0.06042240262031555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,16,power_law_1.2,0.08202239871025085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,16,power_law_1.01,0.06263039708137512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,16,power_law_1.2,0.08117759823799134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,16,power_law_1.01,0.06483839750289917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,16,power_law_1.2,0.08395519852638245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,16,power_law_1.01,0.0726527988910675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,16,power_law_1.2,0.08646399974822998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,16,power_law_1.01,0.07774080038070678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,16,power_law_1.2,0.08910719752311706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,16,power_law_1.01,0.08492799997329711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,32,4,power_law_1.2,0.05729920268058777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,16,power_law_1.2,0.09037439823150635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,32,4,power_law_1.2,0.06282879710197449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,16,power_law_1.01,0.10197759866714477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,16,power_law_1.2,0.09895679950714112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,32,4,power_law_1.2,0.056505602598190305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,16,power_law_1.01,0.11112960577011108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,16,power_law_1.2,0.09964159727096558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,32,4,power_law_1.2,0.06188160181045532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,16,power_law_1.01,0.14254720211029054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,32,4,power_law_1.2,0.06699519753456115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,16,power_law_1.2,0.10629760026931763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,32,4,power_law_1.2,0.07401599884033203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,16,power_law_1.01,0.15932159423828124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,128,power_law_1.01,0.07439360022544861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,32,4,power_law_1.2,0.08476799726486206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,16,power_law_1.01,0.21249918937683104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,16,power_law_1.2,0.11963520050048829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,128,power_law_1.01,0.07448959946632386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,32,4,power_law_1.2,0.08810880184173583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,32,4,power_law_1.2,0.08760319948196411
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,16,power_law_1.01,0.26591360569000244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,128,power_law_1.01,0.07128319740295411
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,16,power_law_1.2,0.13464959859848022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,32,4,power_law_1.2,0.09098240137100219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,128,power_law_1.01,0.06617599725723267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,16,power_law_1.01,0.3675839900970459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,32,4,power_law_1.2,0.09344639778137206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,32,4,power_law_1.2,0.09767040014266967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,16,power_law_1.2,0.15461119413375854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,128,power_law_1.01,0.06289280056953431
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,16,power_law_1.01,0.4484096050262451
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,32,4,power_law_1.2,0.10145920515060425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,128,power_law_1.01,0.07046399712562561
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,16,power_law_1.2,0.18023680448532103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,32,4,power_law_1.2,0.10428800582885742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,16,power_law_1.01,0.6643904209136963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,32,4,power_law_1.2,0.11729919910430908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,128,power_law_1.01,0.06819199919700622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,16,power_law_1.2,0.2269439935684204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,32,4,power_law_1.2,0.11706880331039429
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,16,power_law_1.01,0.8354111671447754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,128,power_law_1.01,0.06692479848861695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,32,4,power_law_1.2,0.12441600561141967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,16,power_law_1.2,0.2617408037185669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,128,power_law_1.01,0.07006080150604248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,32,4,power_law_1.2,0.1414720058441162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,16,power_law_1.01,1.0651328086853027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,16,power_law_1.2,0.36776320934295653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,32,4,power_law_1.2,0.157478404045105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,128,power_law_1.01,0.070169597864151
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,32,4,power_law_1.2,0.19276800155639648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,128,power_law_1.01,0.07123200297355652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,16,power_law_1.2,0.4414656162261963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,16,power_law_1.01,1.685919952392578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,32,4,power_law_1.2,0.21967360973358155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,128,power_law_1.01,0.07281280159950257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,32,4,power_law_1.2,0.2840320110321045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,16,power_law_1.2,0.6212224006652832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,32,4,power_law_1.2,0.3432320117950439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,128,power_law_1.01,0.07111679911613464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,32,4,power_law_1.2,0.5343488216400146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,16,power_law_1.2,0.920633602142334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,128,power_law_1.01,0.07734400033950806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,16,power_law_1.01,3.3399425506591798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,32,4,power_law_1.2,0.6144832134246826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,128,power_law_1.01,0.08110719919204712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,16,power_law_1.2,1.0636287689208985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,32,4,power_law_1.2,0.8958335876464844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,128,power_law_1.01,0.07998719811439514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,128,power_law_1.01,0.08919039964675904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,32,4,power_law_1.2,1.2697216033935548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,16,power_law_1.2,1.5518719673156738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,128,power_law_1.01,0.09537919759750366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,32,4,power_law_1.2,1.485977554321289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,128,power_law_1.01,0.10873600244522094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,128,power_law_1.01,0.13189760446548462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,16,power_law_1.2,3.6149696350097655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,32,4,power_law_1.2,2.493440055847168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,128,power_law_1.01,0.15044480562210083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,128,power_law_1.01,0.18724479675292968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,32,4,power_law_1.2,4.832255935668945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,128,power_law_1.01,0.22491519451141356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,1,4,power_law_1.01,0.07134720087051391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,1,4,power_law_1.01,0.08849920034408569
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,1,4,power_law_1.01,0.06967039704322815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,1,4,power_law_1.01,0.07090560197830201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,128,power_law_1.01,0.31400959491729735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,1,4,power_law_1.01,0.07305600047111512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,1,4,power_law_1.01,0.07423999905586243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,1,4,power_law_1.01,0.07420160174369812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,1,4,power_law_1.01,0.07333760261535645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,1,4,power_law_1.01,0.07580800056457519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,1,4,power_law_1.01,0.07497599720954895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,1,4,power_law_1.01,0.07715839743614197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,1,4,power_law_1.01,0.08064000010490417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,128,power_law_1.01,0.3899264097213745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,1,4,power_law_1.01,0.08295040130615235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,1,4,power_law_1.01,0.09064319729804993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,1,4,power_law_1.01,0.10211199522018433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,1,4,power_law_1.01,0.10198400020599366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,1,4,power_law_1.01,0.12328319549560547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,1,4,power_law_1.01,0.13757439851760864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,1,4,power_law_1.01,0.1752128005027771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,1,4,power_law_1.01,0.22371840476989746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,128,power_law_1.01,0.5418176174163818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,1,4,power_law_1.01,0.25772800445556643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,1,4,power_law_1.01,0.3369343996047974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,1,4,power_law_1.01,0.4555776119232178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,1,4,power_law_1.01,0.6967487812042237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,1,4,power_law_1.01,1.0887231826782227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,128,power_law_1.01,0.7086336135864257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,1,4,power_law_1.01,1.546835231781006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,1,4,power_law_1.01,1.764396858215332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,16,1,power_law_1.01,0.034790399670600894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,1,4,power_law_1.01,2.0380672454833983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,16,1,power_law_1.01,0.03831680119037628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,128,power_law_1.01,0.8849727630615234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,16,1,power_law_1.01,0.04360319972038269
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,16,1,power_law_1.01,0.05608320236206055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,1,4,power_law_1.01,4.241427230834961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,16,1,power_law_1.01,0.057036799192428586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,16,1,power_law_1.01,0.05854719877243042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,16,1,power_law_1.01,0.0617792010307312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,16,1,power_law_1.01,0.06028159856796265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,128,power_law_1.01,1.3367296218872071
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,16,1,power_law_1.01,0.061887997388839724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,1,4,power_law_1.01,5.510752105712891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,16,1,power_law_1.01,0.0634112000465393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,2,4,balanced,0.041738669077555336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,2,4,balanced,0.039674667020638786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,16,1,power_law_1.01,0.06352639794349671
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,2,4,balanced,0.06047999858856201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,2,4,balanced,0.061754668752352394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,2,4,balanced,0.06002666552861532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,2,4,balanced,0.06238399942715963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,2,4,balanced,0.061333333452542625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,2,4,balanced,0.06202666461467743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,16,1,power_law_1.01,0.06948480010032654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,2,4,balanced,0.06253866851329803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,2,4,balanced,0.062309334675470986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,2,4,balanced,0.0644160012404124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,2,4,balanced,0.06388266881306966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,2,4,balanced,0.06433066725730896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,2,4,balanced,0.06855999926726024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,16,1,power_law_1.01,0.07128319740295411
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,2,4,balanced,0.07675200204054515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,2,4,balanced,0.07509333391984303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,2,4,balanced,0.08133333424727122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,2,4,balanced,0.09152000149091084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,2,4,balanced,0.09682133793830872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,16,1,power_law_1.01,0.07032319903373718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,2,4,balanced,0.1193386713663737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,2,4,balanced,0.1288159986337026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,2,4,balanced,0.17037334044774374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,2,4,balanced,0.2049973408381144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,16,1,power_law_1.01,0.08267520070075988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,2,4,balanced,0.27804799874623615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,2,4,balanced,0.35443735122680664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,2,4,balanced,0.5019893248875936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,16,1,power_law_1.01,0.08436480164527893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,2,4,balanced,0.6413600047429403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,2,4,balanced,0.7912906805674235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,16,1,power_law_1.01,0.09612159729003907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,2,4,balanced,1.2207039992014568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,16,1,power_law_1.01,0.11295360326766968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,16,1,power_law_1.01,0.11872639656066894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,2,4,balanced,2.367525259653727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,16,1,power_law_1.01,0.14935040473937988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,128,power_law_1.01,2.77139835357666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,16,1,power_law_1.01,0.1724671959877014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,16,1,power_law_1.01,0.2332927942276001
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,16,1,power_law_1.01,0.28523519039154055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,16,1,power_law_1.01,0.40227842330932617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,16,1,power_law_1.01,0.5145343780517578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,16,1,power_law_1.01,0.7348159790039063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,16,1,power_law_1.01,0.9499520301818848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,16,1,power_law_1.01,1.1753151893615723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,16,1,power_law_1.01,1.8262271881103516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,16,1,power_law_1.01,3.584486389160156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,64,power_law_1.01,0.05912320017814636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,64,power_law_1.01,0.05743359923362732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,64,power_law_1.01,0.04809600114822388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,64,power_law_1.01,0.04898560047149658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,32,power_law_1.2,0.04340479969978332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,2,power_law_1.01,0.03758719861507416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,64,power_law_1.01,0.049081599712371825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,32,power_law_1.2,0.04226559996604919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,2,power_law_1.01,0.03682560026645661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,32,power_law_1.2,0.03607040047645569
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,2,power_law_1.01,0.03988479971885681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,64,power_law_1.01,0.04996480047702789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,32,power_law_1.2,0.03420799970626831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,2,power_law_1.01,0.04807040095329285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,32,power_law_1.2,0.03444480001926422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,64,power_law_1.01,0.05128960013389587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,2,power_law_1.01,0.05280640125274658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,32,power_law_1.2,0.03560320138931274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,2,power_law_1.01,0.07215999960899352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,64,power_law_1.01,0.05177599787712097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,32,power_law_1.2,0.0366784006357193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,2,power_law_1.01,0.07500799894332885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,32,power_law_1.2,0.035955199599266054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,2,power_law_1.01,0.0791167974472046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,64,power_law_1.01,0.05182719826698303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,32,power_law_1.2,0.036211198568344115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,2,power_law_1.01,0.07899519801139832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,32,power_law_1.2,0.03619840145111084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,64,power_law_1.01,0.05287039875984192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,2,16,power_law_1.2,0.028806400299072266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,2,power_law_1.01,0.08182399868965148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,32,power_law_1.2,0.03678719997406006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,2,power_law_1.01,0.08417919874191285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,32,power_law_1.2,0.037747201323509214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,64,power_law_1.01,0.053452801704406736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,2,power_law_1.01,0.08757759928703308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,2,16,power_law_1.2,0.024639999866485594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,32,power_law_1.2,0.038406398892402646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,64,power_law_1.01,0.054745602607727054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,2,power_law_1.01,0.0889792025089264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,32,power_law_1.2,0.03978239893913269
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,2,16,power_law_1.2,0.025446400046348572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,2,power_law_1.01,0.09373440146446228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,32,power_law_1.2,0.04333440065383911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,64,power_law_1.01,0.05440000295639038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,2,power_law_1.01,0.09996160268783569
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,32,power_law_1.2,0.04405759871006012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,2,16,power_law_1.2,0.025830399990081788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,2,power_law_1.01,0.10561280250549317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,64,power_law_1.01,0.05668479800224304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,32,power_law_1.2,0.04705919921398163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,2,power_law_1.01,0.11048320531845093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,32,power_law_1.2,0.051795202493667605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,2,16,power_law_1.2,0.026348799467086792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,64,power_law_1.01,0.06135680079460144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,2,power_law_1.01,0.12239999771118164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,32,power_law_1.2,0.05546879768371582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,2,power_law_1.01,0.13790080547332764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,2,16,power_law_1.2,0.026316800713539125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,32,power_law_1.2,0.06738560199737549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,64,power_law_1.01,0.06320639848709106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,2,power_law_1.01,0.1642624020576477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,32,power_law_1.2,0.06934400200843811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,2,16,power_law_1.2,0.026240000128746034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,64,power_law_1.01,0.06567680239677429
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,2,power_law_1.01,0.18706560134887695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,32,power_law_1.2,0.09337599873542786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,2,power_law_1.01,0.2473344087600708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,2,16,power_law_1.2,0.03249920010566711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,32,power_law_1.2,0.1161344051361084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,64,power_law_1.01,0.07272319793701172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,2,power_law_1.01,0.2932607889175415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,32,power_law_1.2,0.1466815948486328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,2,16,power_law_1.2,0.03400320112705231
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,64,power_law_1.01,0.07959679961204529
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,2,power_law_1.01,0.4471295833587646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,32,power_law_1.2,0.1951359987258911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,2,16,power_law_1.2,0.034771201014518735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,64,power_law_1.01,0.09427840113639832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,2,power_law_1.01,0.5230016231536865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,32,power_law_1.2,0.26642560958862305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,2,16,power_law_1.2,0.03639039993286133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,64,power_law_1.01,0.11238399744033814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,32,power_law_1.2,0.33834240436553953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,2,power_law_1.01,0.7725376129150391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,2,16,power_law_1.2,0.03607040047645569
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,64,power_law_1.01,0.1382848024368286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,2,16,power_law_1.2,0.037376001477241516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,2688,1856,6,128,1,32,power_law_1.2,0.40782718658447265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,2,power_law_1.01,1.0757887840270997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,64,power_law_1.01,0.16943360567092897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,2,16,power_law_1.2,0.03587839901447296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,2688,1856,6,128,1,32,power_law_1.2,0.6330624103546143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,2,power_law_1.01,1.2746560096740722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,2,16,power_law_1.2,0.040601599216461184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,64,power_law_1.01,0.22453761100769043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,2,16,power_law_1.2,0.04261119961738587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,64,power_law_1.01,0.2757055997848511
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,2,16,power_law_1.2,0.04431999921798706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,2,power_law_1.01,1.905900764465332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,2688,1856,6,128,1,32,power_law_1.2,1.4770943641662597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,2,16,power_law_1.2,0.04700160026550293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,64,power_law_1.01,0.39443199634552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,16,power_law_1.2,0.04917759895324707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,16,power_law_1.2,0.05619199872016907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,64,power_law_1.01,0.5213312149047852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,16,power_law_1.2,0.06049280166625977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,2,power_law_1.01,3.9744895935058593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,16,power_law_1.2,0.08141440153121948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,64,power_law_1.01,0.606553602218628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,16,power_law_1.2,0.09319679737091065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,16,power_law_1.2,0.1368127942085266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,64,power_law_1.01,0.9663935661315918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,16,power_law_1.2,0.17044479846954347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,16,power_law_1.2,0.20449280738830566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,16,power_law_1.2,0.2801151990890503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,64,power_law_1.01,1.9205951690673828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,16,power_law_1.2,0.3974208116531372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,16,power_law_1.2,0.599123191833496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,2,power_law_1.01,0.04293760061264038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,2,power_law_1.01,0.03902080059051514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,2,power_law_1.01,0.04145280122756958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,16,power_law_1.2,1.2935104370117188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,2,power_law_1.01,0.05063040256500244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,2,power_law_1.01,0.05358080267906189
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,2,power_law_1.01,0.07022719979286193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,4,power_law_1.01,0.0844543993473053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,2,power_law_1.01,0.07319679856300354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,4,power_law_1.01,0.09493119716644287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,2,power_law_1.01,0.0742143988609314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,4,power_law_1.01,0.0763264000415802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,2,power_law_1.01,0.07574399709701538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,2,power_law_1.01,0.0779263973236084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,4,power_law_1.01,0.12077440023422241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,2,power_law_1.01,0.07989119887351989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,4,power_law_1.01,0.1347839951515198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,2,power_law_1.01,0.08177919983863831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,4,power_law_1.01,0.17643519639968872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,2,power_law_1.01,0.08346880078315735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,4,power_law_1.01,0.19467519521713256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,2,power_law_1.01,0.08703359961509705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,4,power_law_1.01,0.20099198818206787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,2,power_law_1.01,0.091839998960495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,32,4,balanced,0.03182933231194814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,32,4,balanced,0.029722665747006733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,32,4,balanced,0.029274667302767437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,32,4,balanced,0.028058665494124096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,32,4,balanced,0.029482667644818623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,32,4,balanced,0.029322666426499683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,32,4,balanced,0.02940800040960312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,32,4,balanced,0.031370667119820915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,32,4,balanced,0.029450667401154835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,32,4,balanced,0.029482667644818623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,2,power_law_1.01,0.09596800208091735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,4,power_law_1.01,0.19336960315704346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,32,4,balanced,0.031397332747777305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,32,4,balanced,0.03330666571855545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,32,4,balanced,0.03156266609827677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,32,4,balanced,0.03139200061559677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,32,4,balanced,0.03754666695992152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,32,4,balanced,0.039605334401130676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,32,4,balanced,0.03969600051641464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,32,4,balanced,0.04162133236726125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,32,4,balanced,0.043605332573254905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,32,4,balanced,0.04780266682306925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,32,4,balanced,0.051829333106676735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,2,power_law_1.01,0.1025920033454895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,32,4,balanced,0.06443733473618825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,32,4,balanced,0.07856533428033192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,32,4,balanced,0.10481066505114238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,4,power_law_1.01,0.2013887882232666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,32,4,balanced,0.1132480005423228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,32,4,balanced,0.15985600153605142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,32,4,balanced,0.18471999963124594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,32,4,balanced,0.23254400491714478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,32,4,balanced,0.3274719913800557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,2,power_law_1.01,0.11742719411849975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,32,4,balanced,0.586677352587382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,4,power_law_1.01,0.20288639068603515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,2,power_law_1.01,0.12603520154953002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,4,power_law_1.01,0.20810239315032958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,2,power_law_1.01,0.14902399778366088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,4,power_law_1.01,0.2116607904434204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,2,power_law_1.01,0.19219199419021607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,4,power_law_1.01,0.21541121006011962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,2,power_law_1.01,0.23914880752563478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,4,power_law_1.01,0.23176960945129393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,2,power_law_1.01,0.2889280080795288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,1,balanced,0.05133866767088572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,4,power_law_1.01,0.229203200340271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,1,balanced,0.05969599882761637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,1,balanced,0.08119999865690868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,2,power_law_1.01,0.3974656105041504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,1,balanced,0.11145066221555074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,1,balanced,0.17478400468826294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,1,balanced,0.1781760056813558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,1,balanced,0.18276800711949667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,4,power_law_1.01,0.24437758922576905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,1,balanced,0.18217599391937256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,1,balanced,0.18413333098093668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,1,balanced,0.191594660282135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,1,balanced,0.18644267320632935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,1,balanced,0.19529600938161215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,2,power_law_1.01,0.5196095943450928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,1,balanced,0.19887999693552652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,1,balanced,0.20326934258143106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,1,balanced,0.2088586688041687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,4,power_law_1.01,0.2634304046630859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,1,balanced,0.21524800856908163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,1,balanced,0.2246933380762736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,1,balanced,0.2562613288561503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,1,balanced,0.2804746627807617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,1,balanced,0.3369226853052775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,1,balanced,0.3856853246688843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,2,power_law_1.01,0.784339189529419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,4,power_law_1.01,0.28739840984344484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,1,balanced,0.5344160000483195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,1,balanced,0.6292693217595419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,1,balanced,0.9012373288472494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,1,balanced,1.1343093713124592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,4,power_law_1.01,0.3727423906326294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,1,balanced,1.6579519907633464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,2,power_law_1.01,0.9464832305908203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,1,balanced,2.1813066800435386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,4,power_law_1.01,0.3807039976119995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,1,balanced,2.6967360178629556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,4,power_law_1.01,0.47771520614624025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,2,power_law_1.01,1.2460864067077637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,1,balanced,4.161493301391602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,4,power_law_1.01,0.5785024166107178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,4,power_law_1.01,0.739302396774292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,1,balanced,8.10047976175944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,2,power_law_1.01,1.9823999404907227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,4,power_law_1.01,0.9802751541137695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,16,power_law_1.2,0.0451200008392334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,4,power_law_1.01,1.2824895858764649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,16,power_law_1.2,0.04386560022830963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,2,power_law_1.01,3.681081771850586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,16,power_law_1.2,0.0341376006603241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,4,power_law_1.01,1.6262016296386719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,16,power_law_1.2,0.034560000896453856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,16,power_law_1.2,0.035673600435256955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,16,power_law_1.2,0.03854719996452331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,4,power_law_1.01,1.949331283569336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,16,power_law_1.2,0.03903999924659729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,16,power_law_1.2,0.03925760090351105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,16,power_law_1.2,0.038796800374984744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,16,power_law_1.2,0.03948160111904144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,4,power_law_1.01,2.9926591873168946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,16,power_law_1.2,0.040038400888442995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,16,power_law_1.2,0.04097920060157776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,16,power_law_1.2,0.04071680009365082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,16,power_law_1.2,0.04381439983844757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,16,power_law_1.2,0.04776319861412048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,16,power_law_1.2,0.049369600415229795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,16,power_law_1.2,0.05177599787712097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,4,power_law_1.01,5.920223999023437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,16,power_law_1.2,0.05809280276298523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,16,power_law_1.2,0.06343680024147033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,16,power_law_1.2,0.06993920207023621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,16,power_law_1.2,0.08014079928398132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,16,power_law_1.2,0.1056831955909729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,8,power_law_1.2,0.058899199962615965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,16,power_law_1.2,0.12976640462875366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,2,1,power_law_1.2,0.06633599996566772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,2,1,power_law_1.2,0.07538560032844543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,16,power_law_1.2,0.17417600154876708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,2,1,power_law_1.2,0.09534080028533935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,2,1,power_law_1.2,0.1358847975730896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,2,1,power_law_1.2,0.1415295958518982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,2,1,power_law_1.2,0.14293760061264038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,16,power_law_1.2,0.23637120723724364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,2,1,power_law_1.2,0.14304640293121337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,2,1,power_law_1.2,0.1446720004081726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,2,1,power_law_1.2,0.14577280282974242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,8,power_law_1.2,0.06446719765663148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,2,1,power_law_1.2,0.1466879963874817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,16,power_law_1.2,0.3106048107147217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,2,1,power_law_1.2,0.149452805519104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,2,1,power_law_1.2,0.15538560152053832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,2,1,power_law_1.2,0.15868159532546997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,2,1,power_law_1.2,0.16373759508132935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,16,power_law_1.2,0.3690239906311035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,2,1,power_law_1.2,0.1779647946357727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,2,1,power_law_1.2,0.1855679988861084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,2,1,power_law_1.2,0.19743360280990602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,2,1,power_law_1.2,0.22860159873962402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,2,1,power_law_1.2,0.2582655906677246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,2688,1856,6,128,1,16,power_law_1.2,0.45401601791381835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,2,1,power_law_1.2,0.3222208023071289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,8,power_law_1.2,0.06384639739990235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,2,1,power_law_1.2,0.3962815999984741
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,2,1,power_law_1.2,0.523360013961792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,2,1,power_law_1.2,0.6163072109222412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,2688,1856,6,128,1,16,power_law_1.2,0.9235903739929199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,2,1,power_law_1.2,0.852511978149414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,2,1,power_law_1.2,1.0790271759033203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,8,power_law_1.2,0.05923839807510376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,2,1,power_law_1.2,1.5286208152770997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,2,1,power_law_1.2,2.002854347229004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,2688,1856,6,128,1,16,power_law_1.2,1.685830307006836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,2,1,power_law_1.2,2.469843292236328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,4,64,power_law_1.01,0.0357120007276535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,8,power_law_1.2,0.06395519971847534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,4,64,power_law_1.01,0.032902398705482484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,2,1,power_law_1.2,3.882175827026367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,4,64,power_law_1.01,0.033267199993133545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,4,64,power_law_1.01,0.032652801275253295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,8,power_law_1.2,0.06852480173110961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,4,64,power_law_1.01,0.03379839956760407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,4,64,power_law_1.01,0.03373439908027649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,2,1,power_law_1.2,7.622393798828125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,4,64,power_law_1.01,0.03410559892654419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,4,64,power_law_1.01,0.03446399867534637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,8,power_law_1.2,0.07012479901313781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,4,64,power_law_1.01,0.03490560054779053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,4,64,power_law_1.01,0.03532159924507141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,4,64,power_law_1.01,0.03548159897327423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,8,power_law_1.2,0.08033279776573181
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,4,64,power_law_1.01,0.03609600067138672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,4,64,power_law_1.01,0.03653120100498199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,4,64,power_law_1.01,0.037215998768806456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,4,64,power_law_1.01,0.04038400053977966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,8,power_law_1.2,0.07978240251541138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,4,64,power_law_1.01,0.039743998646736146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,4,64,power_law_1.01,0.03986560106277466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,4,64,power_law_1.01,0.04248960018157959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,8,power_law_1.2,0.08405759930610657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,4,64,power_law_1.01,0.044249600172042845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,4,64,power_law_1.01,0.04975999891757965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,8,4,balanced,0.03180266668399175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,8,4,balanced,0.032218667368094124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,4,64,power_law_1.01,0.05310080051422119
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,8,4,balanced,0.033759998778502144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,8,4,balanced,0.03181333343187968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,8,4,balanced,0.03375466664632162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,8,4,balanced,0.033728001018365227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,8,power_law_1.2,0.08135679960250855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,8,4,balanced,0.03373866776625315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,8,4,balanced,0.03370666752258936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,8,4,balanced,0.033546666304270424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,8,4,balanced,0.03572800010442734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,8,4,balanced,0.03603200117746989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,8,4,balanced,0.03577066709597906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,8,4,balanced,0.03746666759252548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,4,64,power_law_1.01,0.05854719877243042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,8,4,balanced,0.03716266651948293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,8,4,balanced,0.04498666524887085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,8,4,balanced,0.04402133325735728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,8,4,balanced,0.045007998744646706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,8,4,balanced,0.05390933156013489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,8,4,balanced,0.05588266750176748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,8,4,balanced,0.07531733314196269
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,8,4,balanced,0.07926400005817413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,8,4,balanced,0.10693333546320598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,8,4,balanced,0.13265599807103476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,8,4,balanced,0.17347200711568198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,4,64,power_law_1.01,0.06631039977073669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,8,4,balanced,0.20773333311080933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,8,4,balanced,0.2912320097287496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,8,4,balanced,0.3636213143666585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,8,4,balanced,0.446666677792867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,8,4,balanced,0.6748800277709961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,4,64,power_law_1.01,0.0852735996246338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,8,power_law_1.2,0.08588799834251404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,8,4,balanced,1.2812586625417073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,4,64,power_law_1.01,0.10832639932632446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,32,4,power_law_1.01,0.032601600885391234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,4,64,power_law_1.01,0.153331196308136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,32,4,power_law_1.01,0.0298880010843277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,8,power_law_1.2,0.08785920143127442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,32,4,power_law_1.01,0.0290367990732193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,32,4,power_law_1.01,0.028672000765800475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,32,4,power_law_1.01,0.028825598955154418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,32,4,power_law_1.01,0.02919679880142212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,32,4,power_law_1.01,0.029977598786354066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,4,64,power_law_1.01,0.1785024046897888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,32,4,power_law_1.01,0.02998400032520294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,32,4,power_law_1.01,0.030432000756263733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,32,4,power_law_1.01,0.03160960078239441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,32,4,power_law_1.01,0.0315775990486145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,32,4,power_law_1.01,0.03180800080299377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,8,power_law_1.2,0.09189119935035706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,32,4,power_law_1.01,0.032671999931335446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,32,4,power_law_1.01,0.035692799091339114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,4,64,power_law_1.01,0.2347775936126709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,32,4,power_law_1.01,0.03991039991378784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,32,4,power_law_1.01,0.040320000052452086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,32,4,power_law_1.01,0.04377599954605103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,32,4,power_law_1.01,0.054150402545928955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,32,4,power_law_1.01,0.057081598043441775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,32,4,power_law_1.01,0.06594560146331788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,32,4,power_law_1.01,0.07032960057258605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,8,power_law_1.2,0.09872000217437744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,4,64,power_law_1.01,0.3655616044998169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,32,4,power_law_1.01,0.08952320218086243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,32,4,power_law_1.01,0.11267839670181275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,32,4,power_law_1.01,0.15983359813690184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,32,4,power_law_1.01,0.18688640594482422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,32,4,power_law_1.01,0.34288640022277833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,32,4,power_law_1.01,0.3755136013031006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,8,power_law_1.2,0.10430719852447509
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,32,4,power_law_1.01,0.3806720018386841
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,32,4,power_law_1.01,0.7853184223175049
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,4,64,power_law_1.01,0.7493760108947753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,32,4,power_law_1.01,1.3610624313354491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,8,power_law_1.2,0.11204479932785034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,8,power_law_1.2,0.13370239734649658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,8,power_law_1.2,0.14974720478057862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,16,power_law_1.01,0.03475199937820435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,16,power_law_1.01,0.031014400720596313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,16,power_law_1.01,0.029344001412391664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,8,power_law_1.2,0.18103040456771852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,16,power_law_1.01,0.029292801022529603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,16,power_law_1.01,0.031231999397277832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,16,power_law_1.01,0.0328000009059906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,8,power_law_1.2,0.19296640157699585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,16,power_law_1.01,0.033081600069999696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,16,power_law_1.01,0.034457600116729735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,16,power_law_1.01,0.033983999490737916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,8,power_law_1.2,0.24494080543518065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,16,power_law_1.01,0.034457600116729735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,16,power_law_1.01,0.03621760010719299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,16,power_law_1.01,0.0361407995223999
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,16,power_law_1.01,0.037299200892448425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,8,power_law_1.2,0.30462720394134524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,16,power_law_1.01,0.037484800815582274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,16,power_law_1.01,0.04229759871959686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,16,power_law_1.01,0.04459519982337952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,16,power_law_1.01,0.046367999911308286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,8,power_law_1.2,0.42524161338806155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,16,balanced,0.03147733211517334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,16,balanced,0.0315786674618721
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,16,balanced,0.03137599925200144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,16,balanced,0.033589333295822144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,16,balanced,0.03345066557327906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,16,balanced,0.03566933423280716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,16,balanced,0.03606399893760681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,16,balanced,0.03558400024970373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,16,balanced,0.03542399903138479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,16,balanced,0.03563733398914337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,16,balanced,0.03774933268626531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,16,balanced,0.03779733429352442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,16,power_law_1.01,0.052288001775741576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,16,balanced,0.03731200098991394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,16,balanced,0.03811733424663544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,16,balanced,0.04373333354791006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,16,balanced,0.04164800047874451
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,16,balanced,0.04346133271853129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,16,balanced,0.04560000201066335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,16,balanced,0.049685334165891014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,16,balanced,0.05669866502285004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,16,balanced,0.06018133461475372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,16,balanced,0.07720000048478444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,16,balanced,0.09781333804130554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,16,power_law_1.01,0.058956801891326904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,16,balanced,0.1405333379904429
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,2,power_law_1.01,0.04598399996757507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,16,balanced,0.17762132485707602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,16,balanced,0.24370666344960532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,16,balanced,0.3020266691843669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,16,balanced,0.37302398681640625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,16,power_law_1.01,0.07390720248222352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,16,balanced,0.5548959970474243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,2,power_law_1.01,0.05121279954910278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,16,balanced,1.054533322652181
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,8,power_law_1.2,0.5598720073699951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,16,power_law_1.01,0.08186240196228027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,2,power_law_1.01,0.05397120118141174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,16,power_law_1.01,0.11598720550537109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,2,power_law_1.01,0.0711359977722168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,16,power_law_1.01,0.14440959692001343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,2,power_law_1.01,0.07675520181655884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,8,power_law_1.2,0.7473919868469239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,16,power_law_1.01,0.18495999574661254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,2,power_law_1.01,0.10794240236282349
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,16,power_law_1.01,0.23153278827667237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,2,power_law_1.01,0.11516159772872925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,2,power_law_1.01,0.12148480415344239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,16,power_law_1.01,0.31999359130859373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,2,power_law_1.01,0.12149120569229126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,8,power_law_1.2,1.0959168434143067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,2,power_law_1.01,0.12865279912948607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,16,power_law_1.01,0.43086719512939453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,2,power_law_1.01,0.1324031949043274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,16,power_law_1.01,0.5429632186889648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,2,power_law_1.01,0.14394880533218385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,8,power_law_1.2,1.2616576194763183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,2,power_law_1.01,0.1466879963874817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,2,power_law_1.01,0.14383360147476196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,16,power_law_1.01,0.8212800025939941
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,2,power_law_1.01,0.1552191972732544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,2,power_law_1.01,0.1605504035949707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,2,power_law_1.01,0.17109760046005248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,8,power_law_1.2,2.1380992889404298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,2,power_law_1.01,0.1984768033027649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,16,power_law_1.01,1.803276824951172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,2,power_law_1.01,0.22632958889007568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,2,power_law_1.01,0.2825536012649536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,32,1,balanced,0.08293866614500682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,32,1,balanced,0.08907199899355571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,32,1,balanced,0.09557333588600159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,32,1,balanced,0.120688001314799
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,32,1,balanced,0.16447466611862183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,32,1,balanced,0.247871994972229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,32,1,balanced,0.2521226604779561
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,32,1,balanced,0.2547626694043477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,32,1,balanced,0.2549546758333842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,32,1,balanced,0.25891733169555664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,2,power_law_1.01,0.3323456048965454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,32,1,balanced,0.25863999128341675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,32,1,balanced,0.2661120096842448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,32,1,balanced,0.26846933364868164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,32,1,balanced,0.2760960062344869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,32,1,balanced,0.2823093334833781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,32,1,balanced,0.29146132866541546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,32,1,balanced,0.3027199904123942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,32,1,balanced,0.3473866780598958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,2,power_law_1.01,0.4335360050201416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,32,1,balanced,0.3763466676076253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,32,1,balanced,0.45157865683237713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,32,1,balanced,0.5182613531748453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,32,1,balanced,0.6714239915211996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,32,1,balanced,0.8277119795481364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,2,power_law_1.01,0.5508992195129394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,32,1,balanced,1.2330880165100098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,8,power_law_1.2,4.288556671142578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,32,1,balanced,1.5593652725219727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,32,1,balanced,2.270143985748291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,2,power_law_1.01,0.7614336013793945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,32,1,balanced,2.9813226064046225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,32,1,balanced,3.752565383911133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,2,power_law_1.01,0.9675840377807617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,32,1,balanced,5.752058664957683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,2,power_law_1.01,1.320736026763916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,2,power_law_1.01,1.7207616806030273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,32,1,balanced,11.158650716145834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,2,power_law_1.01,2.1740671157836915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,2,power_law_1.01,3.2286529541015625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,32,power_law_1.01,0.056908798217773435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,32,power_law_1.01,0.058233600854873654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,2,balanced,0.09781333804130554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,2,balanced,0.14442666371663412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,2,balanced,0.2432159980138143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,2,balanced,0.43518932660420734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,2,balanced,0.8225599924723307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,2,power_law_1.01,6.568275451660156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,2,balanced,1.596362590789795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,2,balanced,1.6011892954508464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,32,power_law_1.01,0.049081599712371825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,2,balanced,1.5989386240641277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,2,balanced,1.6037279764811199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,2,balanced,1.607375939687093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,2,balanced,1.6132532755533855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,2,balanced,1.617029349009196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,32,power_law_1.01,0.05137280225753784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,2,balanced,1.6232266426086426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,2,balanced,1.6357812881469727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,2,balanced,1.6443999608357747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,2,balanced,1.6523200670878093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,2,balanced,1.6718932787577312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,32,power_law_1.01,0.05440000295639038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,2,balanced,1.6930826505025227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,2,balanced,1.7272426287333171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,8,balanced,0.0554720014333725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,8,balanced,0.09030399719874065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,8,balanced,0.09986133376757304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,8,balanced,0.1548799971739451
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,8,balanced,0.2566080093383789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,2,balanced,1.7830400466918945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,8,balanced,0.3139786720275879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,8,balanced,0.31438400348027545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,8,balanced,0.3160159985224406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,8,balanced,0.3188213308652242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,2,balanced,1.8475467363993328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,8,balanced,0.3174293239911397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,8,balanced,0.31717334191004437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,8,balanced,0.3187359968821208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,32,power_law_1.01,0.057081598043441775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,8,balanced,0.3203360040982564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,2,balanced,1.9321386019388835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,8,balanced,0.3198186755180359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,8,balanced,0.32656532526016235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,8,balanced,0.32789333661397296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,8,balanced,0.33100799719492596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,2,balanced,2.001306692759196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,8,balanced,0.3404906590779622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,8,balanced,0.3454506794611613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,8,balanced,0.4196213483810425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,8,balanced,0.3871306578318278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,2,balanced,2.466853300730387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,8,balanced,0.5950986544291178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,8,balanced,0.48257601261138916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,8,balanced,0.7096479733784994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,2,balanced,2.4984374046325684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,8,balanced,0.6922240257263184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,32,power_law_1.01,0.05787519812583923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,8,balanced,1.013866662979126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,8,balanced,1.18394136428833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,2,balanced,3.4049599965413413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,8,balanced,1.3314826488494873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,8,balanced,2.0781866709391275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,2,balanced,3.7478132247924805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,32,power_law_1.01,0.05678079724311828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,8,balanced,3.936335881551107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,2,balanced,4.921589215596517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,32,power_law_1.01,0.057760000228881836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,2,balanced,6.645984013875325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,32,power_law_1.01,0.058220797777175905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,128,power_law_1.2,0.03871360123157501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,128,power_law_1.2,0.03891200125217438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,2,balanced,13.119237263997396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,128,power_law_1.2,0.03797119855880737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,32,power_law_1.01,0.060224002599716185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,128,power_law_1.2,0.03811199963092804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,128,power_law_1.2,0.03878400027751923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,128,power_law_1.2,0.04030719995498657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,32,power_law_1.01,0.060147202014923094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,128,power_law_1.2,0.04023680090904236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,128,power_law_1.2,0.0398719996213913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,32,power_law_1.01,0.06756479740142822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,128,power_law_1.2,0.04059520065784454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,128,power_law_1.2,0.04048640131950378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,128,power_law_1.2,0.04094719886779785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,32,power_law_1.01,0.06775040030479432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,128,power_law_1.2,0.04149760007858276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,128,power_law_1.2,0.041555199027061465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,128,power_law_1.2,0.04221439957618713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,32,power_law_1.01,0.07745919823646545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,128,power_law_1.2,0.04588159918785095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,128,power_law_1.2,0.04669440090656281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,128,power_law_1.2,0.047603198885917665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,32,power_law_1.01,0.08060160279273987
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,128,power_law_1.2,0.0506816029548645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,2,power_law_1.2,0.09634559750556945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,128,power_law_1.2,0.05813120007514953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,128,power_law_1.2,0.06380159854888916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,32,power_law_1.01,0.08145920038223267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,2,power_law_1.2,0.14538240432739258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,128,power_law_1.2,0.07543039917945862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,128,power_law_1.2,0.09370880126953125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,2,power_law_1.2,0.10192639827728271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,32,power_law_1.01,0.09774720072746276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,128,power_law_1.2,0.11706240177154541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,2,power_law_1.2,0.176256000995636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,32,1,power_law_1.2,0.03223679959774017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,128,power_law_1.2,0.15557760000228882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,32,1,power_law_1.2,0.035366401076316833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,32,1,power_law_1.2,0.03452799916267395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,2,power_law_1.2,0.2598400115966797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,128,power_law_1.2,0.20054399967193604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,32,power_law_1.01,0.10376319885253907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,32,1,power_law_1.2,0.03818880021572113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,32,1,power_law_1.2,0.03544319868087768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,32,1,power_law_1.2,0.04725759923458099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,2,power_law_1.2,0.37020800113677976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,128,power_law_1.2,0.3016319990158081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,32,1,power_law_1.2,0.05244160294532776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,32,1,power_law_1.2,0.04039680063724518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,32,power_law_1.01,0.12676479816436767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,32,1,power_law_1.2,0.04387199878692627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,2,power_law_1.2,0.5699007987976075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,128,power_law_1.2,0.3686336040496826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,32,1,power_law_1.2,0.04759680032730103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,32,1,power_law_1.2,0.0435263991355896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,32,1,power_law_1.2,0.04848000109195709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,2,power_law_1.2,0.5942848205566407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,32,1,power_law_1.2,0.049753600358963014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,128,power_law_1.2,0.4414015769958496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,32,1,power_law_1.2,0.051052802801132204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,32,power_law_1.01,0.1557952046394348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,32,1,power_law_1.2,0.057529598474502563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,32,1,power_law_1.2,0.06119679808616638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,2,power_law_1.2,0.6367743968963623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,32,1,power_law_1.2,0.06380800008773804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,32,1,power_law_1.2,0.07649919986724854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,128,power_law_1.2,0.7621376037597656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,32,1,power_law_1.2,0.08181759715080261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,2,power_law_1.2,0.6701119899749756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,32,1,power_law_1.2,0.0979968011379242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,32,power_law_1.01,0.1993664026260376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,32,1,power_law_1.2,0.10584959983825684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,32,1,power_law_1.2,0.14744319915771484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,2,power_law_1.2,0.6750463962554931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,32,1,power_law_1.2,0.15984640121459961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,32,1,power_law_1.2,0.22275838851928711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,32,1,power_law_1.2,0.2833983898162842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,2,power_law_1.2,0.7206719875335693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,128,power_law_1.2,1.3835391998291016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,32,power_law_1.01,0.20467839241027833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,32,1,power_law_1.2,0.4017471790313721
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,32,1,power_law_1.2,0.5155839920043945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,2,power_law_1.2,0.7409599781036377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,32,1,power_law_1.2,0.6369088172912598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,2,power_law_1.2,0.7820735931396484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,32,1,power_law_1.2,0.9550399780273438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,32,power_law_1.01,0.26859519481658933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,2,power_law_1.2,0.7800576210021972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,32,1,power_law_1.2,1.8609407424926758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,2,power_law_1.2,0.805504035949707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,32,power_law_1.01,0.3055232048034668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,2,power_law_1.2,0.8592703819274903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,2,power_law_1.2,0.94717435836792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,32,power_law_1.01,0.42761597633361814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,2,power_law_1.2,0.9590720176696778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,2,power_law_1.2,1.1004480361938476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,2,power_law_1.2,1.1200832366943358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,32,power_law_1.01,0.547212791442871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,2,power_law_1.2,1.3284159660339356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,2,power_law_1.2,1.5003456115722655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,8,power_law_1.2,0.05581439733505249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,8,32,power_law_1.01,0.63854079246521
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,8,power_law_1.2,0.06899840235710145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,2,power_law_1.2,1.8710847854614259
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,8,power_law_1.2,0.04969600141048432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,2,balanced,0.04831466575463613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,2,balanced,0.05040533343950907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,2,balanced,0.05230933427810669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,2,balanced,0.06354133288065593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,2,balanced,0.09285333752632141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,2,balanced,0.13219199577967325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,2,balanced,0.13737066586812338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,2,balanced,0.1329759955406189
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,2,balanced,0.13594133655230203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,2,power_law_1.2,2.157209587097168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,2,balanced,0.137472003698349
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,8,power_law_1.2,0.05374079942703247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,2,balanced,0.1395626664161682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,2,balanced,0.1418880025545756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,2,balanced,0.14631999532381693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,2,balanced,0.14944000045458475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,2,balanced,0.15795733531316122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,2,balanced,0.16309332847595215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,2,balanced,0.1742666761080424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,2,balanced,0.19683200120925903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,2,balanced,0.2183306614557902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,8,power_law_1.2,0.06344320178031922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,2,balanced,0.2629706660906474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,2,balanced,0.3007413347562154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,2,balanced,0.418938676516215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,2,balanced,0.4883999824523926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,2,balanced,0.7143680254618326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,2,power_law_1.2,2.780607986450195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,8,power_law_1.2,0.07398399710655212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,2,balanced,0.8786239624023438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,2,balanced,1.280074675877889
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,8,32,power_law_1.01,1.0203136444091796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,2,balanced,1.6691733996073406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,8,power_law_1.2,0.08724480271339416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,2,balanced,2.0432799657185874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,8,power_law_1.2,0.08912000060081482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,2,balanced,3.1750666300455728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,2,power_law_1.2,3.656032180786133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,8,power_law_1.2,0.08917760252952575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,2,balanced,6.238933563232422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,8,power_law_1.2,0.08922240138053894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,2,power_law_1.2,4.2737281799316404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,8,power_law_1.2,0.09796479940414429
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,8,power_law_1.2,0.09811199903488159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,8,power_law_1.2,0.09848319888114929
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,8,power_law_1.2,0.10126719474792481
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,2,power_law_1.2,6.623033905029297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,8,32,power_law_1.01,2.21777286529541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,8,power_law_1.2,0.11116160154342651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,8,power_law_1.2,0.11670399904251098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,8,power_law_1.2,0.11893759965896607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,8,power_law_1.2,0.1377344012260437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,8,power_law_1.2,0.14216320514678954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,2,power_law_1.2,11.97225570678711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,8,power_law_1.2,0.1754047989845276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,8,power_law_1.2,0.20176000595092775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,8,power_law_1.2,0.23812479972839357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,2,balanced,0.07077333331108093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,2,balanced,0.10749866565068562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,2,balanced,0.12173866232236226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,2,balanced,0.18264534076054892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,2,balanced,0.3022986650466919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,2,balanced,0.5406239827473959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,2,balanced,0.5436160167058309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,8,power_law_1.2,0.31687679290771487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,2,balanced,0.543498675028483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,2,balanced,0.5469973484675089
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,2,balanced,0.5480373303095499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,2,balanced,0.5530080000559489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,2,balanced,0.5563946564992269
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,2,balanced,0.5594506661097208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,2,balanced,0.5641706784566244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,8,power_law_1.2,0.39394559860229494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,2,balanced,0.5725866556167603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,2,balanced,0.5806453227996826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,2,balanced,0.5910986661911011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,2,balanced,0.6131893396377563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,2,balanced,0.6326186656951904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,2,balanced,0.6803893248240153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,2,balanced,0.7403039932250977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,8,power_law_1.2,0.5073472023010254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,2,balanced,0.8326773643493652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,2,balanced,0.91811736424764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,2,balanced,1.2275946935017903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,2,balanced,1.3838826815287273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,2,balanced,1.9841119448343914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,8,power_law_1.2,0.6907264232635498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,2,balanced,2.4535093307495117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,128,power_law_1.2,0.03765760064125061
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,2,balanced,3.0654560724894204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,128,power_law_1.2,0.03758719861507416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,8,power_law_1.2,0.9473919868469238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,128,power_law_1.2,0.03711360096931458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,2,balanced,4.586826642354329
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,128,power_law_1.2,0.03741439878940582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,128,power_law_1.2,0.037567999958992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,8,power_law_1.2,1.014668846130371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,128,power_law_1.2,0.03933440148830414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,2,balanced,8.587146759033203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,128,power_law_1.2,0.038822400569915774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,128,power_law_1.2,0.03878400027751923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,128,power_law_1.2,0.038022398948669434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,8,power_law_1.2,1.8230144500732421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,128,power_law_1.2,0.0382207989692688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,1,power_law_1.01,0.031513598561286923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,128,power_law_1.2,0.038361600041389464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,1,power_law_1.01,0.03522560000419617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,128,power_law_1.2,0.03871360123157501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,1,power_law_1.01,0.04135040044784546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,128,power_law_1.2,0.038873600959777835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,1,power_law_1.01,0.05685120224952698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,128,power_law_1.2,0.039801600575447085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,1,power_law_1.01,0.06620799899101257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,128,power_law_1.2,0.0421887993812561
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,1,power_law_1.01,0.10095360279083251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,128,power_law_1.2,0.04309119880199432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,8,power_law_1.2,3.605817413330078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,1,power_law_1.01,0.10483839511871337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,128,power_law_1.2,0.044095999002456664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,1,power_law_1.01,0.11274880170822144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,128,power_law_1.2,0.04415999948978424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,1,power_law_1.01,0.1121664047241211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,128,power_law_1.2,0.04903680086135864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,1,power_law_1.01,0.1177024006843567
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,128,power_law_1.2,0.05596799850463867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,1,power_law_1.01,0.12070399522781372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,128,power_law_1.2,0.0651199996471405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,1,power_law_1.01,0.12447359561920165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,128,power_law_1.2,0.08687360286712646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,1,power_law_1.01,0.1268928050994873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,128,power_law_1.2,0.09926400184631348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,1,power_law_1.01,0.12725759744644166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,128,power_law_1.2,0.143449604511261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,1,power_law_1.01,0.13771519660949708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,128,power_law_1.2,0.18030719757080077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,1,power_law_1.01,0.14472320079803466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,1,power_law_1.01,0.14988160133361816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,128,power_law_1.2,0.2356544017791748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,8,balanced,0.03521066655715307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,8,balanced,0.034634667138258614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,1,power_law_1.01,0.1657920002937317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,8,balanced,0.038533332447210945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,8,balanced,0.05680533250172933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,8,balanced,0.07964266836643219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,8,balanced,0.08113066852092743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,8,balanced,0.08118933439254761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,128,power_law_1.2,0.3371328115463257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,8,balanced,0.08329600095748901
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,1,power_law_1.01,0.17040640115737915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,8,balanced,0.08185066779454549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,8,balanced,0.08253333469231923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,8,balanced,0.083514670530955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,8,balanced,0.08352532982826233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,8,balanced,0.08541867136955261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,8,balanced,0.0849226713180542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,8,balanced,0.09224533041318257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,1,power_law_1.01,0.20352640151977539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,8,balanced,0.09430932998657227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,8,balanced,0.09516266981760661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,8,balanced,0.10086933771769206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,1,128,power_law_1.2,0.42495999336242674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,8,balanced,0.1163093348344167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,8,balanced,0.13239999612172446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,8,balanced,0.13451199730237326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,1,power_law_1.01,0.23583359718322755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,8,balanced,0.1692053278287252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,8,balanced,0.18953599532445273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,8,balanced,0.25692800680796307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,8,balanced,0.30588799715042114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,8,balanced,0.42031999429066974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,1,power_law_1.01,0.28965120315551757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,8,balanced,0.530410647392273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,8,balanced,0.6447466611862183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,1,128,power_law_1.2,0.7413504123687744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,8,balanced,0.9857813517252604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,1,power_law_1.01,0.35587201118469236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,8,balanced,1.8810666402180989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,1,power_law_1.01,0.4595071792602539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,1,power_law_1.01,0.5253056049346924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,1,128,power_law_1.2,1.4213248252868653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,1,power_law_1.01,0.7237823963165283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,2,power_law_1.01,0.07253760099411011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,1,power_law_1.01,0.938815975189209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,2,power_law_1.01,0.10071040391921997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,2,power_law_1.01,0.08922880291938781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,4,1,power_law_1.01,1.1349120140075684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,2,power_law_1.01,0.13203200101852416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,2,power_law_1.01,0.15667200088500977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,4,1,power_law_1.01,1.7311872482299804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,2,power_law_1.01,0.24789121150970458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,2,power_law_1.01,0.2602240085601807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,2,power_law_1.01,0.27473280429840086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,2,power_law_1.01,0.2752255916595459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,4,1,power_law_1.01,3.3350399017333983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,2,power_law_1.01,0.2900799989700317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,2,power_law_1.01,0.29681921005249023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,2,power_law_1.01,0.29745280742645264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,2,power_law_1.01,0.3116224050521851
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,2,power_law_1.01,0.3192960023880005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,2,power_law_1.01,0.331987190246582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,2,power_law_1.01,0.33242239952087405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,8,balanced,0.04692266881465912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,8,balanced,0.05218133330345154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,8,balanced,0.07089066505432129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,2,power_law_1.01,0.35386879444122316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,8,balanced,0.0962559978167216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,8,balanced,0.14864533146222433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,8,balanced,0.2451039950052897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,1,power_law_1.01,0.07517439723014832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,8,balanced,0.3431466817855835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,8,balanced,0.34168533484141034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,8,balanced,0.3416159947713216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,8,balanced,0.3426506519317627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,2,power_law_1.01,0.39559040069580076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,8,balanced,0.34883733590443927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,8,balanced,0.34882132212320965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,8,balanced,0.35117868582407635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,8,balanced,0.3492266734441121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,8,balanced,0.35634132226308185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,8,balanced,0.35791468620300293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,8,balanced,0.3649066686630249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,2,power_law_1.01,0.4239935874938965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,8,balanced,0.37457601229349774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,8,balanced,0.38435200850168866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,8,balanced,0.40722131729125977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,8,balanced,0.4248640139897664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,8,balanced,0.4654080073038737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,8,balanced,0.5069760084152222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,2,power_law_1.01,0.5192831993103028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,1,power_law_1.01,0.08369280099868774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,8,balanced,0.5757866700490316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,8,balanced,0.6850879987080892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,8,balanced,0.82424529393514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,8,balanced,1.0419200261433919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,2,power_law_1.01,0.5362880229949951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,8,balanced,1.3204053243001301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,8,balanced,1.7988425890604656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,2,power_law_1.01,0.6707968235015869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,1,power_law_1.01,0.09644160270690919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,8,balanced,3.392666816711426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,2,power_law_1.01,0.8320511817932129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,1,power_law_1.01,0.13527679443359375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,2,power_law_1.01,1.0899264335632324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,2,power_law_1.01,1.4127936363220215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,32,2,power_law_1.2,0.03146879971027374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,1,power_law_1.01,0.18184959888458252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,32,2,power_law_1.2,0.03132160007953644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,32,2,power_law_1.2,0.031327998638153075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,32,2,power_law_1.2,0.03278720080852508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,32,2,power_law_1.2,0.03224959969520569
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,32,2,power_law_1.2,0.03412480056285858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,32,2,power_law_1.2,0.03460479974746704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,2,power_law_1.01,1.9822975158691407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,32,2,power_law_1.2,0.03529599905014038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,1,power_law_1.01,0.22187519073486328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,32,2,power_law_1.2,0.03632639944553375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,32,2,power_law_1.2,0.037036800384521486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,32,2,power_law_1.2,0.03734399974346161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,32,2,power_law_1.2,0.038438400626182555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,32,2,power_law_1.2,0.040940800309181215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,32,2,power_law_1.2,0.044121599197387694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,32,2,power_law_1.2,0.05003520250320435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,2,power_law_1.01,2.429203224182129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,32,2,power_law_1.2,0.05348479747772217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,1,power_law_1.01,0.29484798908233645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,32,2,power_law_1.2,0.05656319856643677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,32,2,power_law_1.2,0.07566720247268677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,32,2,power_law_1.2,0.08803840279579163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,32,2,power_law_1.2,0.11233919858932495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,32,2,power_law_1.2,0.14462080001831054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,32,2,power_law_1.2,0.18211840391159057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,32,2,power_law_1.2,0.21676158905029297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,1,power_law_1.01,0.3152704000473022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,2,power_law_1.01,3.044428825378418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,32,2,power_law_1.2,0.2943552017211914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,32,2,power_law_1.2,0.3958336114883423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,32,2,power_law_1.2,0.5854015827178956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,32,2,power_law_1.2,0.6955455780029297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,1,power_law_1.01,0.3267519950866699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,32,2,power_law_1.2,0.980191993713379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,32,2,power_law_1.2,1.4186559677124024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,2,power_law_1.01,4.327577590942383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,1,power_law_1.01,0.33701119422912595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,32,2,power_law_1.2,2.7614784240722656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,1,power_law_1.01,0.3366336107254028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,4,power_law_1.2,0.05648639798164368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,1,power_law_1.01,0.35038080215454104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,2,power_law_1.01,8.123270416259766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,4,power_law_1.2,0.062483197450637816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,1,power_law_1.01,0.3608448028564453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,4,power_law_1.2,0.061939197778701785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,1,power_law_1.01,0.3732736110687256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,4,power_law_1.2,0.06424319744110107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,1,power_law_1.01,0.3838912010192871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,4,power_law_1.2,0.07633919715881347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,1,power_law_1.01,0.38826239109039307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,4,power_law_1.2,0.08911359906196595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,1,power_law_1.01,0.413267183303833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,4,power_law_1.2,0.09783040285110474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,1,power_law_1.01,0.46046719551086424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,4,power_law_1.2,0.11534719467163086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,1,power_law_1.01,0.5078911781311035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,4,power_law_1.2,0.11711360216140747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,1,power_law_1.01,0.6025856018066407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,4,power_law_1.2,0.122489595413208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,32,power_law_1.2,0.059673601388931276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,1,power_law_1.01,0.6723840236663818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,4,power_law_1.2,0.1235967993736267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,32,power_law_1.2,0.06710399985313416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,1,power_law_1.01,0.8546431541442872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,32,power_law_1.01,0.0627776026725769
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,4,power_law_1.2,0.13237119913101197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,32,power_law_1.2,0.05483520030975342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,32,power_law_1.01,0.05394560098648071
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,1,power_law_1.01,1.0391167640686034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,32,power_law_1.01,0.04831359982490539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,4,power_law_1.2,0.13564800024032592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,32,power_law_1.2,0.05148800015449524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,32,power_law_1.01,0.04907520115375519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,1,power_law_1.01,1.4046336174011231
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,32,power_law_1.01,0.05082880258560181
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,4,power_law_1.2,0.13768320083618163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,32,power_law_1.2,0.052211201190948485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,32,power_law_1.01,0.051667201519012454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,32,power_law_1.01,0.052960002422332765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,4,power_law_1.2,0.15029759407043458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,1,power_law_1.01,1.7511743545532226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,32,power_law_1.2,0.05382400155067444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,32,power_law_1.01,0.054041600227355956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,32,power_law_1.01,0.05367040038108826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,4,power_law_1.2,0.15345280170440673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,32,power_law_1.2,0.05493119955062866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,32,power_law_1.01,0.05398399829864502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,1,power_law_1.01,2.466988754272461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,32,power_law_1.01,0.054553598165512085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,4,power_law_1.2,0.16254080533981324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,32,power_law_1.01,0.05633280277252197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,32,power_law_1.2,0.0565887987613678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,32,power_law_1.01,0.05683839917182922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,4,power_law_1.2,0.17832319736480712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,32,power_law_1.01,0.060313600301742556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,32,power_law_1.2,0.05759360194206238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,1,power_law_1.01,3.190559959411621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,32,power_law_1.01,0.06664320230484008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,4,power_law_1.2,0.20178558826446533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,32,power_law_1.01,0.07026559710502625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,32,power_law_1.2,0.057120001316070555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,32,power_law_1.01,0.07942399978637696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,4,power_law_1.2,0.2413952112197876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,32,power_law_1.01,0.08466560244560242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,32,power_law_1.2,0.058745598793029784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,1,power_law_1.01,3.912160110473633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,32,power_law_1.01,0.09353600144386291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,4,power_law_1.2,0.2758656024932861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,32,power_law_1.01,0.11458560228347778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,32,power_law_1.2,0.0609279990196228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,32,power_law_1.01,0.13068799972534179
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,4,power_law_1.2,0.3365504026412964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,32,power_law_1.2,0.06207360029220581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,32,power_law_1.01,0.15478399991989136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,4,power_law_1.2,0.42009601593017576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,32,power_law_1.01,0.1900480031967163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,1,power_law_1.01,6.015065765380859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,32,power_law_1.2,0.062745600938797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,32,power_law_1.01,0.25114240646362307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,4,power_law_1.2,0.6072192192077637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,32,power_law_1.2,0.06884480118751526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,32,power_law_1.01,0.2930624008178711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,32,power_law_1.01,0.4431488037109375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,32,power_law_1.2,0.06973440051078797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,4,power_law_1.2,0.7556928157806396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,32,power_law_1.01,0.5843840122222901
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,32,power_law_1.2,0.0769919991493225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,4,power_law_1.2,1.0156352043151855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,32,power_law_1.01,0.634496021270752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,32,power_law_1.2,0.08732799887657165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,1,power_law_1.01,11.701119995117187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,4,power_law_1.2,1.35731201171875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,32,power_law_1.2,0.09564800262451172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,32,power_law_1.01,1.0771007537841797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,32,power_law_1.2,0.11676160097122193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,4,power_law_1.2,1.7708288192749024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,32,power_law_1.2,0.1442944049835205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,32,power_law_1.01,1.895078468322754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,32,power_law_1.2,0.18821120262145996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,4,power_law_1.2,2.622483253479004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,32,power_law_1.2,0.22243199348449708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,32,power_law_1.2,0.30818560123443606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,32,power_law_1.2,0.36990718841552733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,4,power_law_1.2,4.855110549926758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,32,power_law_1.2,0.5553408145904541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,32,power_law_1.2,0.6777344226837159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,16,power_law_1.01,0.06094719767570496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,32,power_law_1.2,0.8333439826965332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,8,balanced,0.043951998154322304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,16,power_law_1.01,0.0762943983078003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,8,balanced,0.04571733375390371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,8,balanced,0.04574933151404063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,8,balanced,0.049584001302719116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,8,balanced,0.052058666944503784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,8,balanced,0.0641546646753947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,8,balanced,0.08193600177764893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,8,balanced,0.0831413318713506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,8,balanced,0.08435733119646709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,8,balanced,0.08521067102750142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,8,balanced,0.08708266417185466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,8,balanced,0.08877866466840108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,8,balanced,0.0888159970442454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,8,balanced,0.09153067072232564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,8,balanced,0.09923733274141948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,8,balanced,0.10153067111968994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,8,balanced,0.11135466893513997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,8,balanced,0.12261333068211873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,8,balanced,0.13487466176350912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,8,balanced,0.158160001039505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,8,balanced,0.17884800831476846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,8,balanced,0.22613867123921713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,8,balanced,0.28803733984629315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,8,balanced,0.385696013768514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,16,power_law_1.01,0.052832001447677614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,8,balanced,0.49769067764282227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,8,balanced,0.711845318476359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,8,balanced,0.903119961420695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,32,power_law_1.2,1.365171241760254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,8,balanced,1.124949296315511
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,8,balanced,1.7338666915893555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,16,power_law_1.01,0.05312640070915222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,8,balanced,3.386656125386556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,16,power_law_1.01,0.05667200088500977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,8,8,power_law_1.2,0.029407998919486998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,32,power_law_1.2,2.631391906738281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,16,power_law_1.01,0.06110720038414001
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,8,8,power_law_1.2,0.029971200227737426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,8,8,power_law_1.2,0.029151999950408937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,8,8,power_law_1.2,0.029344001412391664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,8,8,power_law_1.2,0.03173120021820068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,8,8,power_law_1.2,0.030585598945617676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,8,8,power_law_1.2,0.034431999921798705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,8,8,power_law_1.2,0.03255040049552917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,8,8,power_law_1.2,0.034585601091384886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,16,power_law_1.01,0.06348159909248352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,8,8,power_law_1.2,0.03619199991226196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,8,8,power_law_1.2,0.034118399024009705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,8,8,power_law_1.2,0.03857919871807099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,8,8,power_law_1.2,0.03863680064678192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,8,8,power_law_1.2,0.03712640106678009
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,8,8,power_law_1.2,0.04631040096282959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,8,8,power_law_1.2,0.048537600040435794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,16,power_law_1.01,0.06606079936027527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,8,8,power_law_1.2,0.04928640127182007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,8,8,power_law_1.2,0.05809919834136963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,8,8,power_law_1.2,0.06615039706230164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,8,8,power_law_1.2,0.08787840008735656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,8,8,power_law_1.2,0.09491199851036072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,8,8,power_law_1.2,0.13747199773788452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,16,power_law_1.01,0.06744319796562195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,8,8,power_law_1.2,0.1597440004348755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,8,8,power_law_1.2,0.22245121002197266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,8,8,power_law_1.2,0.2819519996643066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,8,8,power_law_1.2,0.4111743927001953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,16,power_law_1.01,0.06640639901161194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,8,8,power_law_1.2,0.5227519989013671
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,8,8,power_law_1.2,0.6189504146575928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,8,8,power_law_1.2,0.9809344291687012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,16,power_law_1.01,0.06963840126991272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,8,8,power_law_1.2,1.891744041442871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,16,power_law_1.01,0.07086719870567322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,16,power_law_1.01,0.07051519751548767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,16,power_law_1.01,0.07265920042991639
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,16,power_law_1.01,0.07952640056610108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,16,power_law_1.01,0.08149120211601257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,2,power_law_1.2,0.05931519865989685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,8,power_law_1.01,0.04195840060710907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,16,power_law_1.01,0.09123839735984803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,8,power_law_1.01,0.03521920144557953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,2,power_law_1.2,0.06445440053939819
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,8,power_law_1.01,0.03415679931640625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,16,power_law_1.01,0.10560640096664428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,8,power_law_1.01,0.034534400701522826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,2,power_law_1.2,0.06845440268516541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,8,power_law_1.01,0.036032000184059144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,16,power_law_1.01,0.11605759859085082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,2,power_law_1.2,0.07611520290374756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,8,power_law_1.01,0.041791999340057374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,8,power_law_1.01,0.041580799221992495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,2,power_law_1.2,0.09520000219345093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,16,power_law_1.01,0.139302396774292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,8,power_law_1.01,0.041196799278259276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,8,power_law_1.01,0.041407999396324155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,2,power_law_1.2,0.11370240449905396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,8,power_law_1.01,0.04227199852466583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,16,power_law_1.01,0.154585599899292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,8,power_law_1.01,0.04289279878139496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,2,power_law_1.2,0.14650880098342894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,2,balanced,0.03383466601371765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,2,balanced,0.03389333436886469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,2,balanced,0.037818667789300285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,2,balanced,0.042026668787002563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,8,power_law_1.01,0.04433920085430145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,2,balanced,0.06819200019041698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,2,balanced,0.07266666491826375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,2,balanced,0.07158400118350983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,2,balanced,0.07054933408896129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,2,balanced,0.07301866511503856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,2,balanced,0.0747573326031367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,16,power_law_1.01,0.1914687991142273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,2,balanced,0.07604266703128815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,2,balanced,0.07493866483370464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,2,balanced,0.07701333363850911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,8,power_law_1.01,0.04506239891052246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,2,balanced,0.08226133386294048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,2,power_law_1.2,0.15020159482955933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,2,balanced,0.09126399954160054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,2,balanced,0.092549333969752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,2,balanced,0.0965280036131541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,2,balanced,0.11030399799346924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,2,balanced,0.12703466415405273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,2,balanced,0.15229866902033487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,2,balanced,0.17172799507776895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,8,power_law_1.01,0.04684160053730011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,2,balanced,0.2451840043067932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,2,balanced,0.29362666606903076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,2,balanced,0.42876799901326496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,2,balanced,0.5491306781768799
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,2,power_law_1.2,0.15497599840164183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,8,power_law_1.01,0.05130239725112915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,2,balanced,0.8028799692789713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,16,power_law_1.01,0.22165119647979736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,2,balanced,1.0451839764912922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,2,balanced,1.2990132967631023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,8,power_law_1.01,0.052179199457168576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,2,power_law_1.2,0.16008319854736328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,2,balanced,2.011653264363607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,8,power_law_1.01,0.055180799961090085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,2,balanced,3.9076480865478516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,16,power_law_1.01,0.29012479782104494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,8,power_law_1.01,0.06181120276451111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,2,power_law_1.2,0.16239999532699584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,8,power_law_1.01,0.06836479902267456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,2,power_law_1.2,0.1718783974647522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,8,power_law_1.01,0.07910400032997131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,16,power_law_1.01,0.38143999576568605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,8,power_law_1.01,0.08748800158500672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,2,power_law_1.2,0.17296639680862427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,8,power_law_1.01,0.12128640413284301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,2,power_law_1.2,0.18122880458831786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,8,power_law_1.01,0.13608959913253785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,16,power_law_1.01,0.5256768226623535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,8,power_law_1.01,0.1872063994407654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,2,power_law_1.2,0.19216639995574952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,8,power_law_1.01,0.24833920001983642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,2,power_law_1.2,0.20129919052124023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,16,power_law_1.01,0.6728127956390381
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,8,power_law_1.01,0.35508480072021487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,2,power_law_1.2,0.2121664047241211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,8,power_law_1.01,0.4092288017272949
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,2,power_law_1.2,0.23014400005340577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,16,power_law_1.01,0.8365056037902832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,8,power_law_1.01,0.48883838653564454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,2,power_law_1.2,0.25832960605621336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,2,power_law_1.2,0.3130176067352295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,8,power_law_1.01,0.7955711841583252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,16,power_law_1.01,1.2996352195739747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,2,power_law_1.2,0.3589567899703979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,2,power_law_1.2,0.4615295886993408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,8,power_law_1.01,1.5539520263671875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,2,power_law_1.2,0.5647808074951172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,32,power_law_1.2,0.06245120167732239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,16,power_law_1.01,2.6488576889038087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,32,power_law_1.2,0.054745602607727054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,2,power_law_1.2,0.7661632061004638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,32,power_law_1.2,0.04872959852218628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,32,power_law_1.2,0.04890879988670349
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,2,power_law_1.2,0.921664047241211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,32,power_law_1.2,0.05136640071868896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,32,power_law_1.2,0.052223998308181765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,2,power_law_1.2,1.372217559814453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,32,power_law_1.2,0.05189120173454285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,32,power_law_1.2,0.05413119792938233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,32,power_law_1.2,0.05322239995002746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,2,power_law_1.2,1.7926528930664063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,32,power_law_1.2,0.054611200094223024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,32,power_law_1.2,0.054681599140167236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,32,power_law_1.2,0.05818880200386047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,2,power_law_1.2,2.2060991287231446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,32,power_law_1.2,0.05958399772644043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,32,power_law_1.2,0.062483197450637816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,32,power_law_1.2,0.06899200081825256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,32,power_law_1.2,0.07256320118904114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,2,power_law_1.2,3.6874942779541016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,32,power_law_1.2,0.07648000121116638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,32,power_law_1.2,0.08734719753265381
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,32,power_law_1.2,0.09904639720916748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,32,power_law_1.2,0.09283199906349182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,32,power_law_1.2,0.12042239904403687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,32,8,power_law_1.01,0.03230080008506775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,32,power_law_1.2,0.06209279894828797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,32,8,power_law_1.01,0.03155840039253235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,32,power_law_1.2,0.14443520307540894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,32,8,power_law_1.01,0.03142400085926056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,32,power_law_1.2,0.05575039982795715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,32,8,power_law_1.01,0.030745598673820495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,32,power_law_1.2,0.18077440261840821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,2,power_law_1.2,6.5112060546875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,32,8,power_law_1.01,0.03075839877128601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,32,power_law_1.2,0.06379520297050476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,32,8,power_law_1.01,0.03131519854068756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,32,power_law_1.2,0.21219840049743652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,32,8,power_law_1.01,0.03187839984893799
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,32,power_law_1.2,0.06401919722557067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,32,8,power_law_1.01,0.032307198643684386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,32,power_law_1.2,0.28808960914611814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,32,8,power_law_1.01,0.033036801218986514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,32,power_law_1.2,0.07847679853439331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,32,8,power_law_1.01,0.03221119940280914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,32,8,power_law_1.01,0.032793599367141726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,32,power_law_1.2,0.07921280264854431
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,32,power_law_1.2,0.3439552068710327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,32,8,power_law_1.01,0.03383040130138397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,32,8,power_law_1.01,0.034995201230049136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,32,power_law_1.2,0.08039039969444275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,32,8,power_law_1.01,0.034406399726867674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,32,8,power_law_1.01,0.04068480134010315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,32,power_law_1.2,0.5226367950439453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,32,power_law_1.2,0.08019199967384338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,32,8,power_law_1.01,0.043347200751304625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,32,8,power_law_1.01,0.04527359902858734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,32,power_law_1.2,0.08104320168495179
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,32,8,power_law_1.01,0.0531328022480011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,32,8,power_law_1.01,0.05418879985809326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,32,power_law_1.2,0.6485568046569824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,32,power_law_1.2,0.07904639840126038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,32,8,power_law_1.01,0.06751999855041504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,32,8,power_law_1.01,0.07272959947586059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,32,power_law_1.2,0.08543999791145325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,32,8,power_law_1.01,0.10710400342941284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,32,8,power_law_1.01,0.1257151961326599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,32,power_law_1.2,0.9065024375915527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,32,power_law_1.2,0.08428159952163697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,32,8,power_law_1.01,0.17516160011291504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,32,8,power_law_1.01,0.22930560111999512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,32,power_law_1.2,0.08599039912223816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,32,8,power_law_1.01,0.3278656005859375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,32,power_law_1.2,0.09245439767837524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,32,8,power_law_1.01,0.4103871822357178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,32,power_law_1.2,1.464134407043457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,32,power_law_1.2,0.09671040177345276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,32,8,power_law_1.01,0.5297664165496826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,32,power_law_1.2,0.10217599868774414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,32,8,power_law_1.01,0.7932928085327149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,32,power_law_1.2,0.12177920341491699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,32,8,power_law_1.01,1.5192895889282227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,32,power_law_1.2,0.12124160528182984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,32,power_law_1.2,0.1493183970451355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,32,power_law_1.2,3.036697578430176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,32,power_law_1.2,0.16442240476608277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,8,balanced,0.050437331199645996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,8,balanced,0.07018666466077168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,8,balanced,0.09661333759625752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,8,balanced,0.14471999804178873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,8,balanced,0.2412160038948059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,32,power_law_1.2,0.21262719631195068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,8,balanced,0.4371253252029419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,8,balanced,0.43862398465474445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,8,balanced,0.44091200828552246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,8,balanced,0.44233067830403644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,8,balanced,0.4410560131072998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,8,balanced,0.44522666931152344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,8,balanced,0.4468906720479329
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,32,power_law_1.2,0.24730238914489747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,8,balanced,0.44859198729197186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,8,balanced,0.4486986796061198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,8,balanced,0.4553920030593872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,8,balanced,0.45797332127888996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,8,balanced,0.4604693253835042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,8,balanced,0.4703253507614136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,8,balanced,0.48109865188598633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,32,power_law_1.2,0.3267008066177368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,8,balanced,0.49901866912841797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,8,balanced,0.5275199810663859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,8,balanced,0.5580106576283773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,8,balanced,0.6136159896850586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,8,balanced,0.6929759979248047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,8,balanced,0.7452212969462076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,32,power_law_1.2,0.4256127834320068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,8,balanced,1.0213546752929688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,8,balanced,1.1551360289255779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,8,balanced,1.498693307240804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,32,power_law_1.2,0.7151872158050537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,8,balanced,2.003333409627279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,8,balanced,3.916725476582845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,32,power_law_1.2,0.7711167812347413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,32,power_law_1.2,1.2965184211730958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,32,power_law_1.2,1.93175048828125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,2,8,power_law_1.01,0.05794559717178345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,2,8,power_law_1.01,0.05796480178833008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,2,8,power_law_1.01,0.057785600423812866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,2,8,power_law_1.01,0.05727999806404114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,1,1,power_law_1.01,0.08759040236473084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,1,1,power_law_1.01,0.1091007947921753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,2,8,power_law_1.01,0.058380800485610965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,32,power_law_1.2,3.619142532348633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,1,1,power_law_1.01,0.13496960401535035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,2,8,power_law_1.01,0.060115200281143186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,1,1,power_law_1.01,0.2318592071533203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,1,1,power_law_1.01,0.23627519607543945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,2,8,power_law_1.01,0.06267520189285278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,1,1,power_law_1.01,0.24129281044006348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,2,8,power_law_1.01,0.0648959994316101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,1,1,power_law_1.01,0.24714879989624022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,1,1,power_law_1.01,0.24600958824157715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,2,8,power_law_1.01,0.06356480121612548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,1,1,power_law_1.01,0.2494271993637085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,2,8,power_law_1.01,0.06755840182304382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,1,1,power_law_1.01,0.2533760070800781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,2,8,power_law_1.01,0.07048959732055664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,1,1,power_law_1.01,0.25908479690551756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,2,8,power_law_1.01,0.07191039919853211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,1,1,power_law_1.01,0.2734208106994629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,1,1,power_law_1.01,0.2815104007720947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,2,8,power_law_1.01,0.07449600100517273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,1,1,power_law_1.01,0.2857599973678589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,2,8,power_law_1.01,0.07848320007324219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,1,1,power_law_1.01,0.3067712068557739
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,2,8,power_law_1.01,0.08561919927597046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,1,1,power_law_1.01,0.3190975904464722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,1,1,power_law_1.01,0.33299200534820556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,2,8,power_law_1.01,0.09155200123786926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,1,1,power_law_1.01,0.3864959955215454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,2,8,power_law_1.01,0.11255040168762206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,1,1,power_law_1.01,0.43702402114868166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,2,8,power_law_1.01,0.1363584041595459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,1,1,power_law_1.01,0.5541823863983154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,2,8,power_law_1.01,0.15440640449523926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,1,1,power_law_1.01,0.6465600013732911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,2,8,power_law_1.01,0.20543360710144043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,1,1,power_law_1.01,0.8689087867736817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,4,4,power_law_1.2,0.03891839981079102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,2,8,power_law_1.01,0.22766079902648925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,4,4,power_law_1.2,0.039929598569869995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,1,1,power_law_1.01,1.0776191711425782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,4,4,power_law_1.2,0.03680639863014221
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,2,8,power_law_1.01,0.3489984035491943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,4,4,power_law_1.2,0.03700479865074158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,4,4,power_law_1.2,0.03683840036392212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,1,1,power_law_1.01,1.5286591529846192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,2,8,power_law_1.01,0.4485951900482178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,4,4,power_law_1.2,0.03806720077991486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,4,4,power_law_1.2,0.03854719996452331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,4,4,power_law_1.2,0.03903360068798065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,2,8,power_law_1.01,0.5983551979064942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,4,4,power_law_1.2,0.040761598944664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,1,1,power_law_1.01,1.9377471923828125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,4,4,power_law_1.2,0.04220159947872162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,2,8,power_law_1.01,0.7761919975280762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,4,4,power_law_1.2,0.04286080002784729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,4,4,power_law_1.2,0.04487679898738861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,4,4,power_law_1.2,0.04520959854125976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,1,1,power_law_1.01,2.8246719360351564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,2,8,power_law_1.01,1.1594367980957032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,4,4,power_law_1.2,0.05073919892311096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,4,4,power_law_1.2,0.05681920051574707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,4,4,power_law_1.2,0.060249602794647215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,4,4,power_law_1.2,0.06500480175018311
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,2,8,power_law_1.01,1.7570367813110352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,4,4,power_law_1.2,0.08538240194320679
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,1,1,power_law_1.01,3.694041442871094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,4,4,power_law_1.2,0.08895999789237977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,4,4,power_law_1.2,0.11090559959411621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,4,4,power_law_1.2,0.11386879682540893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,2,8,power_law_1.01,2.3418048858642577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,4,4,power_law_1.2,0.17136640548706056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,4,4,power_law_1.2,0.22680320739746093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,4,4,power_law_1.2,0.32390398979187013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,1,1,power_law_1.01,4.640972900390625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,4,4,power_law_1.2,0.3662911891937256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,2,8,power_law_1.01,3.2609214782714844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,4,4,power_law_1.2,0.5977727890014648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,4,4,power_law_1.2,0.6875967979431152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,4,4,power_law_1.2,0.9451135635375977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,1,1,power_law_1.01,7.445216369628906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,4,4,power_law_1.2,1.320678424835205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,2,power_law_1.2,0.03325439989566803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,2,8,power_law_1.01,5.868441772460938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,2,power_law_1.2,0.031641599535942075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,4,4,power_law_1.2,2.478163146972656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,2,power_law_1.2,0.03521279990673065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,2,power_law_1.2,0.03852800130844116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,2,power_law_1.2,0.04002560079097748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,2,power_law_1.2,0.05435519814491272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,1,1,power_law_1.01,14.26208038330078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,2,power_law_1.2,0.05674239993095398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,2,power_law_1.2,0.05802879929542541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,2,power_law_1.2,0.06060799956321716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,2,power_law_1.2,0.0621504008769989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,2,power_law_1.2,0.06309120059013366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,2,power_law_1.2,0.06788480281829834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,2,power_law_1.2,0.07052159905433655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,2,power_law_1.2,0.07259520292282104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,2,power_law_1.2,0.0791487991809845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,2,power_law_1.2,0.08275840282440186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,2,power_law_1.2,0.0896511971950531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,2,power_law_1.2,0.10186879634857178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,2,power_law_1.2,0.11726720333099365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,2,power_law_1.2,0.14421759843826293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,2,power_law_1.2,0.16094720363616943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,2,power_law_1.2,0.20208640098571778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,2,power_law_1.2,0.25948801040649416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,2,power_law_1.2,0.38339838981628416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,4,32,power_law_1.2,0.024031999707221984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,2,power_law_1.01,0.14803839921951295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,2,power_law_1.2,0.4874112129211426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,4,32,power_law_1.2,0.022937600314617158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,4,32,power_law_1.2,0.0234047994017601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,2,power_law_1.2,0.6758528232574463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,8,2,power_law_1.2,0.037561601400375365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,4,32,power_law_1.2,0.021875199675559998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,8,2,power_law_1.2,0.03751679956912994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,2,power_law_1.01,0.23956480026245117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,8,2,power_law_1.2,0.03729279935359955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,4,32,power_law_1.2,0.023756800591945647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,8,2,power_law_1.2,0.041388800740242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,2,power_law_1.2,0.9316543579101563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,8,2,power_law_1.2,0.04232319891452789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,4,32,power_law_1.2,0.025331199169158936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,8,2,power_law_1.2,0.04168959856033325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,8,2,power_law_1.2,0.04258559942245484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,8,2,power_law_1.2,0.04373759925365448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,4,32,power_law_1.2,0.02561280131340027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,8,2,power_law_1.2,0.04752640128135681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,8,2,power_law_1.2,0.047628799080848695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,4,32,power_law_1.2,0.02773120105266571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,8,2,power_law_1.2,0.04878720045089722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,2,power_law_1.2,1.202444839477539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,2,power_law_1.01,0.21670401096343994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,8,2,power_law_1.2,0.0530239999294281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,8,2,power_law_1.2,0.05414400100708008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,4,32,power_law_1.2,0.02868480086326599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,8,2,power_law_1.2,0.05699840188026428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,8,2,power_law_1.2,0.06659200191497802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,8,2,power_law_1.2,0.06886399984359741
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,4,32,power_law_1.2,0.029017600417137145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,8,2,power_law_1.2,0.08487679958343505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,8,2,power_law_1.2,0.09443839788436889
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,8,2,power_law_1.2,0.10078719854354859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,4,32,power_law_1.2,0.02995840013027191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,2,power_law_1.01,0.32565760612487793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,2,power_law_1.2,1.7632831573486327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,8,2,power_law_1.2,0.11639679670333862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,8,2,power_law_1.2,0.12789119482040406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,4,32,power_law_1.2,0.030399999022483824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,8,2,power_law_1.2,0.19826560020446776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,8,2,power_law_1.2,0.23802878856658935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,4,32,power_law_1.2,0.030995199084281923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,8,2,power_law_1.2,0.295033597946167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,8,2,power_law_1.2,0.39975039958953856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,4,32,power_law_1.2,0.030611199140548707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,2,power_law_1.01,0.5593599796295166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,8,2,power_law_1.2,0.5585919857025147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,4,32,power_law_1.2,0.033369600772857666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,8,2,power_law_1.2,0.7603583812713623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,4,32,power_law_1.2,0.033478400111198424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,8,2,power_law_1.2,0.9205183982849121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,2,power_law_1.2,3.630694580078125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,4,32,power_law_1.2,0.0335999995470047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,2,power_law_1.01,0.7838335990905761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,8,2,power_law_1.2,1.2924287796020508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,4,32,power_law_1.2,0.03800320029258728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,32,power_law_1.2,0.04161919951438904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,8,2,power_law_1.2,2.979884719848633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,32,power_law_1.2,0.04888319969177246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,2,power_law_1.01,1.0911999702453614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,32,power_law_1.2,0.05311359763145447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,32,power_law_1.2,0.06881279945373535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,2,power_law_1.01,1.1748736381530762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,32,power_law_1.2,0.07180160284042358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,32,power_law_1.2,0.09446399807929992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,2,power_law_1.01,1.205190372467041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,32,power_law_1.2,0.11864960193634033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,32,power_law_1.2,0.2123136043548584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,2,power_law_1.01,1.2328960418701171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,32,power_law_1.2,0.23428480625152587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,2,power_law_1.01,1.2370431900024415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,32,power_law_1.2,0.3266304016113281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,2,power_law_1.01,1.3440704345703125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,32,power_law_1.2,0.4235072135925293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,2,power_law_1.01,1.384108829498291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,32,power_law_1.2,0.9551103591918946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,2,power_law_1.01,1.4229567527770997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,128,balanced,0.08534933129946391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,128,balanced,0.08571199576059978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,128,balanced,0.08503466844558716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,128,balanced,0.08335999647776286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,128,balanced,0.08303466439247131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,128,balanced,0.09571199615796407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,128,balanced,0.09683199723561604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,128,balanced,0.09731733798980713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,128,balanced,0.10400000214576721
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,128,balanced,0.09574400385220845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,128,balanced,0.10101333260536194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,128,balanced,0.10585600137710571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,128,balanced,0.10776000221570332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,128,balanced,0.09507200121879578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,128,balanced,0.10538666447003682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,128,balanced,0.10400000214576721
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,128,balanced,0.11272533734639485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,128,balanced,0.10771200060844421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,2,power_law_1.01,1.405887985229492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,4,power_law_1.2,0.04177280068397522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,128,balanced,0.11943999926249187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,128,balanced,0.12454932928085327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,128,balanced,0.12902933359146118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,128,balanced,0.14866133530934653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,128,balanced,0.16875199476877847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,128,balanced,0.21380800008773804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,4,power_law_1.2,0.036550399661064145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,128,balanced,0.2479626735051473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,128,balanced,0.34626134236653644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,128,balanced,0.41656001408894855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,128,balanced,0.5079840024312338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,4,power_law_1.2,0.039084801077842714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,128,balanced,0.7625013192494711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,2,power_law_1.01,1.4467071533203124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,128,balanced,1.4655307133992512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,4,power_law_1.2,0.04348160028457641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,4,power_law_1.2,0.05000960230827332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,4,power_law_1.2,0.06236799955368042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,2,power_law_1.01,1.5241344451904297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,4,power_law_1.2,0.0661184012889862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,4,power_law_1.2,0.06447359919548035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,4,power_law_1.2,0.06755200028419495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,2,power_law_1.01,1.641708755493164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,4,power_law_1.2,0.06620799899101257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,4,power_law_1.2,0.07141119837760926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,4,power_law_1.2,0.07336959838867188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,2,power_law_1.01,1.7072383880615234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,4,power_law_1.2,0.07390080094337463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,4,power_law_1.2,0.07521920204162598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,4,power_law_1.2,0.08213120102882385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,2,power_law_1.01,1.8305152893066405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,4,power_law_1.2,0.08349440097808838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,4,power_law_1.2,0.09048320055007934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,4,power_law_1.2,0.10002559423446655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,2,power_law_1.01,1.820774459838867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,4,power_law_1.2,0.10995199680328369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,4,power_law_1.2,0.13844480514526367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,1,power_law_1.2,0.046387198567390445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,4,power_law_1.2,0.13996800184249877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,2,power_law_1.01,2.066067123413086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,1,power_law_1.2,0.053625601530075076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,4,power_law_1.2,0.17239680290222167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,1,power_law_1.2,0.0667519986629486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,4,power_law_1.2,0.2279360055923462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,1,power_law_1.2,0.08701440095901489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,2,power_law_1.01,2.3188352584838867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,4,power_law_1.2,0.3207808017730713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,1,power_law_1.2,0.1019711971282959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,4,power_law_1.2,0.3617216110229492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,1,power_law_1.2,0.1609279990196228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,4,power_law_1.2,0.5056128025054931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,1,power_law_1.2,0.17262719869613646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,16,power_law_1.2,0.09960960149765015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,2,power_law_1.01,2.85284481048584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,1,power_law_1.2,0.18193279504776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,4,power_law_1.2,0.6940288066864013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,1,power_law_1.2,0.18644479513168336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,16,power_law_1.2,0.151910400390625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,2688,1856,6,128,1,4,power_law_1.2,0.881004810333252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,1,power_law_1.2,0.19315840005874635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,2,power_law_1.01,3.054585647583008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,1,power_law_1.2,0.20615038871765137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,16,power_law_1.2,0.1355455994606018
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,1,power_law_1.2,0.2153536081314087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,2688,1856,6,128,1,4,power_law_1.2,1.342124843597412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,16,power_law_1.2,0.07589759826660156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,1,power_law_1.2,0.22470400333404542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,2,power_law_1.01,3.957843017578125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,1,power_law_1.2,0.2258944034576416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,16,power_law_1.2,0.09429759979248047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,1,power_law_1.2,0.2443903923034668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,2688,1856,6,128,1,4,power_law_1.2,2.847212791442871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,1,power_law_1.2,0.24726400375366211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,16,power_law_1.2,0.12104320526123047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,1,power_law_1.2,0.274399995803833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,2,power_law_1.01,4.830483245849609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,1,power_law_1.2,0.3116224050521851
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,16,power_law_1.2,0.1266816020011902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,1,power_law_1.2,0.3478847980499268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,1,power_law_1.2,0.4383359909057617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,16,power_law_1.2,0.14160000085830687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,1,power_law_1.2,0.506873607635498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,2,power_law_1.01,5.969977569580078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,1,power_law_1.2,0.6571839809417724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,16,power_law_1.2,0.14303359985351563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,1,power_law_1.2,0.8597184181213379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,16,power_law_1.2,0.16365439891815187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,1,power_law_1.2,1.1223360061645509
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,16,power_law_1.2,0.16046080589294434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,1,power_law_1.2,1.4641216278076172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,2,power_law_1.01,8.276019287109374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,16,power_law_1.2,0.17070080041885377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,1,power_law_1.2,1.9833343505859375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,16,power_law_1.2,0.16622079610824586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,1,power_law_1.2,2.5228544235229493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,16,power_law_1.2,0.17260799407958985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,16,balanced,0.048170665899912514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,16,balanced,0.04818666477998098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,16,balanced,0.04994666576385498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,16,balanced,0.07328533132870992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,16,balanced,0.0977280040582021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,16,balanced,0.14613866806030273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,16,balanced,0.14869866768519083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,16,balanced,0.15019733707110086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,16,balanced,0.1513813336690267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,16,balanced,0.1518239974975586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,16,power_law_1.2,0.18742400407791138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,16,balanced,0.1520639955997467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,1,power_law_1.2,3.0564672470092775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,16,balanced,0.15388266245524088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,16,balanced,0.15503467122713724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,16,balanced,0.15794133146603903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,16,balanced,0.16267200311024985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,16,balanced,0.16482667128245035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,16,balanced,0.17100266615549722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,16,balanced,0.18116267522176108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,16,balanced,0.19064533710479736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,16,balanced,0.20716800292332968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,16,balanced,0.23084266980489096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,16,balanced,0.26847465833028156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,16,balanced,0.3018239935239156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,16,balanced,0.39230398337046307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,16,power_law_1.2,0.19516799449920655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,16,balanced,0.4663039843241374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,16,balanced,0.6490613222122192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,16,balanced,0.7800107002258301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,16,balanced,0.9912213484446207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,2,power_law_1.01,16.097625732421875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,16,balanced,1.4505386352539062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,16,power_law_1.2,0.21294078826904297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,1,power_law_1.2,4.703641510009765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,16,balanced,2.806901295979818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,16,power_law_1.2,0.23644800186157228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,16,power_law_1.2,0.2761728048324585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,16,power_law_1.2,0.3226560115814209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,1,power_law_1.2,8.960947418212891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,16,power_law_1.2,0.3387840032577515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,16,power_law_1.2,0.42923521995544434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,16,power_law_1.2,0.5358848094940185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,16,power_law_1.2,0.70382719039917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,1,balanced,0.03908800085385641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,1,balanced,0.036703998843828835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,16,power_law_1.2,0.7553791999816895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,1,balanced,0.04001066585381826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,1,balanced,0.04671466847260793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,1,balanced,0.07313600182533264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,1,balanced,0.06825066606203715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,1,balanced,0.06961599985758464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,1,balanced,0.07239999870459239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,1,balanced,0.07419200241565704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,1,balanced,0.0738506664832433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,1,balanced,0.07550399998823802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,1,balanced,0.07915199796358745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,2,4,power_law_1.01,0.05681279897689819
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,1,balanced,0.08225599924723308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,1,balanced,0.08610666791598003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,1,balanced,0.09314133723576863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,2,4,power_law_1.01,0.06537600159645081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,1,balanced,0.09616532921791077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,1,balanced,0.10431999961535136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,2,4,power_law_1.01,0.05548160076141358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,1,balanced,0.1207413375377655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,1,balanced,0.13943466544151306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,1,balanced,0.17372800906499228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,2,4,power_law_1.01,0.05989760160446167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,1,balanced,0.19773866732915243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,16,power_law_1.2,1.1173248291015625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,1,balanced,0.29740800460179645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,2,4,power_law_1.01,0.05920000076293945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,1,balanced,0.3646080096562703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,2,4,power_law_1.01,0.05999360084533691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,1,balanced,0.5399093230565389
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,2,4,power_law_1.01,0.06241919994354248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,1,balanced,0.6983253161112467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,2,4,power_law_1.01,0.06467199921607972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,1,balanced,1.0432106653849285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,2,4,power_law_1.01,0.06613759994506836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,1,balanced,1.364240010579427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,2,4,power_law_1.01,0.06758400201797485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,2,4,power_law_1.01,0.06656640172004699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,1,balanced,1.6977386474609375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,2,4,power_law_1.01,0.0711296021938324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,2,4,power_law_1.01,0.07442560195922851
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,2,4,power_law_1.01,0.07919999957084656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,1,balanced,2.699157396952311
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,16,power_law_1.2,1.3109824180603027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,2,4,power_law_1.01,0.08911359906196595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,2,4,power_law_1.01,0.0911616027355194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,2,4,power_law_1.01,0.10713599920272827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,2,4,power_law_1.01,0.1309056043624878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,1,balanced,5.274208068847656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,2,4,power_law_1.01,0.14627840518951415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,16,8,power_law_1.01,0.0293503999710083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,2,4,power_law_1.01,0.1676800012588501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,16,8,power_law_1.01,0.029702401161193846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,2,4,power_law_1.01,0.2357503890991211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,16,8,power_law_1.01,0.02942720055580139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,2,4,power_law_1.01,0.3097791910171509
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,16,8,power_law_1.01,0.029407998919486998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,16,8,power_law_1.01,0.02930560111999512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,2,4,power_law_1.01,0.4272064208984375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,16,8,power_law_1.01,0.030508801341056824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,16,power_law_1.2,1.5645695686340333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,16,8,power_law_1.01,0.031353598833084105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,2,4,power_law_1.01,0.605836820602417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,16,8,power_law_1.01,0.030342400074005127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,2,4,power_law_1.01,0.615174388885498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,16,8,power_law_1.01,0.03154560029506683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,16,8,power_law_1.01,0.03160319924354553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,16,8,power_law_1.01,0.033471998572349546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,2,4,power_law_1.01,0.9461376190185546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,16,8,power_law_1.01,0.03436160087585449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,16,8,power_law_1.01,0.03530240058898926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,16,8,power_law_1.01,0.035020801424980166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,2,4,power_law_1.01,1.1301504135131837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,16,8,power_law_1.01,0.04013440012931824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,16,8,power_law_1.01,0.04408319890499115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,16,8,power_law_1.01,0.04690560102462769
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,2,4,power_law_1.01,1.641913604736328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,16,8,power_law_1.01,0.053401601314544675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,16,8,power_law_1.01,0.0555840015411377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,16,8,power_law_1.01,0.06643840074539184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,16,power_law_1.2,2.436319923400879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,16,8,power_law_1.01,0.07879040241241456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,16,8,power_law_1.01,0.11694719791412353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,2,4,power_law_1.01,2.5607872009277344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,16,8,power_law_1.01,0.134662401676178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,16,8,power_law_1.01,0.1881152033805847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,16,8,power_law_1.01,0.24878079891204835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,16,8,power_law_1.01,0.3515647888183594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,16,8,power_law_1.01,0.46799359321594236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,2,4,power_law_1.01,5.027199935913086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,16,8,power_law_1.01,0.5416895866394043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,16,8,power_law_1.01,0.8778304100036621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,16,8,power_law_1.01,1.6304191589355468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,16,power_law_1.2,4.798668670654297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,32,power_law_1.2,0.032416000962257385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,32,power_law_1.2,0.028921601176261903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,32,power_law_1.2,0.028857600688934327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,32,power_law_1.2,0.03079040050506592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,32,power_law_1.2,0.0297791987657547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,32,power_law_1.2,0.031142398715019226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,32,power_law_1.2,0.031123200058937074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,32,power_law_1.2,0.031148800253868104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,32,power_law_1.2,0.031673601269721983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,32,power_law_1.2,0.03182719945907593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,32,balanced,0.043951998154322304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,32,balanced,0.043712000052134194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,32,balanced,0.04181866844495138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,32,balanced,0.045456002155939736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,32,balanced,0.04569066564242045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,32,balanced,0.048165331284205117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,32,balanced,0.048842668533325195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,32,balanced,0.049770668148994446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,32,balanced,0.04990933338801066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,32,balanced,0.05073600014050802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,32,power_law_1.2,0.032441601157188416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,32,balanced,0.051674668987592064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,32,balanced,0.0499946673711141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,32,balanced,0.05403199791908264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,32,balanced,0.05383466680844625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,32,balanced,0.05793066819508871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,32,balanced,0.058287998040517174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,32,balanced,0.06126933296521505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,32,balanced,0.07338133454322815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,32,balanced,0.07372800012429555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,32,balanced,0.09896000226338704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,32,balanced,0.11956266562143962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,32,balanced,0.15090133746465048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,32,power_law_1.2,0.03247359991073608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,32,balanced,0.17162134250005087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,32,balanced,0.2276960015296936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,32,balanced,0.27224000295003253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,32,balanced,0.38551998138427734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,32,balanced,0.4845279852549235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,2,balanced,0.07202133536338806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,2,balanced,0.09603200356165568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,32,balanced,0.5825226704279581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,2,balanced,0.12727999687194824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,2,balanced,0.18735466400782266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,32,power_law_1.2,0.033478400111198424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,2,balanced,0.3073920011520386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,32,balanced,0.8911413351694742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,2,balanced,0.5510239998499552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,128,power_law_1.2,0.03460479974746704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,2,balanced,0.790992021560669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,32,balanced,1.7260373433430989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,2,balanced,0.7860213120778402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,2,balanced,0.7883360385894775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,128,power_law_1.2,0.0341376006603241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,32,power_law_1.2,0.03537279963493347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,2,balanced,0.7913866837819418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,2,balanced,0.7955359617869059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,2,balanced,0.7999359766642252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,128,power_law_1.2,0.03473280072212219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,2,balanced,0.8047786553700765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,2,balanced,0.8086400032043457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,32,power_law_1.2,0.03894400000572205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,2,balanced,0.8207093079884847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,2,balanced,0.8288586934407552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,128,power_law_1.2,0.03447679877281189
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,2,balanced,0.8403200308481852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,2,balanced,0.86517866452535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,32,power_law_1.2,0.039955198764801025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,2,balanced,0.8884853521982828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,128,power_law_1.2,0.03503359854221344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,2,balanced,0.9382080237070719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,2,balanced,0.9828373591105143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,128,power_law_1.2,0.03606399893760681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,2,balanced,1.0961226622263591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,32,power_law_1.2,0.041817599534988405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,2,balanced,1.1989013353983562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,128,power_law_1.2,0.03573760092258453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,2,balanced,1.40227206548055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,32,power_law_1.2,0.04997760057449341
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,2,balanced,1.7140533129374187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,128,power_law_1.2,0.03587839901447296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,2,balanced,2.1056319872538247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,128,power_law_1.2,0.03648000061511993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,32,power_law_1.2,0.04628480076789856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,2,balanced,2.8859945933024087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,128,power_law_1.2,0.036601600050926206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,32,power_law_1.2,0.05957760214805603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,2,balanced,3.4029385248819985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,128,power_law_1.2,0.03811199963092804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,32,power_law_1.2,0.06370559930801392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,128,power_law_1.2,0.03800320029258728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,2,balanced,5.141279856363933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,128,power_law_1.2,0.03847039937973022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,32,power_law_1.2,0.07885439991950989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,128,power_law_1.2,0.03834240138530731
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,32,power_law_1.2,0.09267200231552124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,128,power_law_1.2,0.042054399847984314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,2,balanced,9.664074579874674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,128,power_law_1.2,0.04256640076637268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,32,power_law_1.2,0.13864320516586304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,128,power_law_1.2,0.043628799915313723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,32,power_law_1.2,0.17994879484176635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,128,power_law_1.2,0.045747199654579164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,128,power_law_1.2,0.04881280064582825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,32,power_law_1.2,0.282694411277771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,128,power_law_1.2,0.055206400156021115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,128,power_law_1.2,0.06295040249824524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,32,power_law_1.2,0.3259711980819702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,128,power_law_1.2,0.08095999956130981
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,1,1,power_law_1.01,0.03922559916973114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,128,power_law_1.2,0.10142079591751099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,1,1,power_law_1.01,0.0584384024143219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,32,power_law_1.2,0.5430463790893555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,128,power_law_1.2,0.140665602684021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,1,1,power_law_1.01,0.07177600264549255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,128,power_law_1.2,0.17908480167388915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,1,1,power_law_1.01,0.09804159998893738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,1,1,power_law_1.01,0.12451839447021484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,128,power_law_1.2,0.28334081172943115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,32,power_law_1.2,0.7299327850341797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,1,1,power_law_1.01,0.1892351984977722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,1,1,power_law_1.01,0.19326080083847047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,128,power_law_1.2,0.36484479904174805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,1,1,power_law_1.01,0.19809279441833497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,1,128,power_law_1.2,0.3998080015182495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,1,1,power_law_1.01,0.20527360439300538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,1,1,power_law_1.01,0.20348799228668213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,32,power_law_1.2,1.5510592460632324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,1,1,power_law_1.01,0.21107199192047119
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,1,128,power_law_1.2,0.6433728218078614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,1,1,power_law_1.01,0.21688320636749267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,1,1,power_law_1.01,0.21818881034851073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,1,1,power_law_1.01,0.24011518955230712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,1,1,power_law_1.01,0.24118399620056152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,1,128,power_law_1.2,1.326784038543701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,1,1,power_law_1.01,0.2526400089263916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,1,1,power_law_1.01,0.2612351894378662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,64,power_law_1.01,0.07244160175323486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,1,1,power_law_1.01,0.28660480976104735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,1,power_law_1.01,0.308351993560791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,1,power_law_1.01,0.36097280979156493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,64,power_law_1.01,0.074726402759552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,1,power_law_1.01,0.4021759986877441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,1,power_law_1.01,0.5155072212219238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,64,power_law_1.01,0.05937920212745666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,1,power_law_1.01,0.6069183826446534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,1,power_law_1.01,0.8068608283996582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,64,power_law_1.01,0.06045439839363098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,1,power_law_1.01,0.9981632232666016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,1,power_law_1.01,1.3873600006103515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,64,power_law_1.01,0.06308479905128479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,1,power_law_1.01,1.7499071121215821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,32,1,power_law_1.01,0.031014400720596313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,32,1,power_law_1.01,0.03457280099391937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,64,power_law_1.01,0.0661952018737793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,32,1,power_law_1.01,0.03524479866027832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,32,1,power_law_1.01,0.03790720105171204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,1,power_law_1.01,2.1458879470825196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,32,1,power_law_1.01,0.03431040048599243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,32,1,power_law_1.01,0.046982398629188536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,32,1,power_law_1.01,0.04755840003490448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,64,power_law_1.01,0.06534399986267089
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,32,1,power_law_1.01,0.039654400944709775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,32,1,power_law_1.01,0.04538240134716034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,32,1,power_law_1.01,0.04746240079402923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,1,power_law_1.01,3.306304168701172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,32,1,power_law_1.01,0.04298880100250244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,32,1,power_law_1.01,0.0470335990190506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,16,power_law_1.2,0.098854398727417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,64,power_law_1.01,0.06772480010986329
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,32,1,power_law_1.01,0.05000320076942444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,32,1,power_law_1.01,0.052313601970672606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,32,1,power_law_1.01,0.05856639742851257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,32,1,power_law_1.01,0.05976960062980652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,16,power_law_1.2,0.18668800592422485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,32,1,power_law_1.01,0.061375999450683595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,64,power_law_1.01,0.06906880140304565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,32,1,power_law_1.01,0.07512959837913513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,32,1,power_law_1.01,0.08085119724273682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,32,1,power_law_1.01,0.09957759976387023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,16,power_law_1.2,0.10814080238342286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,32,1,power_law_1.01,0.10785919427871704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,1,power_law_1.01,6.452057647705078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,32,1,power_law_1.01,0.1463744044303894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,64,power_law_1.01,0.06692479848861695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,32,1,power_law_1.01,0.16142079830169678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,32,1,power_law_1.01,0.224288010597229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,16,power_law_1.2,0.08067839741706848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,32,1,power_law_1.01,0.28243839740753174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,32,1,power_law_1.01,0.40037121772766116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,64,power_law_1.01,0.0689791977405548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,32,1,power_law_1.01,0.5164095878601074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,16,power_law_1.2,0.09648000001907349
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,32,1,power_law_1.01,0.6353280067443847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,32,1,power_law_1.01,0.9530176162719727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,16,power_law_1.2,0.11483520269393921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,64,power_law_1.01,0.07310079932212829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,32,1,power_law_1.01,1.8556095123291017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,16,power_law_1.2,0.12157440185546875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,64,power_law_1.01,0.07094399929046631
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,16,power_law_1.2,0.15397759675979614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,64,power_law_1.01,0.07439360022544861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,16,power_law_1.2,0.15582720041275025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,64,power_law_1.01,0.08169599771499633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,16,power_law_1.2,0.15998079776763915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,16,power_law_1.2,0.16482559442520142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,64,power_law_1.01,0.08427519798278808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,16,power_law_1.2,0.16658560037612916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,64,power_law_1.01,0.09071360230445862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,1,2,power_law_1.01,0.10077439546585083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,1,2,power_law_1.01,0.13715200424194335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,16,power_law_1.2,0.16497279405593873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,1,2,power_law_1.01,0.14108799695968627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,1,2,power_law_1.01,0.20366721153259276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,1,2,power_law_1.01,0.2127295970916748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,64,power_law_1.01,0.10647679567337036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,1,2,power_law_1.01,0.21389439105987548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,1,2,power_law_1.01,0.22583041191101075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,16,power_law_1.2,0.1828160047531128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,1,2,power_law_1.01,0.21903998851776124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,1,2,power_law_1.01,0.22381439208984374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,1,2,power_law_1.01,0.23259520530700684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,1,2,power_law_1.01,0.2322688102722168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,1,2,power_law_1.01,0.24086399078369142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,64,power_law_1.01,0.1189568042755127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,1,2,power_law_1.01,0.24099841117858886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,16,power_law_1.2,0.19044480323791504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,1,2,power_law_1.01,0.2571199893951416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,1,2,power_law_1.01,0.28250880241394044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,1,2,power_law_1.01,0.3118016004562378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,1,2,power_law_1.01,0.3243904113769531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,1,2,power_law_1.01,0.3978816032409668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,1,balanced,0.09902399778366089
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,1,balanced,0.1360479990641276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,16,power_law_1.2,0.1831936001777649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,1,2,power_law_1.01,0.4188864231109619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,1,balanced,0.1502293348312378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,1,balanced,0.21340266863505045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,1,2,power_law_1.01,0.5487679958343505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,64,power_law_1.01,0.1449280023574829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,1,balanced,0.33907198905944824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,1,balanced,0.5874933401743571
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,1,2,power_law_1.01,0.7155519962310791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,1,balanced,0.8363040288289388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,1,balanced,0.8440106709798177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,1,2,power_law_1.01,0.8922687530517578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,1,balanced,0.8425707022349039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,16,power_law_1.2,0.19793920516967772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,1,balanced,0.8475146293640137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,1,2,power_law_1.01,1.1459839820861817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,1,balanced,0.8553120295206705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,1,balanced,0.8590826988220215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,1,2,power_law_1.01,1.6633855819702148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,1,balanced,0.865061362584432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,64,power_law_1.01,0.17204480171203612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,1,balanced,0.8785973389943441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,1,balanced,0.8907360235850016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,16,power_law_1.2,0.22347519397735596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,1,balanced,0.8957333564758301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,1,2,power_law_1.01,2.525555229187012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,1,balanced,0.9394453366597494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,1,balanced,0.9757493336995443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,1,balanced,1.0370559692382812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,1,balanced,1.1398506959279378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,1,2,power_law_1.01,3.6440895080566404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,16,power_law_1.2,0.24083199501037597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,1,balanced,1.1997919877370198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,64,power_law_1.01,0.21442561149597167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,1,balanced,1.3653279940287273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,1,balanced,1.5496373176574707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,1,balanced,1.8653333981831868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,1,2,power_law_1.01,4.422355270385742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,16,power_law_1.2,0.3005183935165405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,1,balanced,2.4051946004231772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,1,balanced,3.0685653686523438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,64,power_law_1.01,0.23401598930358886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,16,power_law_1.2,0.29633278846740724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,1,2,power_law_1.01,5.879270553588867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,1,balanced,4.205264091491699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,1,balanced,5.062783877054851
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,16,power_law_1.2,0.35676159858703616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,64,power_law_1.01,0.2965888023376465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,1,2,power_law_1.01,9.166649627685548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,1,balanced,7.552064259847005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,16,power_law_1.2,0.4182079792022705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,64,power_law_1.01,0.37107200622558595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,16,power_law_1.2,0.5508863925933838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,1,balanced,14.217812856038412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,16,power_law_1.2,0.6252799987792969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,1,2,power_law_1.01,16.73297882080078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,64,power_law_1.01,0.5172544002532959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,16,power_law_1.2,0.8842495918273926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,16,power_law_1.2,1.0934207916259766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,64,power_law_1.01,0.6380735874176026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,16,power_law_1.2,1.396134376525879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,1,64,power_law_1.01,0.7619391918182373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,16,power_law_1.2,1.769260787963867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,8,4,power_law_1.01,0.046316799521446225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,8,4,power_law_1.01,0.03935999870300293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,1,64,power_law_1.01,1.161849594116211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,16,8,balanced,0.031680000325044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,16,8,balanced,0.031189332405726116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,16,power_law_1.2,3.5046016693115236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,16,8,balanced,0.03345066557327906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,16,8,balanced,0.0314026673634847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,16,8,balanced,0.031898667414983116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,8,4,power_law_1.01,0.03906559944152832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,16,8,balanced,0.03152533372243246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,16,8,balanced,0.031397332747777305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,16,8,balanced,0.03372266640265783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,16,8,balanced,0.03357866654793421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,16,8,balanced,0.03378133227427801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,16,8,balanced,0.03344533344109853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,16,8,balanced,0.03365866591533025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,16,8,balanced,0.033674667278925575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,8,4,power_law_1.01,0.03903360068798065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,16,8,balanced,0.03552533437808355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,16,8,balanced,0.040149333576361336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,16,8,balanced,0.041690667470296226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,16,8,balanced,0.0414986660083135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,16,8,balanced,0.04632533093293508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,16,8,balanced,0.04979733129342397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,16,8,balanced,0.05596266686916351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,8,4,power_law_1.01,0.03856639862060547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,16,8,balanced,0.07563200096289317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,16,8,balanced,0.10502400000890096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,16,8,balanced,0.12500799695650736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,16,8,balanced,0.16921067237854004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,16,8,balanced,0.20723199844360352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,8,4,power_law_1.01,0.039263999462127684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,16,8,balanced,0.2937493324279785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,16,8,balanced,0.3680320183436076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,8,4,power_law_1.01,0.03857280015945434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,16,8,balanced,0.44918398062388104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,16,8,balanced,0.6781546274820963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,8,4,power_law_1.01,0.038899201154708865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,16,8,balanced,1.290618658065796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,8,4,power_law_1.01,0.04048640131950378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,8,4,power_law_1.01,0.039980798959732056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,8,4,power_law_1.01,0.04129279851913452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,8,4,power_law_1.01,0.04328320026397705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,8,4,power_law_1.01,0.04338560104370117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,8,4,power_law_1.01,0.048390400409698484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,8,4,power_law_1.01,0.05575039982795715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,8,4,power_law_1.01,0.06051200032234192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,8,4,power_law_1.01,0.060678398609161376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,8,4,power_law_1.01,0.07122560143470764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,8,4,power_law_1.01,0.08517119884490967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,8,4,power_law_1.01,0.10920959711074829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,8,4,power_law_1.01,0.11495039463043213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,8,4,power_law_1.01,0.1595199942588806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,8,4,power_law_1.01,0.15935360193252562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,8,4,power_law_1.01,0.2651263952255249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,8,4,power_law_1.01,0.3124480009078979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,1,64,power_law_1.01,2.365830421447754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,8,4,power_law_1.01,0.5215551853179932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,8,4,power_law_1.01,0.6511168003082275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,8,4,power_law_1.01,0.681056022644043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,8,4,power_law_1.01,0.9979264259338378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,8,4,power_law_1.01,2.011961555480957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,32,4,power_law_1.01,0.04245119988918304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,32,4,power_law_1.01,0.03329919874668121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,32,4,power_law_1.01,0.03189760148525238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,32,4,power_law_1.01,0.031948798894882204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,32,4,power_law_1.01,0.03178240060806274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,32,4,power_law_1.01,0.032467201352119446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,32,4,power_law_1.01,0.03283840119838714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,32,4,power_law_1.01,0.03421440124511719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,32,4,power_law_1.01,0.03473919928073883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,32,4,power_law_1.01,0.03468799889087677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,32,4,power_law_1.01,0.0345984011888504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,32,4,power_law_1.01,0.03661440014839172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,16,balanced,0.04382933179537455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,16,balanced,0.04786666731039683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,16,balanced,0.04554133117198944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,32,4,power_law_1.01,0.03761279881000519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,16,balanced,0.04568533102671305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,16,balanced,0.05171200136343638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,16,balanced,0.07507200042406718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,32,4,power_law_1.01,0.03818880021572113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,16,balanced,0.0916426678498586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,16,balanced,0.09212799866994222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,16,balanced,0.09291733304659526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,32,4,power_law_1.01,0.04270719885826111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,16,balanced,0.09218666950861613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,16,balanced,0.09502399961153667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,16,balanced,0.09854400157928467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,32,4,power_law_1.01,0.045433598756790164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,16,balanced,0.09661866227785747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,16,balanced,0.10002666711807251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,16,balanced,0.10683199763298035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,16,balanced,0.10848533113797505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,32,4,power_law_1.01,0.050419199466705325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,16,balanced,0.11691733201344807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,16,balanced,0.12763200203577676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,32,4,power_law_1.01,0.061510401964187625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,16,balanced,0.13703999916712442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,16,balanced,0.1586026648680369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,16,balanced,0.17984533309936523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,32,4,power_law_1.01,0.06528639793395996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,16,balanced,0.22286933660507202
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,16,balanced,0.2672533392906189
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,32,4,power_law_1.01,0.07948160171508789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,16,balanced,0.3520266612370809
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,16,balanced,0.4409173329671224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,32,4,power_law_1.01,0.10473599433898925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,16,balanced,0.6256853342056274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,32,4,power_law_1.01,0.12151679992675782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,16,balanced,0.7879947026570638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,32,4,power_law_1.01,0.17825280427932738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,16,balanced,0.9723467032114664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,32,4,power_law_1.01,0.2191231966018677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,16,balanced,1.4784480730692546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,32,4,power_law_1.01,0.24520320892333985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,32,4,power_law_1.01,0.3296128034591675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,16,balanced,2.881930669148763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,32,4,power_law_1.01,0.5322432041168212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,32,4,power_law_1.01,0.5246975898742676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,32,4,power_law_1.01,1.0172224044799805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,16,power_law_1.01,0.05637120008468628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,16,power_law_1.01,0.04058879911899567
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,32,4,power_law_1.01,1.9525184631347656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,16,power_law_1.01,0.036671999096870425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,16,power_law_1.01,0.03856000006198883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,2,balanced,0.041589332123597465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,16,power_law_1.01,0.04353919923305512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,2,balanced,0.05851200222969055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,2,balanced,0.08029866715272267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,2,balanced,0.12110400199890137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,2,balanced,0.2067520022392273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,2,balanced,0.21039466063181558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,2,balanced,0.20956265926361084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,16,power_law_1.01,0.05177599787712097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,2,balanced,0.2087413271268209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,2,balanced,0.21020267407099405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,2,balanced,0.2120373249053955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,2,balanced,0.21515733003616333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,2,balanced,0.2158613403638204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,16,power_law_1.01,0.05418239831924439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,2,balanced,0.21811199188232422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,2,balanced,0.223306675752004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,2,balanced,0.23110399643580118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,2,balanced,0.2333866755167643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,2,balanced,0.24116265773773193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,2,balanced,0.25809599955876666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,16,power_law_1.01,0.05692800283432007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,2,balanced,0.28178133567174274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,2,balanced,0.31692800919214886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,2,balanced,0.3474666674931844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,2,balanced,0.4424053430557251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,16,power_law_1.01,0.05772160291671753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,2,balanced,0.48659201463063556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,2,balanced,0.6763573487599691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,2,balanced,0.8076266447703043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,16,power_law_1.01,0.057222402095794676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,2,balanced,1.1345760027567546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,2,balanced,1.4599146842956543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,16,power_law_1.01,0.05902079939842224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,2,balanced,1.8016266822814941
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,16,power_law_1.01,0.060159999132156375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,2,balanced,2.8155412673950195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,16,power_law_1.01,0.06087039709091187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,16,power_law_1.01,0.06254079937934875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,2,balanced,5.4820906321207685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,16,power_law_1.01,0.06796799898147583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,16,power_law_1.01,0.07159680128097534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,16,power_law_1.01,0.07728639841079712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,16,power_law_1.01,0.08727679848670959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,16,power_law_1.01,0.09148799777030944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,16,power_law_1.01,0.11227519512176513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,16,power_law_1.01,0.11862399578094482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,8,power_law_1.01,0.030905601382255555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,16,power_law_1.01,0.16132479906082153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,8,power_law_1.01,0.02998400032520294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,16,power_law_1.01,0.2006335973739624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,8,power_law_1.01,0.02885119915008545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,8,power_law_1.01,0.030943998694419862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,16,power_law_1.01,0.2715967893600464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,32,balanced,0.05013333261013031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,32,balanced,0.052101333936055504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,32,balanced,0.050111999114354454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,32,balanced,0.049770668148994446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,32,balanced,0.05400000015894572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,32,balanced,0.05421866476535797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,32,balanced,0.055776000022888184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,32,balanced,0.05602666735649109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,32,balanced,0.0561653325955073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,8,power_law_1.01,0.03221760094165802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,32,balanced,0.058090666929880776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,32,balanced,0.05839466551939646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,32,balanced,0.0580266664425532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,32,balanced,0.059936001896858215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,32,balanced,0.06005333364009857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,16,power_law_1.01,0.31738240718841554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,32,balanced,0.06558933357397716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,32,balanced,0.06646933158238728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,32,balanced,0.06817600131034851
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,32,balanced,0.07523199915885925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,8,power_law_1.01,0.03487359881401062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,32,balanced,0.08030400176843007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,32,balanced,0.09276266892751057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,32,balanced,0.10936533411343892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,32,balanced,0.13371200362841287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,32,balanced,0.15423466761906943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,32,balanced,0.2025173306465149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,8,power_law_1.01,0.03563520014286041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,32,balanced,0.24814399083455405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,32,balanced,0.3383893171946208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,16,power_law_1.01,0.48676481246948244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,32,balanced,0.42485864957173664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,32,balanced,0.5268959999084473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,8,power_law_1.01,0.035622400045394895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,32,balanced,0.7759573459625244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,8,power_law_1.01,0.03671039938926697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,32,balanced,1.518602689107259
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,16,power_law_1.01,0.5724991798400879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,8,power_law_1.01,0.03804160058498383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,8,power_law_1.01,0.03925760090351105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,16,power_law_1.01,0.7260543823242187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,8,power_law_1.01,0.0414463996887207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,8,power_law_1.01,0.04227199852466583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,8,power_law_1.01,0.04327679872512817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,16,power_law_1.01,1.2281599998474122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,8,power_law_1.01,0.04710400104522705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,8,power_law_1.01,0.0486272007226944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,8,power_law_1.01,0.05111680030822754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,8,power_law_1.01,0.05868800282478333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,8,power_law_1.01,0.06659839749336242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,32,8,power_law_1.01,0.04864639937877655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,16,power_law_1.01,2.217900848388672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,8,power_law_1.01,0.08079360127449035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,8,power_law_1.01,0.08758400082588196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,8,power_law_1.01,0.13355519771575927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,8,power_law_1.01,0.1686527967453003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,32,8,power_law_1.01,0.051648002862930295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,8,power_law_1.01,0.24169600009918213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,8,power_law_1.01,0.2991552114486694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,8,power_law_1.01,0.4587071895599365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,32,8,power_law_1.01,0.06041600108146668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,16,balanced,0.052576000491778054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,16,balanced,0.05273599922657013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,16,balanced,0.05197866757710775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,8,power_law_1.01,0.5389823913574219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,16,balanced,0.06700266897678375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,16,balanced,0.08609599868456523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,16,balanced,0.09404266873995464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,16,balanced,0.09645866354306538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,16,balanced,0.09579199552536011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,16,balanced,0.09642133116722107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,16,balanced,0.09616532921791077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,16,balanced,0.09782399733861287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,16,balanced,0.09844799836476643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,16,balanced,0.09936533371607463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,16,balanced,0.09861333171526591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,16,balanced,0.10471466183662415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,16,balanced,0.10672533512115479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,16,balanced,0.10733333230018616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,16,balanced,0.11571733156840007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,16,balanced,0.12194666266441345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,16,balanced,0.14486400286356607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,32,8,power_law_1.01,0.04709759950637817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,8,power_law_1.01,0.6351424217224121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,16,balanced,0.14850133657455444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,16,balanced,0.1813653310139974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,16,balanced,0.19767999649047852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,16,balanced,0.27430933713912964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,16,balanced,0.31100799640019733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,16,balanced,0.44620800018310547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,16,balanced,0.5288266738255819
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,16,balanced,0.6164693435033163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,16,balanced,0.9581759770711263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,16,balanced,1.8057173093159993
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,8,power_law_1.01,1.0324543952941894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,32,8,power_law_1.01,0.05030400156974792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,32,8,power_law_1.01,0.05551360249519348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,8,power_law_1.01,2.0520959854125977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,1,balanced,0.05808533231417338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,1,balanced,0.08950400352478027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,1,balanced,0.10213866829872131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,1,balanced,0.1545866628487905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,1,balanced,0.2541439930597941
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,1,balanced,0.45041600863138836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,1,balanced,0.4520159959793091
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,1,balanced,0.45846935113271076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,1,balanced,0.4639413356781006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,1,balanced,0.46533334255218506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,32,8,power_law_1.01,0.05745919942855835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,1,balanced,0.4681653181711833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,1,balanced,0.4800853331883748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,1,balanced,0.48310399055480957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,1,balanced,0.4880746603012085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,1,balanced,0.5029173294703165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,1,balanced,0.5080906550089518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,1,balanced,0.5278079907099406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,1,balanced,0.5623626708984375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,1,balanced,0.5926186641057333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,1,balanced,0.6605120102564493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,1,balanced,0.7320160071055094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,1,balanced,0.9301919937133789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,32,8,power_law_1.01,0.06170880198478699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,1,balanced,1.0365173021952312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,1,balanced,1.4355252583821614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,1,balanced,1.698421319325765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,1,balanced,2.39192533493042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,1,balanced,3.009530703226725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,32,8,power_law_1.01,0.06501759886741638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,1,balanced,3.728917439778646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,1,balanced,5.5365549723307295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,32,8,power_law_1.01,0.06584320068359376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,32,8,power_law_1.01,0.06833279728889466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,1,balanced,10.539786656697592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,4,balanced,0.07337066531181335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,4,balanced,0.10288533568382263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,4,balanced,0.16145066420237222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,4,balanced,0.25960532824198407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,4,balanced,0.46562667687733966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,4,balanced,0.5710826714833578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,32,8,power_law_1.01,0.07297279834747314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,4,balanced,0.5737706820170084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,4,balanced,0.574997345606486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,4,balanced,0.575055996576945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,4,balanced,0.5785653193791708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,4,balanced,0.5781386693318685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,4,balanced,0.5799733400344849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,4,balanced,0.5814773241678873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,4,balanced,0.5873759984970093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,4,balanced,0.5950506528218588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,4,balanced,0.5997440020243326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,4,balanced,0.6086879968643188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,4,balanced,0.6499626636505127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,32,8,power_law_1.01,0.07406719923019409
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,4,balanced,0.6352373361587524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,4,balanced,0.8208693663279215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,4,balanced,0.7028480370839437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,4,balanced,0.9095626672108968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,4,balanced,0.8309760093688965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,4,balanced,1.3746293385823567
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,4,balanced,1.2396480242411296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,32,8,power_law_1.01,0.07939839959144593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,4,balanced,1.6784000396728516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,4,balanced,2.2184853553771973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,128,power_law_1.2,0.05839999914169312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,4,balanced,2.5113706588745117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,4,balanced,3.842858632405599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,32,8,power_law_1.01,0.08969600200653076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,128,power_law_1.2,0.0573248028755188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,4,balanced,7.21992556254069
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,32,8,power_law_1.01,0.09185280203819275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,128,power_law_1.2,0.0527616024017334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,32,8,power_law_1.01,0.10131200551986694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,128,power_law_1.2,0.053446400165557864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,32,8,power_law_1.01,0.12209279537200927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,128,power_law_1.2,0.05430399775505066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,1,power_law_1.01,0.04921599924564361
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,32,8,power_law_1.01,0.1281599998474121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,1,power_law_1.01,0.05613440275192261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,128,power_law_1.2,0.05443840026855469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,1,power_law_1.01,0.06511359810829162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,32,8,power_law_1.01,0.16186239719390869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,128,power_law_1.2,0.05452160239219665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,1,power_law_1.01,0.09638400077819824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,1,power_law_1.01,0.13155839443206788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,32,8,power_law_1.01,0.18670079708099366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,128,power_law_1.2,0.05504000186920166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,1,power_law_1.01,0.1604864001274109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,1,power_law_1.01,0.21838719844818116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,128,power_law_1.2,0.055929601192474365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,32,8,power_law_1.01,0.2533760070800781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,1,power_law_1.01,0.22544000148773194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,64,power_law_1.01,0.03308799862861633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,128,power_law_1.2,0.05619840025901794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,64,power_law_1.01,0.028518399596214293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,1,power_law_1.01,0.23796479701995848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,32,8,power_law_1.01,0.31531519889831544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,64,power_law_1.01,0.02852480113506317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,1,power_law_1.01,0.23664638996124268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,64,power_law_1.01,0.029139199852943422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,128,power_law_1.2,0.05591679811477661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,64,power_law_1.01,0.02980479896068573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,1,power_law_1.01,0.25226240158081054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,64,power_law_1.01,0.030444800853729248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,32,8,power_law_1.01,0.42595839500427246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,64,power_law_1.01,0.0308351993560791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,1,power_law_1.01,0.25966720581054686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,128,power_law_1.2,0.05942400097846985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,64,power_law_1.01,0.030707201361656188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,64,power_law_1.01,0.03076480031013489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,1,power_law_1.01,0.27409920692443845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,32,8,power_law_1.01,0.5335616111755371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,64,power_law_1.01,0.030847999453544616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,128,power_law_1.2,0.060236799716949466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,1,power_law_1.01,0.2896320104598999
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,64,power_law_1.01,0.031513598561286923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,64,power_law_1.01,0.03240320086479187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,1,power_law_1.01,0.31656320095062257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,64,power_law_1.01,0.032543998956680295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,128,power_law_1.2,0.06202239990234375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,64,power_law_1.01,0.03367680013179779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,32,8,power_law_1.01,0.7944640159606934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,1,power_law_1.01,0.3325952053070068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,64,power_law_1.01,0.036371201276779175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,64,power_law_1.01,0.037510401010513304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,1,power_law_1.01,0.33697919845581054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,128,power_law_1.2,0.06836479902267456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,64,power_law_1.01,0.03806079924106598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,64,power_law_1.01,0.04136320054531097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,1,power_law_1.01,0.3979648113250732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,32,8,power_law_1.01,1.061945629119873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,64,power_law_1.01,0.04481280148029328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,128,power_law_1.2,0.06937599778175355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,1,power_law_1.01,0.4499199867248535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,64,power_law_1.01,0.04787839949131012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,64,power_law_1.01,0.053395199775695804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,1,power_law_1.01,0.5461376190185547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,64,power_law_1.01,0.06753280162811279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,128,power_law_1.2,0.0738431990146637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,32,8,power_law_1.01,1.2884991645812989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,64,power_law_1.01,0.0735040009021759
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,1,power_law_1.01,0.6151872158050538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,64,power_law_1.01,0.09937279820442199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,1,power_law_1.01,0.7855552196502685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,128,power_law_1.2,0.08863999843597412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,64,power_law_1.01,0.13040000200271606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,64,power_law_1.01,0.19972480535507203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,1,power_law_1.01,0.9691647529602051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,32,8,power_law_1.01,1.9705791473388672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,128,power_law_1.2,0.09596160054206848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,64,power_law_1.01,0.24068479537963866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,1,power_law_1.01,1.3021759986877441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,64,power_law_1.01,0.2825151920318604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,128,power_law_1.2,0.12602880001068115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,1,power_law_1.01,1.6417152404785156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,64,power_law_1.01,0.49027199745178224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,128,power_law_1.2,0.15075199604034423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,1,power_law_1.01,2.310092735290527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,32,8,power_law_1.01,3.8808128356933596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,128,power_law_1.2,0.19941120147705077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,64,power_law_1.01,0.966214370727539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,1,power_law_1.01,2.9893951416015625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,128,power_law_1.2,0.20982398986816406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,1,power_law_1.01,3.641004943847656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,128,power_law_1.2,0.2737663984298706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,1,power_law_1.01,5.605670547485351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,128,power_law_1.2,0.36368000507354736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,128,power_law_1.2,0.5140416145324707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,1,power_law_1.01,10.854764556884765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,64,power_law_1.2,0.0604095995426178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,8,2,power_law_1.2,0.03301759958267212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,128,power_law_1.2,0.5696896076202392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,8,2,power_law_1.2,0.031763198971748355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,8,2,power_law_1.2,0.031744000315666196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,8,2,power_law_1.2,0.03336319923400879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,8,2,power_law_1.2,0.03449600040912628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,8,2,power_law_1.2,0.03688960075378418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,8,2,power_law_1.2,0.04230400025844574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,64,power_law_1.2,0.0546176016330719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,8,2,power_law_1.2,0.03946239948272705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,8,2,power_law_1.2,0.041119998693466185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,8,2,power_law_1.2,0.04492799937725067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,8,2,power_law_1.2,0.045440000295639035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,8,2,power_law_1.2,0.04810880124568939
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,8,2,power_law_1.2,0.0500927984714508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,8,2,power_law_1.2,0.05560320019721985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,64,power_law_1.2,0.048614400625228885
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,8,2,power_law_1.2,0.060031998157501223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,8,2,power_law_1.2,0.0659775972366333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,8,2,power_law_1.2,0.07154560089111328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,8,2,power_law_1.2,0.0827072024345398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,8,2,power_law_1.2,0.09293439984321594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,2,128,power_law_1.2,0.7965184211730957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,8,2,power_law_1.2,0.11206400394439697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,64,power_law_1.2,0.04981760084629059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,8,2,power_law_1.2,0.14276479482650756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,8,2,power_law_1.2,0.168614399433136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,8,2,power_law_1.2,0.22948479652404785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,8,2,power_law_1.2,0.32415359020233153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,8,2,power_law_1.2,0.4057663917541504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,64,power_law_1.2,0.049593600630760196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,8,2,power_law_1.2,0.5355519771575927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,8,2,power_law_1.2,0.702400016784668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,8,2,power_law_1.2,0.9494912147521972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,64,power_law_1.2,0.0506496012210846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,8,2,power_law_1.2,1.358022403717041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,64,power_law_1.2,0.05246719717979431
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,2,128,power_law_1.2,1.1788928031921386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,8,2,power_law_1.2,2.7436031341552733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,64,power_law_1.2,0.05237759947776795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,64,power_law_1.2,0.052934402227401735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,64,power_law_1.2,0.05262079834938049
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,2,power_law_1.01,0.04671359956264496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,64,power_law_1.2,0.05367680191993714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,64,power_law_1.2,0.055622398853302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,2,power_law_1.01,0.05666559934616089
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,64,power_law_1.2,0.05605760216712952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,2,power_law_1.01,0.07639679908752442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,2,128,power_law_1.2,3.339539337158203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,64,power_law_1.2,0.056908798217773435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,2,power_law_1.01,0.07211520075798035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,64,power_law_1.2,0.062489598989486694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,2,power_law_1.01,0.09850879907608032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,64,power_law_1.2,0.06484479904174804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,2,power_law_1.01,0.12004480361938477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,64,power_law_1.2,0.06857600212097167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,4,balanced,0.0904266635576884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,4,balanced,0.16301332910855612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,4,balanced,0.2004106640815735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,4,balanced,0.3500426610310872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,4,balanced,0.6099786758422852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,2,power_law_1.01,0.13645440340042114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,4,balanced,0.8318560123443604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,64,power_law_1.2,0.07393280267715455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,4,balanced,0.837994654973348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,4,balanced,0.8395360310872396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,4,balanced,0.8398773670196533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,4,balanced,0.8441279729207357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,4,balanced,0.8459093570709229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,4,balanced,0.8536586761474609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,2,power_law_1.01,0.16353280544281007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,4,balanced,0.8560372988382975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,64,power_law_1.2,0.08412799835205079
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,4,balanced,0.859813372294108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,4,balanced,0.8733973503112793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,4,balanced,0.8797279993693033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,4,balanced,0.8869439760843912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,4,balanced,1.0663093725840251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,4,balanced,0.9390026728312174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,2,power_law_1.01,0.16799360513687134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,64,power_law_1.2,0.10640000104904175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,4,balanced,1.2274773120880127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,4,balanced,1.0500266551971436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,4,balanced,1.5637706120808919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,4,balanced,1.2358506520589192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,2,power_law_1.01,0.17690240144729613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,4,balanced,2.0680160522460938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,64,power_law_1.2,0.12302080392837525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,4,balanced,1.8498934110005696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,4,balanced,3.074437459309896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,2,power_law_1.01,0.18883199691772462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,4,balanced,2.7959893544514975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,1,4,balanced,3.7447732289632163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,8,balanced,0.030048000315825146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,64,power_law_1.2,0.157151997089386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,8,balanced,0.032298666735490165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,8,balanced,0.029557332396507263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,8,balanced,0.03162133445342382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,8,balanced,0.03389333436886469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,8,balanced,0.03781333317359289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,2,power_law_1.01,0.19329919815063476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,8,balanced,0.03803733239571253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,8,balanced,0.03797333439191183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,8,balanced,0.036271999279658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,8,balanced,0.03751999884843826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,8,balanced,0.037903999288876854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,8,balanced,0.037717332442601524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,8,balanced,0.037871999045213066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,8,balanced,0.0378506655494372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,8,balanced,0.03994666785001755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,8,balanced,0.0418453315893809
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,8,balanced,0.043840001026789345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,1,4,balanced,5.257274627685547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,8,balanced,0.05397333204746246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,8,balanced,0.04621866842110952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,8,balanced,0.05921066800753275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,8,balanced,0.05894933144251505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,8,balanced,0.07633066674073537
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,8,balanced,0.07800533374150594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,8,balanced,0.10106133421262105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,8,balanced,0.12050666411717732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,8,balanced,0.16721065839131674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,8,balanced,0.17653866608937582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,4,8,balanced,0.21369065841039023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,2,power_law_1.01,0.2001215934753418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,4,8,balanced,0.2948586742083232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,4,8,balanced,0.5355626742045084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,64,power_law_1.2,0.18269439935684204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,1,4,balanced,10.120341618855795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,2,power_law_1.01,0.21028480529785157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,64,power_law_1.2,0.2579648017883301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,2,power_law_1.01,0.2257983922958374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,64,power_law_1.2,0.3497600078582764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,2,power_law_1.01,0.24644479751586915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,2,power_law_1.01,0.25905280113220214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,64,power_law_1.2,0.4711359977722168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,2,power_law_1.01,0.3066688060760498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,64,power_law_1.2,0.5950208187103272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,2,power_law_1.01,0.3212480068206787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,64,power_law_1.2,0.7529664039611816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,2,power_law_1.01,0.38631041049957277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,2,power_law_1.01,0.44664320945739744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,64,power_law_1.2,1.1570176124572753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,2,power_law_1.01,0.566431999206543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,2,power_law_1.01,0.6885056018829345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,1,4,power_law_1.2,0.0749504029750824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,1,4,power_law_1.2,0.077702397108078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,1,4,power_law_1.2,0.06355839967727661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,32,power_law_1.01,0.08520320057868958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,1,4,power_law_1.2,0.07230719923973083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,1,4,power_law_1.2,0.07336320281028748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,1,4,power_law_1.2,0.07457919716835022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,1,4,power_law_1.2,0.07479040026664734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,1,4,power_law_1.2,0.0741375982761383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,64,power_law_1.2,2.5956352233886717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,2,power_law_1.01,0.9005696296691894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,1,4,power_law_1.2,0.07710719704627991
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,1,4,power_law_1.2,0.07707520127296448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,1,4,power_law_1.2,0.07879040241241456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,1,4,power_law_1.2,0.08153600096702576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,32,power_law_1.01,0.12108160257339477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,1,4,power_law_1.2,0.08650239706039428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,1,4,power_law_1.2,0.09259520173072815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,1,4,power_law_1.2,0.0996288001537323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,1,4,power_law_1.2,0.10184320211410522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,1,4,power_law_1.2,0.12352639436721802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,2,power_law_1.01,1.1728896141052245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,1,4,power_law_1.2,0.14034559726715087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,1,4,power_law_1.2,0.1679744005203247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,32,power_law_1.01,0.10509439706802368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,1,4,power_law_1.2,0.22862720489501953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,1,4,power_law_1.2,0.2969919919967651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,1,4,power_law_1.2,0.42912001609802247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,1,4,power_law_1.2,0.43815040588378906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,32,power_law_1.01,0.06542720198631287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,1,4,power_law_1.2,0.7741312026977539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,2,power_law_1.01,1.5647040367126466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,1,4,power_law_1.2,0.9785152435302734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,1,4,power_law_1.2,1.468115234375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,32,power_law_1.01,0.0753279983997345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,1,4,power_law_1.2,1.9812351226806642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,2,power_law_1.01,1.9684415817260743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,32,power_law_1.01,0.08055040240287781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,1,4,power_law_1.2,2.8880128860473633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,32,power_law_1.01,0.0880895972251892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,1,4,power_law_1.2,3.768735885620117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,2,power_law_1.01,2.457804870605469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,32,power_law_1.01,0.09352959990501404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,1,4,power_law_1.2,7.98115234375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,32,power_law_1.01,0.0945792019367218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,32,power_law_1.01,0.09317759871482849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,2,power_law_1.01,3.85362548828125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,4,balanced,0.05328000088532766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,4,balanced,0.08428266644477844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,4,balanced,0.09851732850074768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,4,balanced,0.1465173363685608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,32,power_law_1.01,0.09660159945487976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,4,balanced,0.24889600276947021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,4,balanced,0.4429973363876343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,4,balanced,0.6365013519922892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,4,balanced,0.6381280024846395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,4,balanced,0.6402773459752401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,4,balanced,0.6397493282953898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,4,balanced,0.6434666713078817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,32,power_law_1.01,0.09989759922027588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,4,balanced,0.6497759819030762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,4,balanced,0.6501226822535197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,4,balanced,0.6572746833165487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,4,balanced,0.6677760283152262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,4,balanced,0.6698293685913086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,4,balanced,0.680351972579956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,4,balanced,0.6997919877370199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,32,power_law_1.01,0.09542400240898133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,4,balanced,0.717034657796224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,4,balanced,0.7568000157674154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,4,balanced,0.792090654373169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,4,balanced,0.8594133059183756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,2,power_law_1.01,7.261318206787109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,4,balanced,0.9376107056935629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,4,balanced,1.076906681060791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,32,power_law_1.01,0.10309120416641235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,4,balanced,1.2538987000783284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,4,balanced,1.5137333869934082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,4,balanced,1.969312032063802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,32,power_law_1.01,0.11189759969711303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,4,balanced,2.4064586957295737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,4,balanced,3.4661439259847007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,32,power_law_1.01,0.10947840213775635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,4,balanced,6.580448150634766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,32,power_law_1.01,0.1161471962928772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,32,power_law_1.01,0.1213312029838562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,1,power_law_1.2,0.030873599648475646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,1,power_law_1.2,0.03628160059452057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,32,power_law_1.01,0.13553279638290405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,1,power_law_1.2,0.043808001279830935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,1,power_law_1.2,0.05472000241279602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,32,power_law_1.01,0.15070719718933107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,1,power_law_1.2,0.06783360242843628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,1,power_law_1.2,0.1028223991394043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,32,power_law_1.01,0.17634559869766236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,1,power_law_1.2,0.11000319719314575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,1,power_law_1.2,0.11265920400619507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,1,power_law_1.2,0.11662080287933349
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,32,power_law_1.01,0.22302720546722413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,1,power_law_1.2,0.11993600130081176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,1,power_law_1.2,0.12319359779357911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,32,power_law_1.01,0.26216959953308105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,1,power_law_1.2,0.12780159711837769
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,16,power_law_1.2,0.05451520085334778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,1,power_law_1.2,0.133951997756958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,16,power_law_1.2,0.04560000002384186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,1,power_law_1.2,0.13335039615631103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,32,power_law_1.01,0.34079999923706056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,16,power_law_1.2,0.043219199776649474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,1,power_law_1.2,0.14030079841613768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,1,power_law_1.2,0.14381439685821534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,16,power_law_1.2,0.04392319917678833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,1,power_law_1.2,0.150655996799469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,32,power_law_1.01,0.44741120338439944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,16,power_law_1.2,0.045132800936698914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,1,power_law_1.2,0.16899839639663697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,16,power_law_1.2,0.049158400297164916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,1,power_law_1.2,0.18595839738845826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,16,power_law_1.2,0.049472001194953916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,1,power_law_1.2,0.2249216079711914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,32,power_law_1.01,0.6135615825653076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,16,power_law_1.2,0.051641601324081424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,1,power_law_1.2,0.23758718967437745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,16,power_law_1.2,0.05082880258560181
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,1,power_law_1.2,0.2911871910095215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,16,power_law_1.2,0.05287039875984192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,32,2,power_law_1.01,0.05084800124168396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,1,power_law_1.2,0.345414400100708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,32,power_law_1.01,0.7768959999084473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,16,power_law_1.2,0.057120001316070555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,1,power_law_1.2,0.4512320041656494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,16,power_law_1.2,0.05792639851570129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,1,power_law_1.2,0.5579904079437256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,16,power_law_1.2,0.060703998804092406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,32,power_law_1.01,1.0072128295898437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,16,power_law_1.2,0.06376320123672485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,1,power_law_1.2,0.7728256225585938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,32,2,power_law_1.01,0.052102398872375486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,16,power_law_1.2,0.07033600211143494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,1,power_law_1.2,0.9912639617919922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,16,power_law_1.2,0.07278079986572265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,32,power_law_1.01,1.47957124710083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,16,power_law_1.2,0.07332479953765869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,2,1,power_law_1.2,1.2043968200683595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,32,2,power_law_1.01,0.060761600732803345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,16,power_law_1.2,0.08756480216979981
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,16,power_law_1.2,0.10255359411239624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,2,1,power_law_1.2,1.840870475769043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,16,power_law_1.2,0.13048959970474244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,16,power_law_1.2,0.16128640174865722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,32,2,power_law_1.01,0.06228479743003845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,16,power_law_1.2,0.19530240297317505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,32,power_law_1.01,3.031488037109375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,2,1,power_law_1.2,3.589990234375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,16,power_law_1.2,0.25829119682312013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,32,2,power_law_1.01,0.08212479948997498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,16,power_law_1.2,0.3837696075439453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,16,power_law_1.2,0.4506944179534912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,32,2,power_law_1.01,0.10124800205230713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,16,power_law_1.2,0.6802048206329345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,16,power_law_1.2,0.9587072372436524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,32,2,power_law_1.01,0.1147968053817749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,16,power_law_1.2,1.0210944175720216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,32,2,power_law_1.01,0.13014400005340576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,16,power_law_1.2,1.9031679153442382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,32,2,power_law_1.01,0.1317952036857605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,32,2,power_law_1.01,0.13880959749221802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,16,power_law_1.2,3.3527294158935548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,32,2,power_law_1.01,0.1409216046333313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,16,8,balanced,0.031557333966096245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,16,8,balanced,0.033488000432650246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,16,8,balanced,0.031541332602500916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,16,8,balanced,0.029461334149042766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,16,8,balanced,0.029802667597929638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,16,8,balanced,0.029391999046007793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,16,8,balanced,0.029461334149042766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,8,8,power_law_1.01,0.03701759874820709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,16,8,balanced,0.03136533250411352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,16,8,balanced,0.03138133386770884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,16,8,balanced,0.029311999678611755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,16,8,balanced,0.03133866687615713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,16,8,balanced,0.031328000128269196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,8,8,power_law_1.01,0.037555199861526486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,16,8,balanced,0.03143466760714849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,16,8,balanced,0.03278933217128118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,16,8,balanced,0.035536001125971474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,16,8,balanced,0.03699733316898346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,8,8,power_law_1.01,0.03547520041465759
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,16,8,balanced,0.03788800040880839
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,16,8,balanced,0.038880000511805214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,16,8,balanced,0.03980266551176707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,16,8,balanced,0.04377066592375437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,8,8,power_law_1.01,0.03522560000419617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,16,8,balanced,0.04684266448020935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,16,8,balanced,0.061903998255729675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,16,8,balanced,0.06136000156402588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,8,8,power_law_1.01,0.035206401348114015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,16,8,balanced,0.08266133566697438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,16,8,balanced,0.09005332986513774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,16,8,balanced,0.12430399656295776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,32,2,power_law_1.01,0.1503872036933899
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,8,8,power_law_1.01,0.03623040020465851
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,16,8,balanced,0.13864533106486002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,16,8,balanced,0.16262933611869812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,8,8,power_law_1.01,0.03709439933300018
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,16,8,balanced,0.23769599199295044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,16,8,balanced,0.40940264860788983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,8,8,power_law_1.01,0.03751679956912994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,8,8,power_law_1.01,0.0377344012260437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,8,8,power_law_1.01,0.03776639997959137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,8,8,power_law_1.01,0.03813759982585907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,8,8,power_law_1.01,0.039724799990653994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,8,8,power_law_1.01,0.04071680009365082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,8,8,power_law_1.01,0.043136000633239746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,8,8,power_law_1.01,0.05106559991836548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,32,2,power_law_1.01,0.15345280170440673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,8,8,power_law_1.01,0.054688000679016115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,8,8,power_law_1.01,0.057683199644088745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,8,8,power_law_1.01,0.06577280163764954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,8,8,power_law_1.01,0.06986879706382751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,8,8,power_law_1.01,0.08229119777679443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,8,8,power_law_1.01,0.09708799719810486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,8,8,power_law_1.01,0.13634560108184815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,8,8,power_law_1.01,0.16167680025100709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,32,2,power_law_1.01,0.15917439460754396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,8,8,power_law_1.01,0.2110208034515381
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,8,8,power_law_1.01,0.27381119728088377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,8,8,power_law_1.01,0.36122879981994627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,8,8,power_law_1.01,0.5062272071838378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,32,2,power_law_1.01,0.16749440431594848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,8,8,power_law_1.01,0.6249599933624268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,8,8,power_law_1.01,0.9166912078857422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,8,8,power_law_1.01,1.636729621887207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,32,2,power_law_1.01,0.17612160444259645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,32,2,power_law_1.01,0.18744959831237792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,32,4,power_law_1.01,0.035206401348114015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,32,4,power_law_1.01,0.032313600182533264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,32,2,power_law_1.01,0.2117055892944336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,32,4,power_law_1.01,0.032128000259399415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,32,4,power_law_1.01,0.031711998581886294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,32,4,power_law_1.01,0.03160960078239441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,32,4,power_law_1.01,0.033292800188064575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,32,4,power_law_1.01,0.03362559974193573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,32,4,power_law_1.01,0.0341376006603241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,32,4,power_law_1.01,0.033958399295806886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,32,4,power_law_1.01,0.03480319976806641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,32,2,power_law_1.01,0.23288319110870362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,32,4,power_law_1.01,0.035750401020050046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,32,4,power_law_1.01,0.03651840090751648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,32,4,power_law_1.01,0.03716480135917664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,32,4,power_law_1.01,0.038700801134109494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,32,4,power_law_1.01,0.04475519955158234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,32,4,power_law_1.01,0.0472896009683609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,32,4,power_law_1.01,0.05059199929237366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,1,power_law_1.01,0.05098239779472351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,32,4,power_law_1.01,0.06277120113372803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,32,4,power_law_1.01,0.07197440266609192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,32,2,power_law_1.01,0.28871679306030273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,32,4,power_law_1.01,0.08764799833297729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,1,power_law_1.01,0.07245439887046815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,32,4,power_law_1.01,0.11308159828186035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,32,4,power_law_1.01,0.16091519594192505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,32,4,power_law_1.01,0.2043071985244751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,1,power_law_1.01,0.08984959721565247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,32,4,power_law_1.01,0.2745856046676636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,32,4,power_law_1.01,0.3125760078430176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,1,power_law_1.01,0.12128000259399414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,32,4,power_law_1.01,0.52740478515625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,32,2,power_law_1.01,0.347052788734436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,32,4,power_law_1.01,0.6541056156158447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,1,power_law_1.01,0.1593984007835388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,32,4,power_law_1.01,0.8065728187561035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,1,power_law_1.01,0.24403200149536133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,32,4,power_law_1.01,1.3720959663391112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,32,2,power_law_1.01,0.439251184463501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,1,power_law_1.01,0.268505597114563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,1,power_law_1.01,0.2736128091812134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,32,4,power_law_1.01,2.6356224060058593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,1,power_law_1.01,0.28524160385131836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,2,balanced,0.05773866673310598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,2,balanced,0.06632000207901001
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,2,balanced,0.0916426678498586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,2,balanced,0.12533866365750632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,32,2,power_law_1.01,0.5749055862426757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,2,balanced,0.19200533628463745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,2,balanced,0.22494399547576904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,2,balanced,0.22920000553131104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,2,balanced,0.22779200474421182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,2,balanced,0.23123733202616373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,1,power_law_1.01,0.2927808046340942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,2,balanced,0.23348265886306763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,2,balanced,0.23317867517471313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,2,balanced,0.23665066560109457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,2,balanced,0.23885866006215414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,2,balanced,0.24551467100779215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,2,balanced,0.253493328889211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,2,balanced,0.25706666707992554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,2,balanced,0.2645546595255534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,1,power_law_1.01,0.3038144111633301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,2,balanced,0.28723732630411786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,2,balanced,0.311791996161143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,2,balanced,0.43779198328653973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,2,balanced,0.39191468556722003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,2,balanced,0.5458293358484904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,2,balanced,0.5819520155588785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,1,power_law_1.01,0.3160576105117798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,32,2,power_law_1.01,0.7460288047790528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,2,balanced,0.8350240389506022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,2,balanced,0.9901493390401205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,1,power_law_1.01,0.31877760887145995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,2,balanced,1.4236586888631184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,2,balanced,1.814341386159261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,1,power_law_1.01,0.32597761154174804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,2,balanced,2.1104960441589355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,1,power_law_1.01,0.35493760108947753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,2,balanced,3.3890612920125327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,32,2,power_law_1.01,0.927462387084961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,1,power_law_1.01,0.36250879764556887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,2,balanced,6.421552022298177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,1,power_law_1.01,0.38064639568328856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,1,power_law_1.01,0.4254144191741943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,32,2,power_law_1.01,1.3197567939758301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,1,power_law_1.01,0.4745791912078857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,1,power_law_1.01,0.5863679885864258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,1,power_law_1.01,0.7093567848205566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,32,2,power_law_1.01,1.6367488861083985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,1,power_law_1.01,0.9010623931884766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,1,power_law_1.01,1.027948760986328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,8,2,balanced,0.034789333740870156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,8,2,balanced,0.03580799947182337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,32,2,power_law_1.01,1.9929216384887696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,8,2,balanced,0.03947199881076813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,1,power_law_1.01,1.343513584136963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,8,2,balanced,0.03750933210055033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,8,2,balanced,0.03750933210055033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,8,2,balanced,0.03921599934498469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,8,2,balanced,0.03945599993069967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,8,2,balanced,0.03976000100374222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,8,2,balanced,0.03789333254098892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,8,2,balanced,0.03939733405907949
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,8,2,balanced,0.04027199993530909
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,8,2,balanced,0.039994666973749794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,8,2,balanced,0.041536000867684685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,8,2,balanced,0.043578664461771645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,8,2,balanced,0.04975466430187225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,8,2,balanced,0.05128000179926554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,8,2,balanced,0.04987200101216634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,8,2,balanced,0.06241600215435028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,8,2,balanced,0.0683840016523997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,8,2,balanced,0.0805866668621699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,8,2,balanced,0.09825600186983745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,8,2,balanced,0.12168533603350322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,8,2,balanced,0.14567466576894125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,8,2,balanced,0.1913813352584839
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,8,2,balanced,0.2257173260052999
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,1,power_law_1.01,1.7382080078125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,8,2,balanced,0.3174720009167989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,8,2,balanced,0.4129493236541748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,8,2,balanced,0.4994453191757202
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,8,2,balanced,0.7534666856129965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,8,2,balanced,1.44379727045695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,1,power_law_1.01,2.395955276489258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,32,2,power_law_1.01,3.1612415313720703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,1,power_law_1.01,3.0606399536132813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,1,power_law_1.01,3.7285694122314452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,8,power_law_1.01,0.0565887987613678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,32,2,power_law_1.01,6.098073577880859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,8,power_law_1.01,0.062463998794555664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,1,power_law_1.01,5.570240020751953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,8,power_law_1.01,0.07482240200042725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,8,power_law_1.01,0.060915201902389526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,8,power_law_1.01,0.06419839859008789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,8,power_law_1.01,0.07260159850120544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,1,power_law_1.01,10.823072052001953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,8,power_law_1.01,0.07626240253448487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,64,power_law_1.01,0.07873280048370361
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,8,power_law_1.01,0.08589439988136291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,64,power_law_1.01,0.07793279886245727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,8,power_law_1.01,0.08789759874343872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,64,power_law_1.01,0.07526400089263915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,8,power_law_1.01,0.09318400025367737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,64,power_law_1.01,0.06202239990234375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,8,power_law_1.01,0.09687039852142335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,64,power_law_1.01,0.06680319905281067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,8,power_law_1.01,0.09656959772109985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,64,power_law_1.01,0.06780800223350525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,16,power_law_1.2,0.08475520014762879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,8,power_law_1.01,0.0945088028907776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,64,power_law_1.01,0.06821759939193725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,8,power_law_1.01,0.10287359952926636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,64,power_law_1.01,0.07171199917793274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,16,power_law_1.2,0.13023359775543214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,8,power_law_1.01,0.10856319665908813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,64,power_law_1.01,0.07166720032691956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,8,power_law_1.01,0.11312639713287354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,64,power_law_1.01,0.08462719917297364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,64,power_law_1.01,0.0723136007785797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,16,power_law_1.2,0.10053119659423829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,8,power_law_1.01,0.12328319549560547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,64,power_law_1.01,0.07237759828567505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,64,power_law_1.01,0.0809216022491455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,8,power_law_1.01,0.13857280015945433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,64,power_law_1.01,0.0737280011177063
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,16,power_law_1.2,0.07587839961051941
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,64,power_law_1.01,0.07998080253601074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,8,power_law_1.01,0.15212160348892212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,64,power_law_1.01,0.07345920205116271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,64,power_law_1.01,0.06273279786109924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,8,power_law_1.01,0.18440959453582764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,16,power_law_1.2,0.08044800162315369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,64,power_law_1.01,0.07578240036964416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,64,power_law_1.01,0.06813439726829529
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,8,power_law_1.01,0.18599040508270265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,64,power_law_1.01,0.08370559811592101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,64,power_law_1.01,0.07549440264701843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,16,power_law_1.2,0.09152640104293823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,8,power_law_1.01,0.24558720588684083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,64,power_law_1.01,0.08147839903831482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,64,power_law_1.01,0.07639039754867553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,8,power_law_1.01,0.2973504066467285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,16,power_law_1.2,0.1075711965560913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,64,power_law_1.01,0.08569599986076355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,64,power_law_1.01,0.07996159791946411
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,8,power_law_1.01,0.3940799951553345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,64,power_law_1.01,0.09731199741363525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,64,power_law_1.01,0.07987200021743775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,16,power_law_1.2,0.11731200218200684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,8,power_law_1.01,0.5286528110504151
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,64,power_law_1.01,0.10729600191116333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,64,power_law_1.01,0.08162559866905213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,64,power_law_1.01,0.12598400115966796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,16,power_law_1.2,0.11538560390472412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,64,power_law_1.01,0.08080000281333924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,8,power_law_1.01,0.7202367782592773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,64,power_law_1.01,0.15389440059661866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,64,power_law_1.01,0.08320639729499817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,16,power_law_1.2,0.12208000421524048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,8,power_law_1.01,0.953593635559082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,64,power_law_1.01,0.19032959938049315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,64,power_law_1.01,0.08145279884338379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,16,power_law_1.2,0.12762880325317383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,64,power_law_1.01,0.08328319787979126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,64,power_law_1.01,0.2232192039489746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,8,power_law_1.01,1.1920127868652344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,64,power_law_1.01,0.09229440093040467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,64,power_law_1.01,0.3077375888824463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,16,power_law_1.2,0.13050880432128906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,64,power_law_1.01,0.09435520172119141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,8,power_law_1.01,1.7863744735717773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,64,power_law_1.01,0.37200639247894285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,16,power_law_1.2,0.13490560054779052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,64,power_law_1.01,0.09846400022506714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,64,power_law_1.01,0.5308671951293945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,64,power_law_1.01,0.10896639823913574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,16,power_law_1.2,0.13570560216903688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,64,power_law_1.01,0.12035200595855713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,8,power_law_1.01,3.4940223693847656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,64,power_law_1.01,0.6963583946228027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,16,power_law_1.2,0.1442304015159607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,64,power_law_1.01,0.1415552020072937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,64,power_law_1.01,0.8417023658752442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,64,power_law_1.01,0.17360639572143555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,16,power_law_1.2,0.1500864028930664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,64,power_law_1.01,0.22704000473022462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,16,power_law_1.2,0.15786880254745483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,64,power_law_1.01,1.2992768287658691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,64,power_law_1.01,0.26746881008148193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,16,power_law_1.2,0.17015680074691772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,64,power_law_1.01,0.36357119083404543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,16,power_law_1.2,0.1887488007545471
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,64,power_law_1.01,0.4662655830383301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,64,power_law_1.01,2.6690879821777345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,16,power_law_1.2,0.21766400337219238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,64,power_law_1.01,0.6687615871429443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,16,power_law_1.2,0.24805760383605957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,64,power_law_1.01,0.8656000137329102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,8,1,power_law_1.2,0.038873600959777835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,8,1,power_law_1.2,0.04759680032730103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,8,1,power_law_1.2,0.05125759840011597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,8,1,power_law_1.2,0.07512320280075073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,8,1,power_law_1.2,0.07707520127296448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,8,1,power_law_1.2,0.07734400033950806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,8,1,power_law_1.2,0.07827839851379395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,16,power_law_1.2,0.3282880067825317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,8,1,power_law_1.2,0.07913600206375122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,64,power_law_1.01,1.097708797454834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,8,1,power_law_1.2,0.08128640055656433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,8,1,power_law_1.2,0.08223360180854797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,8,1,power_law_1.2,0.0819263994693756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,8,1,power_law_1.2,0.08823040127754211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,8,1,power_law_1.2,0.08885120153427124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,8,1,power_law_1.2,0.09903360009193421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,8,1,power_law_1.2,0.10791679620742797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,8,1,power_law_1.2,0.11094399690628051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,8,1,power_law_1.2,0.11340160369873047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,16,power_law_1.2,0.35671679973602294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,8,1,power_law_1.2,0.13193600177764891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,8,1,power_law_1.2,0.15238399505615235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,8,1,power_law_1.2,0.1858944058418274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,64,power_law_1.01,1.644428825378418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,8,1,power_law_1.2,0.2273087978363037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,8,1,power_law_1.2,0.3022527933120728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,8,1,power_law_1.2,0.36056320667266845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,8,1,power_law_1.2,0.5081600189208985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,8,1,power_law_1.2,0.6406079769134522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,16,power_law_1.2,0.46517119407653806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,8,1,power_law_1.2,0.9183615684509278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,8,1,power_law_1.2,1.1797439575195312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,8,1,power_law_1.2,1.4489855766296387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,16,power_law_1.2,0.5808767795562744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,8,1,power_law_1.2,2.284979248046875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,64,power_law_1.01,3.2199871063232424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,16,balanced,0.031184000273545582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,16,balanced,0.031301334500312805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,16,balanced,0.03138133386770884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,16,balanced,0.03331200033426285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,16,balanced,0.03175999969244003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,16,balanced,0.033717334270477295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,16,balanced,0.03532266616821289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,16,balanced,0.03385599950949351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,16,balanced,0.035386666655540466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,16,balanced,0.0335359995563825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,16,balanced,0.035546667873859406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,16,balanced,0.036533333361148834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,16,balanced,0.03545066714286804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,16,balanced,0.03743999948104223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,16,balanced,0.040031999349594116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,16,balanced,0.04161066561937332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,16,balanced,0.04172799984614054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,16,balanced,0.04388799766699473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,16,balanced,0.04719999929269155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,16,balanced,0.05182399849096934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,16,balanced,0.0558240016301473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,16,balanced,0.07285333176453908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,16,balanced,0.08259200056393941
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,16,balanced,0.1093280017375946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,16,balanced,0.13608533143997192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,16,balanced,0.1801919937133789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,16,balanced,0.22099200884501138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,16,balanced,0.26525332530339557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,8,1,power_law_1.2,4.436896133422851
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,16,balanced,0.3930773337682088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,16,balanced,0.7496693134307861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,16,power_law_1.2,0.7970304012298584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,16,power_law_1.2,1.0312895774841309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,16,power_law_1.2,1.3561535835266114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,16,power_law_1.2,1.9430015563964844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,16,power_law_1.2,3.979596710205078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,64,power_law_1.01,0.03861120045185089
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,64,power_law_1.01,0.031564798951148984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,32,1,balanced,0.033546666304270424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,32,1,balanced,0.03549333413441976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,32,1,balanced,0.05351466437180837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,32,1,balanced,0.05342400074005127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,8,2,balanced,0.05601066847642263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,32,1,balanced,0.055642664432525635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,32,1,balanced,0.055173332492510475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,8,2,balanced,0.062037333846092224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,64,power_law_1.01,0.027321600914001466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,32,1,balanced,0.05657599866390228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,32,1,balanced,0.055999999245007835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,8,2,balanced,0.06435200075308482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,32,1,balanced,0.05492799977461497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,8,2,balanced,0.08412800232569377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,32,1,balanced,0.05650666852792104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,8,2,balanced,0.11569600303967793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,32,1,balanced,0.05795733133951823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,32,1,balanced,0.058277333776156105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,8,2,balanced,0.13396799564361572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,32,1,balanced,0.05880533158779144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,8,2,balanced,0.13500799735387167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,32,1,balanced,0.06515733400980632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,8,2,balanced,0.13829333583513895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,32,1,balanced,0.07276799778143565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,32,1,balanced,0.07482133309046428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,8,2,balanced,0.13847466309865317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,32,1,balanced,0.07852800190448761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,64,power_law_1.01,0.0281792014837265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,8,2,balanced,0.1386613349119822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,32,1,balanced,0.09055466453234355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,8,2,balanced,0.1377066671848297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,32,1,balanced,0.10309333602587382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,8,2,balanced,0.13853866855303446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,32,1,balanced,0.12981866796811423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,8,2,balanced,0.14036267002423605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,32,1,balanced,0.1366933286190033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,8,2,balanced,0.1458239952723185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,32,1,balanced,0.18249066670735678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,8,2,balanced,0.15239466230074564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,32,1,balanced,0.21838400761286417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,8,2,balanced,0.1551520029703776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,32,1,balanced,0.29985066254933673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,8,2,balanced,0.16339199741681418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,64,power_law_1.01,0.028505599498748778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,32,1,balanced,0.38200000921885174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,8,2,balanced,0.18522133429845175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,8,2,balanced,0.20306666692097983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,32,1,balanced,0.5445493459701538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,8,2,balanced,0.2458453377087911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,8,2,balanced,0.2740959922472636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,32,1,balanced,0.684175968170166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,8,2,balanced,0.3943733374277751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,64,power_law_1.01,0.02933120131492615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,32,1,balanced,0.8457439740498861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,8,2,balanced,0.466261347134908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,8,2,balanced,0.6858346462249756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,32,1,balanced,1.3109493255615234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,8,2,balanced,0.8303146362304688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,64,power_law_1.01,0.029651200771331786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,8,2,balanced,1.2165119647979736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,32,1,balanced,2.5460373560587564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,8,2,balanced,1.5540107091267903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,64,power_law_1.01,0.029356798529624938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,8,2,balanced,1.8508960405985515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,64,power_law_1.01,0.02887679934501648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,8,2,balanced,2.8971999486287436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,64,power_law_1.01,0.029817599058151244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,64,power_law_1.01,0.029606398940086365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,8,2,balanced,5.65773328145345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,64,power_law_1.01,0.029926401376724244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,64,power_law_1.01,0.03133440017700195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,64,power_law_1.01,0.03150080144405365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,64,power_law_1.01,0.03493120074272156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,64,power_law_1.01,0.03656960129737854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,64,power_law_1.01,0.0371071994304657
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,64,power_law_1.01,0.03919360041618347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,64,power_law_1.01,0.04201599955558777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,64,power_law_1.01,0.04863359928131104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,64,power_law_1.01,0.04831359982490539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,64,power_law_1.01,0.06069759726524353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,64,power_law_1.01,0.07194240093231201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,64,power_law_1.01,0.08743680119514466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,16,power_law_1.2,0.05727360248565674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,64,power_law_1.01,0.09894400238990783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,64,power_law_1.01,0.13043839931488038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,16,power_law_1.2,0.054028797149658206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,64,power_law_1.01,0.16974079608917236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,16,power_law_1.2,0.04568319916725159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,2,64,power_law_1.01,0.174508798122406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,16,power_law_1.2,0.04778240025043488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,4,power_law_1.2,0.048793599009513855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,2,64,power_law_1.01,0.30178558826446533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,16,power_law_1.2,0.05136640071868896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,4,power_law_1.2,0.07006080150604248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,1,balanced,0.08433066805203755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,1,balanced,0.12171199917793274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,1,balanced,0.13853333393732706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,1,balanced,0.19978666305541992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,1,balanced,0.3276639978090922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,1,balanced,0.5700000127156576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,1,balanced,0.5749226808547974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,1,balanced,0.5791626771291097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,2,64,power_law_1.01,0.5451327800750733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,1,balanced,0.5838666756947836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,1,balanced,0.5853066841761271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,1,balanced,0.5888319810231527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,16,power_law_1.2,0.054079997539520266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,4,power_law_1.2,0.06419199705123901
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,1,balanced,0.5988959868748983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,1,balanced,0.6040853261947632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,1,balanced,0.6172159910202026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,1,balanced,0.6404159863789877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,1,balanced,0.641594648361206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,1,balanced,0.6625653505325317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,1,balanced,0.7108960151672363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,1,balanced,0.7566666603088379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,4,power_law_1.2,0.06398079991340637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,1,balanced,0.860309362411499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,16,power_law_1.2,0.05683839917182922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,1,balanced,0.9068693319956461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,1,balanced,1.069925308227539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,1,balanced,1.2430559794108074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,1,balanced,1.6648319562276204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,4,power_law_1.2,0.08843520283699036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,16,power_law_1.2,0.05783680081367493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,1,balanced,1.9971307118733723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,1,balanced,2.896165211995443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,4,power_law_1.2,0.10033919811248779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,16,power_law_1.2,0.05850239992141724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,1,balanced,3.5851786931355796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,1,balanced,4.61737600962321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,4,power_law_1.2,0.11509120464324951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,16,power_law_1.2,0.05793920159339905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,1,balanced,6.765946706136067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,4,power_law_1.2,0.14831360578536987
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,16,power_law_1.2,0.05934079885482788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,4,power_law_1.2,0.15202560424804687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,16,power_law_1.2,0.05980799794197082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,1,balanced,13.062442779541016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,4,power_law_1.2,0.15985920429229736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,16,power_law_1.2,0.06380800008773804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,4,power_law_1.2,0.15945600271224974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,16,power_law_1.2,0.06442239880561829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,2,2,balanced,0.08222400148709615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,2,2,balanced,0.12147200107574463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,2,2,balanced,0.20936532815297446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,2,2,balanced,0.21589332818984985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,4,power_law_1.2,0.1706496000289917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,2,2,balanced,0.21251199642817178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,16,power_law_1.2,0.07273600101470948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,2,2,balanced,0.2087306578954061
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,2,2,balanced,0.20907199382781982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,2,2,balanced,0.20887466271718344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,2,2,balanced,0.20937599738438925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,2,2,balanced,0.2116159995396932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,2,2,balanced,0.21863466501235962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,2,2,balanced,0.21076265970865884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,2,2,balanced,0.2133386731147766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,2,2,balanced,0.22728532552719116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,2,2,balanced,0.24799466133117676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,2,2,balanced,0.24943466981252035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,2,2,balanced,0.24672534068425497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,2,2,balanced,0.3001493414243062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,2,2,balanced,0.29588266213734943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,2,2,balanced,0.3691893418629964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,2,2,balanced,0.3959413369496663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,16,power_law_1.2,0.07383040189743043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,4,power_law_1.2,0.1736448049545288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,2,2,balanced,0.5356959899266561
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,2,2,balanced,0.6917706330617269
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,2,2,balanced,0.9652907053629557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,2,2,balanced,1.2530986467997234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,2,2,balanced,1.8508639335632324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,16,power_law_1.2,0.07884160280227662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,4,power_law_1.2,0.18443520069122316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,2,2,balanced,2.4443413416544595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,2,2,balanced,3.0718132654825845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,16,power_law_1.2,0.08949120044708252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,4,power_law_1.2,0.19420160055160524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,2,2,balanced,4.908224105834961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,4,power_law_1.2,0.20090880393981933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,16,power_law_1.2,0.09685760140419006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,2,2,balanced,9.573114395141602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,4,power_law_1.2,0.21477758884429932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,16,power_law_1.2,0.11644159555435181
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,4,power_law_1.2,0.23457279205322265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,16,power_law_1.2,0.13691519498825072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,2,power_law_1.2,0.03209599852561951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,4,power_law_1.2,0.24893438816070557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,2,power_law_1.2,0.031974399089813234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,16,power_law_1.2,0.17632639408111572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,2,power_law_1.2,0.03672960102558136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,4,power_law_1.2,0.3015552043914795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,2,power_law_1.2,0.040268799662590025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,16,power_law_1.2,0.20609281063079835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,2,power_law_1.2,0.04339199960231781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,2,power_law_1.2,0.06257280111312866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,4,power_law_1.2,0.3342463970184326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,2,power_law_1.2,0.06672639846801758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,16,power_law_1.2,0.2749311923980713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,2,power_law_1.2,0.06769919991493226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,4,power_law_1.2,0.4059455871582031
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,2,power_law_1.2,0.07071359753608704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,2,power_law_1.2,0.07274879813194275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,16,power_law_1.2,0.3825088024139404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,2,power_law_1.2,0.07453439831733703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,4,power_law_1.2,0.5025472164154052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,2,power_law_1.2,0.07741439938545228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,2,power_law_1.2,0.08163840174674988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,16,power_law_1.2,0.543552017211914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,2,power_law_1.2,0.08483200073242188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,4,power_law_1.2,0.6617663860321045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,2,power_law_1.2,0.09155840277671815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,2,power_law_1.2,0.09477760195732117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,1,4,balanced,0.06446399788061778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,1,4,balanced,0.07684266567230225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,16,power_law_1.2,0.6523392200469971
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,1,4,balanced,0.09110400080680847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,1,4,balanced,0.0897653301556905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,1,4,balanced,0.0888853371143341
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,2,power_law_1.2,0.10183680057525635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,1,4,balanced,0.08965866764386494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,4,power_law_1.2,0.8018431663513184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,1,4,balanced,0.09014399846394856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,1,4,balanced,0.09025599559148152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,1,4,balanced,0.0918933351834615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,1,4,balanced,0.09298666318257649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,1,4,balanced,0.09428266684214275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,1,4,balanced,0.09381332993507385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,1,4,balanced,0.09444800019264221
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,1,4,balanced,0.10179199775060017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,1,4,balanced,0.10666666428248088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,2,power_law_1.2,0.11841919422149658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,1,4,balanced,0.10750933488210042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,1,4,balanced,0.11417599519093831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,1,4,balanced,0.13369066516558328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,1,4,balanced,0.13988266388575235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,1,4,balanced,0.17870932817459106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,1,4,balanced,0.19370667139689127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,1,4,balanced,0.26341867446899414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,1,4,balanced,0.3123253385225932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,2,power_law_1.2,0.14252159595489503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,1,4,balanced,0.44219199816385907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,1,4,balanced,0.5734719832738241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,1,4,balanced,0.8232159614562988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,1,4,balanced,1.05402668317159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,2,power_law_1.2,0.17158399820327758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,16,power_law_1.2,0.8561856269836425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,4,power_law_1.2,1.121388816833496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,1,4,balanced,1.3064160346984863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,2,power_law_1.2,0.20892159938812255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,1,4,balanced,2.0499626795450845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,2,power_law_1.2,0.2670975923538208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,1,4,balanced,4.083461443583171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,2,power_law_1.2,0.36281599998474123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,4,power_law_1.2,1.3525119781494142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,16,power_law_1.2,1.2450495719909669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,2,power_law_1.2,0.492460823059082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,2,power_law_1.2,0.635916805267334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,4,power_law_1.2,1.642745590209961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,2,power_law_1.2,0.9106880187988281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,2,power_law_1.2,1.268179225921631
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,16,power_law_1.2,2.803641510009766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,4,power_law_1.2,2.495084762573242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,2,power_law_1.2,1.628531265258789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,2,power_law_1.2,2.271763229370117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,4,power_law_1.2,4.863251113891602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,2,power_law_1.2,5.006208038330078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,4,1,power_law_1.01,0.046515199542045596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,4,1,power_law_1.01,0.070169597864151
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,4,1,power_law_1.01,0.06342399716377259
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,4,1,power_law_1.01,0.08776959776878357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,4,1,power_law_1.01,0.08726400136947632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,4,1,power_law_1.01,0.09015679955482483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,4,1,power_law_1.01,0.09022719860076904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,4,1,power_law_1.01,0.09073920249938965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,4,1,power_law_1.01,0.08980479836463928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,4,1,power_law_1.01,0.09114239811897278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,4,1,power_law_1.01,0.0944703996181488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,4,1,power_law_1.01,0.09607040286064147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,4,1,power_law_1.01,0.09969279766082764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,4,1,power_law_1.01,0.10765440464019775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,4,1,power_law_1.01,0.11841280460357666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,4,1,power_law_1.01,0.12043520212173461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,4,1,power_law_1.01,0.12885119915008544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,4,1,power_law_1.01,0.15125759840011596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,4,1,power_law_1.01,0.1752768039703369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,4,1,power_law_1.01,0.21823360919952392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,4,1,power_law_1.01,0.24146559238433837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,4,1,power_law_1.01,0.3170752048492432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,4,1,power_law_1.01,0.38087680339813235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,1,balanced,0.052000001072883606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,1,balanced,0.05151999990145365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,4,1,power_law_1.01,0.5249023914337159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,1,balanced,0.06150400141874949
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,1,balanced,0.0822026679913203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,1,balanced,0.11982400218645732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,1,balanced,0.12026133139928182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,4,1,power_law_1.01,0.6587071895599366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,1,balanced,0.12412266929944356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,1,balanced,0.12418666481971741
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,1,balanced,0.12498133381207784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,1,balanced,0.12757333119710287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,1,balanced,0.12898666659990946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,4,1,power_law_1.01,0.9368703842163086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,1,balanced,0.133050670226415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,1,balanced,0.13366933663686117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,1,balanced,0.13943466544151306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,1,balanced,0.1453439990679423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,1,balanced,0.1495146652062734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,1,balanced,0.1573973298072815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,4,1,power_law_1.01,1.2153663635253906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,1,balanced,0.17945067087809244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,1,balanced,0.20046399037043253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,1,balanced,0.2424373428026835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,1,balanced,0.28429333368937176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,1,balanced,0.39661868413289386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,4,1,power_law_1.01,1.4932607650756835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,1,balanced,0.48285333315531415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,1,balanced,0.6933759848276774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,1,balanced,0.8793493111928304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,1,balanced,1.2820533116658528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,4,1,power_law_1.01,2.330668830871582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,1,balanced,1.6818666458129883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,1,balanced,2.0802879333496094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,1,balanced,3.3011999130249023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,4,1,power_law_1.01,4.55670394897461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,1,balanced,6.3886292775472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,256,power_law_1.2,0.0648576021194458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,1,power_law_1.01,0.03278079926967621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,1,power_law_1.01,0.035411199927330016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,1,power_law_1.01,0.040140798687934874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,256,power_law_1.2,0.06584960222244263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,1,power_law_1.01,0.04650880098342895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,1,power_law_1.01,0.05679360032081604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,1,power_law_1.01,0.07276160120964051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,256,power_law_1.2,0.06027519702911377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,1,power_law_1.01,0.0773311972618103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,1,power_law_1.01,0.0782975971698761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,1,power_law_1.01,0.08423039913177491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,256,power_law_1.2,0.06384639739990235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,1,power_law_1.01,0.08718079924583436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,1,power_law_1.01,0.09144960045814514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,1,power_law_1.01,0.09605119824409485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,256,power_law_1.2,0.06379520297050476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,1,power_law_1.01,0.10266239643096924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,1,power_law_1.01,0.10229120254516602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,1,power_law_1.01,0.11080960035324097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,256,power_law_1.2,0.06388480067253113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,1,power_law_1.01,0.11635199785232545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,1,2,power_law_1.01,0.07022719979286193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,1,power_law_1.01,0.1276352047920227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,1,2,power_law_1.01,0.08295680284500122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,16,balanced,0.043925335009892784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,16,balanced,0.03292799989382426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,1,2,power_law_1.01,0.08294399976730346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,16,balanced,0.03143466760714849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,1,power_law_1.01,0.15728000402450562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,16,balanced,0.03302400062481562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,1,2,power_law_1.01,0.10146559476852417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,16,balanced,0.033285332222779594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,16,balanced,0.036981334288915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,1,2,power_law_1.01,0.10133119821548461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,16,balanced,0.03656533360481262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,256,power_law_1.2,0.06492159962654113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,16,balanced,0.03669333209594091
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,1,2,power_law_1.01,0.10264960527420045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,16,balanced,0.03551999976237615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,1,power_law_1.01,0.18817280530929564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,16,balanced,0.036618667344252266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,16,balanced,0.03730666637420654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,1,2,power_law_1.01,0.10348800420761109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,16,balanced,0.036490666369597115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,16,balanced,0.03681600093841553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,1,2,power_law_1.01,0.10347520112991333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,16,balanced,0.03855466594298681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,16,balanced,0.043162668744723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,16,balanced,0.0425600012143453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,1,2,power_law_1.01,0.10776319503784179
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,1,power_law_1.01,0.24142079353332518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,16,balanced,0.04264000058174133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,16,balanced,0.04863999783992767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,1,2,power_law_1.01,0.10598399639129638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,16,balanced,0.045109331607818604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,16,balanced,0.0525546669960022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,1,2,power_law_1.01,0.11140480041503906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,16,balanced,0.05309333403905233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,16,balanced,0.07726400097211202
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,1,2,power_law_1.01,0.11459200382232666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,16,balanced,0.07991999884446462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,1,power_law_1.01,0.2917759895324707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,256,power_law_1.2,0.06483200192451477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,16,balanced,0.10813333590825398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,1,2,power_law_1.01,0.11827839612960815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,16,balanced,0.1265120009581248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,16,balanced,0.17108800013860068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,1,2,power_law_1.01,0.12855679988861085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,16,balanced,0.2056586742401123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,1,2,power_law_1.01,0.1389248013496399
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,2,16,balanced,0.24657066663106283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,1,power_law_1.01,0.4008959770202637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,2,16,balanced,0.35042134920756024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,1,2,power_law_1.01,0.14841599464416505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,2,16,balanced,0.6533759832382202
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,1,2,power_law_1.01,0.15994880199432374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,1,2,power_law_1.01,0.18290560245513915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,1,2,power_law_1.01,0.2160128116607666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,1,power_law_1.01,0.49993600845336916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,256,power_law_1.2,0.06496000289916992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,1,2,power_law_1.01,0.2753024101257324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,1,2,power_law_1.01,0.3117824077606201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,1,2,power_law_1.01,0.49441280364990237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,1,power_law_1.01,0.7116352081298828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,1,2,power_law_1.01,0.5222655773162842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,1,2,power_law_1.01,0.8227264404296875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,1,power_law_1.01,0.9194496154785157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,256,power_law_1.2,0.0656000018119812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,1,2,power_law_1.01,1.0697983741760253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,1,2,power_law_1.01,1.6326400756835937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,1,power_law_1.01,1.3374527931213378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,1,2,power_law_1.01,2.2289920806884767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,256,power_law_1.2,0.06687359809875489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,1,2,power_law_1.01,2.9383871078491213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,1,power_law_1.01,1.7449663162231446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,256,power_law_1.2,0.06723840236663818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,1,2,power_law_1.01,4.060851287841797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,1,power_law_1.01,2.1761024475097654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,256,power_law_1.2,0.06734079718589783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,1,2,power_law_1.01,7.2628227233886715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,1,power_law_1.01,3.402675247192383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,256,power_law_1.2,0.06829439997673034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,256,power_law_1.2,0.07216640114784241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,256,power_law_1.2,0.07342720031738281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,4,1,power_law_1.01,0.045875200629234315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,1,power_law_1.01,6.686534118652344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,4,1,power_law_1.01,0.053855997323989865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,4,1,power_law_1.01,0.05960959792137146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,4,1,power_law_1.01,0.08165760040283203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,256,power_law_1.2,0.08047999739646912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,4,1,power_law_1.01,0.07968000173568726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,4,1,power_law_1.01,0.08472959995269776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,4,1,power_law_1.01,0.08588160276412964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,4,1,power_law_1.01,0.08611199855804444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,4,1,power_law_1.01,0.08724480271339416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,4,1,power_law_1.01,0.09056000113487243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,256,power_law_1.2,0.09403520226478576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,4,1,power_law_1.01,0.08924800157546997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,4,1,power_law_1.01,0.0937279999256134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,4,1,power_law_1.01,0.09386879801750184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,4,1,power_law_1.01,0.0987712025642395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,4,1,power_law_1.01,0.10766079425811767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,256,power_law_1.2,0.11418880224227905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,4,1,power_law_1.01,0.11148159503936768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,4,1,power_law_1.01,0.11838079690933227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,4,1,power_law_1.01,0.13716479539871215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,4,1,power_law_1.01,0.15294079780578612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,4,1,power_law_1.01,0.1880511999130249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,256,power_law_1.2,0.14455039501190187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,4,1,power_law_1.01,0.22262399196624755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,4,1,power_law_1.01,0.28497920036315916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,4,1,power_law_1.01,0.3513535976409912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,4,1,power_law_1.01,0.48888959884643557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,8,power_law_1.2,0.03976959884166718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,4,1,power_law_1.01,0.629030418395996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,256,power_law_1.2,0.17099519968032836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,8,power_law_1.2,0.03501439988613129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,4,1,power_law_1.01,0.8856320381164551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,8,power_law_1.2,0.03511680066585541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,4,1,power_law_1.01,1.1368831634521483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,8,power_law_1.2,0.03798399865627289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,4,1,power_law_1.01,1.399500846862793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,256,power_law_1.2,0.22470400333404542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,8,power_law_1.2,0.04009599983692169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,8,power_law_1.2,0.04692479968070984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,4,1,power_law_1.01,2.1595455169677735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,8,power_law_1.2,0.04785920083522797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,8,power_law_1.2,0.048102399706840514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,2,balanced,0.049738665421803795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,2,balanced,0.052111998200416565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,2,balanced,0.05843733251094818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,2,balanced,0.08122133215268452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,256,power_law_1.2,0.25061759948730467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,2,balanced,0.11264533797899882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,2,balanced,0.16962132851282755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,2,balanced,0.22797866662343344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,2,balanced,0.22062400976816812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,8,power_law_1.2,0.04933759868144989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,2,balanced,0.2227039933204651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,2,balanced,0.22529600063959757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,2,balanced,0.2266133427619934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,4,1,power_law_1.01,4.215935897827149
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,2,balanced,0.2300693392753601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,2,balanced,0.23227733373641968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,2,balanced,0.23963199059168497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,8,power_law_1.2,0.050419199466705325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,2,balanced,0.2507306734720866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,2,balanced,0.25429866711298627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,2,balanced,0.26916799942652386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,2,balanced,0.30169065793355304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,2,balanced,0.33907198905944824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,2,balanced,0.3861440022786458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,8,power_law_1.2,0.05017600059509277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,2,balanced,0.42870934804280597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,2,balanced,0.5259626706441244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,2,balanced,0.6386613448460897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,2,balanced,0.8297279675801595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,8,power_law_1.2,0.052249598503112796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,2,balanced,1.0889386336008708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,256,power_law_1.2,0.3597503900527954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,2,balanced,1.469098726908366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,8,power_law_1.2,0.05422080159187317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,2,balanced,1.9615947405497234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,8,power_law_1.2,0.05912960171699524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,2,balanced,2.460906664530436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,8,power_law_1.2,0.06530560255050659
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,2,balanced,3.677114804585775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,8,power_law_1.2,0.06591359972953796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,256,power_law_1.2,0.45525121688842773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,8,power_law_1.2,0.07145599722862243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,8,power_law_1.2,0.08444160223007202
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,2,balanced,7.092981338500977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,8,power_law_1.2,0.10139520168304443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,8,power_law_1.2,0.1157696008682251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,8,power_law_1.2,0.13159040212631226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,256,power_law_1.2,0.6540671825408936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,8,power_law_1.2,0.17120640277862548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,8,power_law_1.2,0.23248639106750488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,8,power_law_1.2,0.30497920513153076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,8,power_law_1.2,0.37242879867553713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,256,power_law_1.2,0.7903039932250977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,8,power_law_1.2,0.5283391952514649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,8,power_law_1.2,0.5880576133728027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,4,2,power_law_1.2,0.056883198022842404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,4,2,power_law_1.2,0.07888000011444092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,8,power_law_1.2,1.0344575881958007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,4,2,power_law_1.2,0.060889601707458496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,1,256,power_law_1.2,1.0477503776550292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,4,2,power_law_1.2,0.07574399709701538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,4,2,power_law_1.2,0.08208640217781067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,4,2,power_law_1.2,0.08468480110168457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,4,2,power_law_1.2,0.0843455970287323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,8,power_law_1.2,1.5255999565124512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,4,2,power_law_1.2,0.09015679955482483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,4,2,power_law_1.2,0.09089919924736023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,4,2,power_law_1.2,0.09011840224266052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,4,2,power_law_1.2,0.08902400135993957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,4,2,power_law_1.2,0.09163519740104675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,4,2,power_law_1.2,0.09496960043907166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,4,2,power_law_1.2,0.10442880392074586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,32,power_law_1.2,0.03155199885368347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,4,2,power_law_1.2,0.1156991958618164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,8,power_law_1.2,2.9256704330444334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,4,2,power_law_1.2,0.11759999990463257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,1,256,power_law_1.2,1.7611711502075196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,32,power_law_1.2,0.02815360128879547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,4,2,power_law_1.2,0.11874560117721558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,4,2,power_law_1.2,0.14880640506744386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,4,2,power_law_1.2,0.1634559988975525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,32,power_law_1.2,0.029311999678611755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,4,2,power_law_1.2,0.18343679904937743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,4,2,power_law_1.2,0.23496320247650146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,32,power_law_1.2,0.029369598627090453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,4,2,power_law_1.2,0.3163455963134766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,32,power_law_1.2,0.029651200771331786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,4,2,power_law_1.2,0.3529088020324707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,32,power_law_1.2,0.030003198981285097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,4,2,power_law_1.2,0.533951997756958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,32,power_law_1.2,0.030707201361656188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,4,2,power_law_1.2,0.7032447814941406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,32,power_law_1.2,0.030976000428199767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,4,2,power_law_1.2,1.046617603302002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,32,power_law_1.2,0.031014400720596313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,4,2,power_law_1.2,1.302233600616455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,32,power_law_1.2,0.031539198756217954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,4,2,power_law_1.2,1.5928319931030273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,32,power_law_1.2,0.03230080008506775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,32,power_law_1.2,0.033062401413917544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,4,2,power_law_1.2,2.5518592834472655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,32,power_law_1.2,0.0336896002292633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,32,power_law_1.2,0.035071998834609985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,1,256,power_law_1.2,3.7998977661132813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,32,power_law_1.2,0.03963519930839539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,4,2,power_law_1.2,5.351180648803711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,32,power_law_1.2,0.03914240002632141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,32,power_law_1.2,0.040428799390792844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,32,power_law_1.2,0.04466559886932373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,32,power_law_1.2,0.046598398685455324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,32,power_law_1.2,0.05591679811477661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,32,power_law_1.2,0.06111360192298889
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,32,power_law_1.2,0.08186240196228027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,32,power_law_1.2,0.09029759764671326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,32,power_law_1.2,0.07942399978637696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,32,power_law_1.2,0.1490623950958252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,32,power_law_1.2,0.06825600266456604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,32,power_law_1.2,0.1699455976486206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,32,power_law_1.2,0.04819200038909912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,32,power_law_1.2,0.28344318866729734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,32,power_law_1.2,0.05066879987716675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,32,power_law_1.2,0.3184000015258789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,32,power_law_1.2,0.052076798677444455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,32,power_law_1.2,0.5372352123260498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,32,power_law_1.2,0.06069120168685913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,32,power_law_1.2,0.06474239826202392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,32,power_law_1.2,0.7353216171264648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,32,power_law_1.2,0.06704000234603882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,64,balanced,0.036837334434191384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,64,balanced,0.03595199932654699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,64,balanced,0.03404266635576884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,64,balanced,0.03390933324893316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,64,balanced,0.03606399893760681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,32,power_law_1.2,0.06914560198783874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,64,balanced,0.03608000030120214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,64,balanced,0.03581333408753077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,64,balanced,0.03629866739114126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,64,balanced,0.035760000348091125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,64,balanced,0.03590933233499527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,64,balanced,0.03562666724125544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,64,balanced,0.035760000348091125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,64,balanced,0.037674665451049805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,64,balanced,0.03755199909210205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,64,balanced,0.03961600114901861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,64,balanced,0.03958933303753535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,64,balanced,0.04182399809360504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,64,balanced,0.04155733436346054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,64,balanced,0.04152533411979675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,64,balanced,0.04424533247947693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,64,balanced,0.045509333411852516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,64,balanced,0.047770669062932335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,32,power_law_1.2,0.06930559873580933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,64,balanced,0.05202666421731313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,64,balanced,0.06211733321348826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,64,balanced,0.0701386680205663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,64,balanced,0.09087999661763509
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,64,balanced,0.10797866185506184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,64,balanced,0.12692266702651978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,64,balanced,0.17588800191879272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,64,balanced,0.2946293354034424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,32,2,power_law_1.2,0.04830079972743988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,32,power_law_1.2,2.15665283203125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,32,power_law_1.2,0.06900479793548583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,32,power_law_1.2,0.0726207971572876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,32,2,power_law_1.2,0.04799999892711639
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,32,power_law_1.2,0.07283200025558471
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,32,2,power_law_1.2,0.053420799970626834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,32,power_law_1.2,0.07763199806213379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,32,power_law_1.2,0.08302720189094544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,32,2,power_law_1.2,0.05982720255851746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,32,power_law_1.2,0.08679680228233337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,32,2,power_law_1.2,0.06574079990386963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,32,power_law_1.2,0.09052799940109253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,32,2,power_law_1.2,0.09093760251998902
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,32,power_law_1.2,0.0971455991268158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,32,2,power_law_1.2,0.09616640210151672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,32,power_law_1.2,0.10671360492706299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,32,2,power_law_1.2,0.09940479993820191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,32,power_law_1.2,0.1323904037475586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,32,2,power_law_1.2,0.10103679895401001
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,32,power_law_1.2,0.14006400108337402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,32,2,power_law_1.2,0.10088959932327271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,32,power_law_1.2,0.18137600421905517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,32,2,power_law_1.2,0.10476160049438477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,32,power_law_1.2,0.23790080547332765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,32,2,power_law_1.2,0.10766079425811767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,32,balanced,0.04808000226815542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,32,balanced,0.04780800143877665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,32,power_law_1.2,0.2747391939163208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,32,balanced,0.046682665745417275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,32,balanced,0.05410666763782501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,16,1,balanced,0.031888000667095184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,32,balanced,0.07580799857775371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,16,1,balanced,0.035962666074434914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,32,balanced,0.08258666594823201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,16,1,balanced,0.05590933561325073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,32,balanced,0.08468799789746602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,16,1,balanced,0.05778133372465769
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,32,balanced,0.08430400490760803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,16,1,balanced,0.05686399837334951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,32,balanced,0.08498133222262065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,16,1,balanced,0.05725333094596863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,32,balanced,0.0846666693687439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,16,1,balanced,0.05937600135803223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,32,2,power_law_1.2,0.11039999723434449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,16,1,balanced,0.05937600135803223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,32,balanced,0.08683199683825175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,16,1,balanced,0.06047466893990835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,32,balanced,0.08757332960764568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,16,1,balanced,0.059008002281188965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,32,balanced,0.08644266923268636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,16,1,balanced,0.06206933160622915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,32,balanced,0.0911253293355306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,16,1,balanced,0.0666186660528183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,32,balanced,0.0962666670481364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,16,1,balanced,0.06764799853165944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,16,1,balanced,0.07045866549015045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,32,balanced,0.09749333063761394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,16,1,balanced,0.07810133198897044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,32,balanced,0.10032533605893452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,16,1,balanced,0.08071466783682506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,32,balanced,0.10714667042096455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,16,1,balanced,0.08849599957466125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,32,balanced,0.11188266674677531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,16,1,balanced,0.10234133402506511
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,32,balanced,0.13775466879208884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,16,1,balanced,0.11169067025184631
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,16,1,balanced,0.1446399986743927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,32,balanced,0.1400373379389445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,32,power_law_1.2,0.3969727993011475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,16,1,balanced,0.1702400048573812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,32,balanced,0.18888533115386963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,16,1,balanced,0.22590400775273642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,32,balanced,0.20107199748357138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,16,1,balanced,0.28082666794459027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,32,balanced,0.27129600445429486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,32,2,power_law_1.2,0.11561599969863892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,16,1,balanced,0.3983413378397624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,32,balanced,0.31565866867701214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,16,1,balanced,0.5012480020523071
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,32,balanced,0.43723734219868976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,32,balanced,0.5575360059738159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,16,1,balanced,0.7316160202026367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,32,balanced,0.6563946803410848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,16,1,balanced,0.9493546485900879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,32,balanced,1.0053013165791829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,16,1,balanced,1.1656373341878254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,32,power_law_1.2,0.5872000217437744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,32,2,power_law_1.2,0.12501120567321777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,16,1,balanced,1.8119573593139648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,32,balanced,1.930448055267334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,16,1,balanced,3.5666774113972983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,32,2,power_law_1.2,0.13033599853515626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,32,power_law_1.2,0.613043212890625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,32,2,power_law_1.2,0.13645440340042114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,32,2,power_law_1.2,0.16239360570907593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,32,power_law_1.2,0.7783999919891358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,32,2,power_law_1.2,0.19783040285110473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,32,2,power_law_1.2,0.23197441101074218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,32,power_law_1.2,1.2543680191040039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,32,2,power_law_1.2,0.30353920459747313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,32,2,power_law_1.2,0.39305601119995115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,32,2,power_law_1.2,0.461407995223999
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,32,power_law_1.2,2.4626943588256838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,32,2,power_law_1.2,0.6257215976715088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,32,2,power_law_1.2,0.8690431594848633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,32,2,power_law_1.2,1.1175295829772949
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,1,balanced,0.05600533386071523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,32,2,power_law_1.2,1.531436824798584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,1,balanced,0.0650079995393753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,1,balanced,0.07969066500663757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,1,balanced,0.11185066898663838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,1,balanced,0.16843199729919434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,1,balanced,0.2844533324241638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,1,balanced,0.28748265902201336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,1,balanced,0.2934933304786682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,1,balanced,0.2953546643257141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,1,balanced,0.2971946597099304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,1,balanced,0.30433066685994464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,1,balanced,0.31194667021433514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,1,balanced,0.3147733410199483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,1,balanced,0.3179466724395752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,1,balanced,0.3245226740837097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,1,balanced,0.3344106674194336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,1,balanced,0.3459999958674113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,1,balanced,0.3890293439229329
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,1,balanced,0.42074668407440186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,1,balanced,0.4935679833094279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,32,2,power_law_1.2,1.8181568145751954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,1,balanced,0.5572426716486613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,1,balanced,0.7543040116628011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,8,power_law_1.2,0.06993920207023621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,1,balanced,0.8472800254821777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,1,balanced,1.2203413645426433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,1,balanced,1.4716159502665203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,1,balanced,2.130016009012858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,1,balanced,2.7466293970743814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,8,power_law_1.2,0.06021760106086731
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,32,2,power_law_1.2,2.603091239929199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,1,balanced,3.3352158864339194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,1,balanced,5.109925270080566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,8,power_law_1.2,0.056576001644134524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,1,balanced,9.802997589111328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,4,8,power_law_1.2,0.036320000886917114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,8,power_law_1.2,0.06420480012893677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,4,8,power_law_1.2,0.036550399661064145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,4,8,power_law_1.2,0.03404799997806549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,32,2,power_law_1.2,5.608985519409179
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,4,8,power_law_1.2,0.03470079898834229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,4,8,power_law_1.2,0.03528960049152374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,4,8,power_law_1.2,0.035795199871063235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,4,8,power_law_1.2,0.03596799969673157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,4,8,power_law_1.2,0.036883199214935304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,4,8,power_law_1.2,0.037376001477241516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,4,8,power_law_1.2,0.037945601344108584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,8,power_law_1.2,0.07073919773101807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,4,8,power_law_1.2,0.037964800000190736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,4,8,power_law_1.2,0.039743998646736146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,4,8,power_law_1.2,0.041920000314712526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,4,8,power_law_1.2,0.043424001336097716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,4,8,power_law_1.2,0.04906240105628967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,4,8,power_law_1.2,0.05305600166320801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,4,8,power_law_1.2,0.055193597078323366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,4,8,power_law_1.2,0.06469759941101075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,8,power_law_1.2,0.09016960263252258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,4,8,power_law_1.2,0.0712768018245697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,4,8,power_law_1.2,0.09048960208892823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,4,8,power_law_1.2,0.09461119771003723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,4,8,power_law_1.2,0.13507200479507447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,4,8,power_law_1.2,0.1600383996963501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,4,8,power_law_1.2,0.22943999767303466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,4,8,power_law_1.2,0.28142080307006834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,8,power_law_1.2,0.09660159945487976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,4,8,power_law_1.2,0.39375360012054444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,4,8,power_law_1.2,0.4939583778381348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,4,8,power_law_1.2,0.5688896179199219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,8,power_law_1.2,0.10084480047225952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,4,8,power_law_1.2,0.8981311798095704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,4,8,power_law_1.2,1.8209344863891601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,8,power_law_1.2,0.10188800096511841
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,8,power_law_1.2,0.10576640367507935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,8,power_law_1.2,0.10609279870986939
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,1,power_law_1.2,0.047219198942184445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,1,power_law_1.2,0.05632640123367309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,8,power_law_1.2,0.1092479944229126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,1,power_law_1.2,0.06808320283889771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,1,power_law_1.2,0.08907520174980163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,32,power_law_1.01,0.07825919985771179
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,8,power_law_1.2,0.11207040548324584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,1,power_law_1.2,0.10778239965438843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,1,power_law_1.2,0.16563839912414552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,1,power_law_1.2,0.17432960271835327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,8,power_law_1.2,0.11552000045776367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,1,power_law_1.2,0.18795520067214966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,32,power_law_1.01,0.0858560025691986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,1,power_law_1.2,0.1985152006149292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,8,power_law_1.2,0.12443519830703735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,1,power_law_1.2,0.19882880449295043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,1,power_law_1.2,0.20465281009674072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,1,power_law_1.2,0.2207871913909912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,32,power_law_1.01,0.08005759716033936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,8,power_law_1.2,0.1255679965019226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,1,power_law_1.2,0.22213759422302246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,1,power_law_1.2,0.23347198963165283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,1,power_law_1.2,0.24705920219421387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,8,power_law_1.2,0.13865599632263184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,32,power_law_1.01,0.06299520134925843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,1,power_law_1.2,0.2553920030593872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,4,2,balanced,0.05641599992911021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,4,2,balanced,0.10124267141024272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,4,2,balanced,0.12029332915941875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,4,2,balanced,0.12133866548538208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,4,2,balanced,0.11959999799728394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,4,2,balanced,0.12026666601498921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,4,2,balanced,0.12158399820327759
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,4,2,balanced,0.12322133779525757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,4,2,balanced,0.12217600146929423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,4,2,balanced,0.12370133399963379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,1,power_law_1.2,0.27998080253601076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,4,2,balanced,0.12598400314648947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,4,2,balanced,0.1241386632124583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,4,2,balanced,0.12537599603335062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,4,2,balanced,0.1330880026022593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,4,2,balanced,0.14017066359519958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,4,2,balanced,0.14305599530537924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,4,2,balanced,0.14481066664059958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,4,2,balanced,0.17617066701253256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,4,2,balanced,0.18093333641688028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,8,power_law_1.2,0.17347840070724488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,4,2,balanced,0.22411733865737915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,1,power_law_1.2,0.32768640518188474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,4,2,balanced,0.24865599473317465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,4,2,balanced,0.32711466153462726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,4,2,balanced,0.38254932562510174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,4,2,balanced,0.5401386817296346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,4,2,balanced,0.6870240370432535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,32,power_law_1.01,0.06684799790382386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,4,2,balanced,1.0005919933319092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,1,power_law_1.2,0.36252799034118655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,4,2,balanced,1.2919893264770508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,4,2,balanced,1.6087573369344075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,1,power_law_1.2,0.45027837753295896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,8,power_law_1.2,0.17749760150909424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,4,2,balanced,2.5493332544962564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,1,power_law_1.2,0.5283967971801757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,32,power_law_1.01,0.06976640224456787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,4,2,balanced,4.961402575174968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,1,power_law_1.2,0.6919807910919189
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,8,power_law_1.2,0.21344640254974365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,1,power_law_1.2,0.9194304466247558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,32,power_law_1.01,0.07264000177383423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,1,power_law_1.2,1.2227840423583984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,8,power_law_1.2,0.2509439945220947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,1,power_law_1.2,1.56626558303833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,32,power_law_1.01,0.07495679855346679
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,8,power_law_1.2,0.32214400768280027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,1,power_law_1.2,2.163718414306641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,32,power_law_1.01,0.07518079876899719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,8,power_law_1.2,0.3673151969909668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,1,power_law_1.2,2.778003120422363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,32,power_law_1.01,0.07667199969291687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,1,power_law_1.2,3.3746177673339846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,8,power_law_1.2,0.49814400672912595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,32,power_law_1.01,0.07756159901618957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,1,power_law_1.2,5.159334564208985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,8,power_law_1.2,0.6200511932373047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,32,power_law_1.01,0.08181759715080261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,1,8,power_law_1.2,0.07083520293235779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,1,8,power_law_1.2,0.07775999903678894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,1,8,power_law_1.2,0.07092480063438415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,1,8,power_law_1.2,0.06863999962806702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,1,8,power_law_1.2,0.07056639790534973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,1,8,power_law_1.2,0.07117440104484558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,1,8,power_law_1.2,0.07156479954719544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,1,8,power_law_1.2,0.07345280051231384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,1,8,power_law_1.2,0.0754751980304718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,32,power_law_1.01,0.07901440262794494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,1,8,power_law_1.2,0.07674880027770996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,8,power_law_1.2,0.8521599769592285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,1,8,power_law_1.2,0.07793279886245727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,1,8,power_law_1.2,0.08414720296859741
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,1,8,power_law_1.2,0.08476160168647766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,1,8,power_law_1.2,0.09572479724884034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,1,8,power_law_1.2,0.10648319721221924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,1,8,power_law_1.2,0.10744960308074951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,1,8,power_law_1.2,0.1269055962562561
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,1,8,power_law_1.2,0.15857280492782594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,1,power_law_1.2,9.916876983642577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,32,power_law_1.01,0.08039039969444275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,1,8,power_law_1.2,0.17312639951705933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,1,8,power_law_1.2,0.232422399520874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,1,8,power_law_1.2,0.2735743999481201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,1,8,power_law_1.2,0.4117119789123535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,1,8,power_law_1.2,0.4895296096801758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,8,power_law_1.2,1.1235199928283692
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,1,8,power_law_1.2,0.7034880161285401
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,32,power_law_1.01,0.08557440042495727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,1,8,power_law_1.2,1.048799991607666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,1,8,power_law_1.2,1.6203136444091797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,1,8,power_law_1.2,2.075347137451172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,32,power_law_1.01,0.08670719861984252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,1,8,power_law_1.2,2.772774314880371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,8,8,power_law_1.2,1.4045120239257813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,32,power_law_1.01,0.09402880072593689
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,1,8,power_law_1.2,3.761363220214844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,32,power_law_1.01,0.10257279872894287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,1,8,power_law_1.2,7.317529296875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,32,power_law_1.01,0.10994559526443481
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,8,8,power_law_1.2,2.1265279769897463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,32,power_law_1.01,0.13403520584106446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,2,2,power_law_1.2,0.07918720245361328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,2,2,power_law_1.2,0.07441920042037964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,32,power_law_1.01,0.15429120063781737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,2,2,power_law_1.2,0.09080960154533387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,2,2,power_law_1.2,0.12184319496154786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,2,2,power_law_1.2,0.12254079580307006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,2,2,power_law_1.2,0.12431999444961547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,2,2,power_law_1.2,0.1266111969947815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,2,2,power_law_1.2,0.1293503999710083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,2,2,power_law_1.2,0.1348031997680664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,2,2,power_law_1.2,0.13381119966506957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,32,power_law_1.01,0.19500800371170043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,2,2,power_law_1.2,0.1337407946586609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,2,2,power_law_1.2,0.14036480188369752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,2,2,power_law_1.2,0.1477311968803406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,2,2,power_law_1.2,0.15312639474868775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,2,2,power_law_1.2,0.16976640224456788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,2,2,power_law_1.2,0.1775231957435608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,2,2,power_law_1.2,0.19307520389556884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,2,2,power_law_1.2,0.2405440092086792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,2,2,power_law_1.2,0.2625279903411865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,32,power_law_1.01,0.23301761150360106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,2,2,power_law_1.2,0.33512320518493655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,2,2,power_law_1.2,0.3836544036865234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,2,2,power_law_1.2,0.5046016216278076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,2,2,power_law_1.2,0.7097023963928223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,2,2,power_law_1.2,0.920531177520752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,8,8,power_law_1.2,5.1880638122558596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,32,power_law_1.01,0.30328960418701173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,2,2,power_law_1.2,1.2694144248962402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,2,2,power_law_1.2,1.9826751708984376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,2,2,power_law_1.2,2.6916799545288086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,32,power_law_1.01,0.3887295961380005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,64,power_law_1.2,0.03909760117530823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,2,2,power_law_1.2,3.260153579711914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,64,power_law_1.2,0.03198080062866211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,64,power_law_1.2,0.02771199941635132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,64,power_law_1.2,0.027654400467872618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,32,power_law_1.01,0.5703551769256592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,2,2,power_law_1.2,4.990060806274414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,64,power_law_1.2,0.029023998975753786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,64,power_law_1.2,0.02930560111999512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,64,power_law_1.2,0.029363200068473816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,64,power_law_1.2,0.029209598898887634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,32,power_law_1.01,0.7131135940551758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,64,power_law_1.2,0.029465600848197937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,2,2,power_law_1.2,9.55823974609375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,64,power_law_1.2,0.029606398940086365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,64,power_law_1.2,0.03000960052013397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,64,power_law_1.2,0.030899199843406677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,32,power_law_1.01,0.8616703987121582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,64,power_law_1.2,0.03146879971027374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,64,power_law_1.2,0.03274880051612854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,64,power_law_1.2,0.03689599931240082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,64,power_law_1.2,0.03619199991226196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,64,power_law_1.2,0.037868800759315493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,64,power_law_1.2,0.04206080138683319
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,32,power_law_1.01,1.4118592262268066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,4,balanced,0.04009599983692169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,4,balanced,0.03988266736268997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,4,balanced,0.0476746658484141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,4,balanced,0.06886399785677592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,64,power_law_1.2,0.047040000557899475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,4,balanced,0.09691199660301208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,4,balanced,0.09558399518330891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,4,balanced,0.09526933232943217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,4,balanced,0.09735999504725139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,4,balanced,0.09773332873980205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,4,balanced,0.09967466195424397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,4,balanced,0.09925333658854167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,4,balanced,0.1009226640065511
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,4,balanced,0.10401599605878194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,4,balanced,0.10451199611028035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,64,power_law_1.2,0.05242239832878113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,4,balanced,0.11202133695284526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,4,balanced,0.11183466513951619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,4,balanced,0.11904000242551167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,4,balanced,0.1262986660003662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,4,balanced,0.1367733379205068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,4,balanced,0.16077333688735962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,4,balanced,0.1720586617787679
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,4,balanced,0.22114133834838867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,64,power_law_1.2,0.06244480013847351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,4,balanced,0.24978667497634888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,4,balanced,0.3384373188018799
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,4,balanced,0.41332801183064777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,4,balanced,0.5790079832077026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,4,balanced,0.7425599892934164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,64,power_law_1.2,0.0648256003856659
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,4,balanced,0.9111573696136475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,4,balanced,1.40829865137736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,64,power_law_1.2,0.09422079920768738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,4,balanced,2.6987358729044595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,64,power_law_1.2,0.11869440078735352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,32,power_law_1.01,2.7543872833251952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,64,power_law_1.2,0.1486143946647644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,64,power_law_1.2,0.2064255952835083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,64,power_law_1.2,0.24725759029388428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,32,8,balanced,0.05481066803137461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,2,64,power_law_1.2,0.3007296085357666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,32,8,balanced,0.056186666091283165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,32,8,balanced,0.0584853341182073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,32,8,balanced,0.058634668588638306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,32,8,balanced,0.0666133314371109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,32,8,balanced,0.07671999931335449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,32,8,balanced,0.08705600102742513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,32,8,balanced,0.08727467060089111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,32,8,balanced,0.08942400415738423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,32,8,balanced,0.08929066856702168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,32,8,balanced,0.0900426705678304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,32,8,balanced,0.09126399954160054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,32,8,balanced,0.09332799911499023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,32,8,balanced,0.09406933188438416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,32,8,balanced,0.0997759997844696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,32,8,balanced,0.1011840005715688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,32,8,balanced,0.10573866963386536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,32,8,balanced,0.11581866939862569
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,2,64,power_law_1.2,0.4906303882598877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,32,8,balanced,0.12627200285593668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,32,8,balanced,0.1506186624368032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,32,8,balanced,0.1699519952138265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,32,8,balanced,0.20988800128300986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,32,8,balanced,0.25750933090845746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,32,8,balanced,0.33604268232981366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,32,8,balanced,0.44017601013183594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,32,8,balanced,0.6035573482513428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,32,8,balanced,0.7884426911671957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,32,8,balanced,0.9767839908599854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,32,8,balanced,1.5106453895568848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,2,64,power_law_1.2,1.0526975631713866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,32,8,balanced,2.93721612294515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,1,power_law_1.2,0.047539201378822324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,1,power_law_1.2,0.052025598287582395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,1,power_law_1.2,0.059935998916625974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,1,power_law_1.2,0.07095040082931518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,1,power_law_1.2,0.09030399918556213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,1,power_law_1.2,0.12713600397109986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,1,power_law_1.2,0.13597439527511596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,1,power_law_1.2,0.12085119485855103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,1,power_law_1.2,0.14131840467453002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,1,power_law_1.2,0.1488960027694702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,1,power_law_1.2,0.15427199602127076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,1,power_law_1.2,0.16106879711151123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,8,4,balanced,0.03875733415285746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,8,4,balanced,0.03976533313592275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,8,4,balanced,0.04159999887148539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,8,4,balanced,0.045834665497144066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,8,4,balanced,0.05213333169619242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,1,power_law_1.2,0.17216639518737792
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,8,4,balanced,0.055888002117474876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,1,power_law_1.2,0.14482560157775878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,8,4,balanced,0.05594133337338766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,8,4,balanced,0.05605866511662801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,8,4,balanced,0.055861334005991616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,8,4,balanced,0.05506666501363119
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,8,4,balanced,0.05583466589450836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,8,4,balanced,0.05825600028038025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,8,4,balanced,0.058133333921432495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,8,4,balanced,0.06035733222961426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,8,4,balanced,0.06512533128261566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,8,4,balanced,0.06823466718196869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,8,4,balanced,0.07195200026035309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,8,2,power_law_1.2,0.05685120224952698
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,8,4,balanced,0.0786293347676595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,1,power_law_1.2,0.17875200510025024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,8,4,balanced,0.08119999865690868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,8,4,balanced,0.1014453371365865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,8,4,balanced,0.11549866199493408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,8,4,balanced,0.15556266903877258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,8,4,balanced,0.17690134048461914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,8,4,balanced,0.25072532892227173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,8,4,balanced,0.3109653393427531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,8,4,balanced,0.4526880184809367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,1,power_law_1.2,0.17970559597015381
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,8,2,power_law_1.2,0.05692800283432007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,8,4,balanced,0.5930240154266357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,8,4,balanced,0.7377013365427653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,8,4,balanced,1.131717363993327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,1,power_law_1.2,0.195251202583313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,8,2,power_law_1.2,0.0652671992778778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,8,4,balanced,2.1985066731770835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,1,power_law_1.2,0.20347518920898439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,8,2,power_law_1.2,0.07200639843940734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,1,power_law_1.2,0.2265984058380127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,1,power_law_1.2,0.21722879409790039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,8,2,power_law_1.2,0.08766720294952393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,1,power_law_1.2,0.26501760482788084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,8,2,power_law_1.2,0.11165440082550049
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,1,power_law_1.2,0.3088896036148071
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,8,2,power_law_1.2,0.11760640144348145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,1,power_law_1.2,0.3971328020095825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,8,2,power_law_1.2,0.12214399576187134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,1,power_law_1.2,0.3807039976119995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,8,2,power_law_1.2,0.1255679965019226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,1,power_law_1.2,0.46592001914978026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,8,2,power_law_1.2,0.13030400276184081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,1,power_law_1.2,0.617907190322876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,8,2,power_law_1.2,0.13414399623870848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,1,power_law_1.2,0.8148863792419434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,8,2,power_law_1.2,0.13626879453659058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,1,power_law_1.2,0.5571775913238526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,8,2,power_law_1.2,0.14169600009918212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,1,power_law_1.2,1.0959808349609375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,8,2,power_law_1.2,0.14371199607849122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,1,power_law_1.2,1.4584511756896972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,8,2,power_law_1.2,0.15408639907836913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,8,2,power_law_1.2,0.15546879768371583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,1,power_law_1.2,1.1181695938110352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,1,power_law_1.2,2.0374847412109376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,8,2,power_law_1.2,0.16984319686889648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,8,2,power_law_1.2,0.19541759490966798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,8,2,power_law_1.2,0.22008960247039794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,1,power_law_1.2,2.6122623443603517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,8,2,power_law_1.2,0.2720767974853516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,1,power_law_1.2,1.2226176261901855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,8,2,power_law_1.2,0.31598720550537107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,32,power_law_1.01,0.07682560086250305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,1,power_law_1.2,3.1942399978637694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,8,2,power_law_1.2,0.41617918014526367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,32,power_law_1.01,0.06929919719696045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,8,2,power_law_1.2,0.5389952182769775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,1,power_law_1.2,1.3046784400939941
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,32,power_law_1.01,0.05608959794044495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,8,2,power_law_1.2,0.7353472232818603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,32,power_law_1.01,0.05107839703559876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,1,power_law_1.2,4.845625686645508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,8,2,power_law_1.2,1.0103296279907226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,32,power_law_1.01,0.059596800804138185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,1,power_law_1.2,1.3430848121643066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,32,power_law_1.01,0.060627198219299315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,8,2,power_law_1.2,1.3040191650390625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,32,power_law_1.01,0.07071999907493591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,8,2,power_law_1.2,1.6267200469970704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,32,power_law_1.01,0.07044479846954346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,1,power_law_1.2,1.3791616439819336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,1,power_law_1.2,9.478368377685547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,32,power_law_1.01,0.06958079934120179
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,8,2,power_law_1.2,2.0694143295288088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,32,power_law_1.01,0.07196159958839417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,1,power_law_1.2,1.4515775680541991
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,32,power_law_1.01,0.0746944010257721
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,8,2,power_law_1.2,3.2974334716796876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,32,power_law_1.01,0.07653120160102844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,32,power_law_1.01,0.08120959997177124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,1,power_law_1.2,1.5172991752624512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,32,power_law_1.01,0.0843392014503479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,32,power_law_1.01,0.09173759818077087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,8,2,power_law_1.2,6.60742416381836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,32,power_law_1.01,0.09258880019187928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,1,power_law_1.2,1.528428840637207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,32,power_law_1.01,0.09411839842796325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,32,power_law_1.01,0.10936959981918334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,32,power_law_1.01,0.11733759641647339
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,1,power_law_1.2,1.6221120834350586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,32,power_law_1.01,0.14289920330047606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,32,power_law_1.01,0.16042239665985109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,1,power_law_1.2,1.6655807495117188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,2,balanced,0.08316799998283386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,2,balanced,0.10444266597429912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,32,power_law_1.01,0.20352640151977539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,2,balanced,0.12830400466918945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,2,balanced,0.18179200092951456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,2,balanced,0.2737226684888204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,2,balanced,0.36425598462422687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,2,balanced,0.37166400750478107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,2,balanced,0.3738826513290405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,2,balanced,0.373418649037679
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,2,balanced,0.37643734614054364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,2,balanced,0.3868906497955322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,2,balanced,0.38731733957926434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,2,balanced,0.3903573354085286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,32,power_law_1.01,0.25408639907836916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,2,balanced,0.40035732587178546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,2,balanced,0.4147040049235026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,2,balanced,0.42218132813771564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,2,balanced,0.4437439839045207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,2,balanced,0.4976266622543335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,2,balanced,0.5501333475112915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,2,balanced,0.676042636235555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,2,balanced,0.7255787054697672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,1,power_law_1.2,1.6989952087402345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,32,power_law_1.01,0.3274751901626587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,2,balanced,1.0676746368408203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,2,balanced,1.162922700246175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,2,balanced,1.708672046661377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,2,balanced,2.0135199228922525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,32,power_law_1.01,0.4490496158599854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,2,balanced,3.019359906514486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,2,balanced,3.648261388142904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,32,power_law_1.01,0.6151552200317383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,1,power_law_1.2,1.7901504516601563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,2,2,balanced,4.587589263916016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,32,power_law_1.01,0.7380928039550781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,2,2,balanced,7.043008168538411
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,1,power_law_1.2,1.9458431243896483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,1,power_law_1.2,0.11943680047988892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,32,power_law_1.01,0.9572352409362793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,2,2,balanced,13.69826634724935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,1,power_law_1.2,0.1512895941734314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,32,power_law_1.01,1.4963135719299316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,1,power_law_1.2,1.9989952087402343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,1,power_law_1.2,0.23657600879669188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,1,power_law_1.2,2.263475227355957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,1,power_law_1.2,0.3886143922805786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,32,power_law_1.01,3.037651252746582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,1,power_law_1.2,0.564243221282959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,1,power_law_1.2,2.4895423889160155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,1,power_law_1.2,1.1276800155639648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,1,power_law_1.2,1.2185471534729004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,1,power_law_1.2,3.0118207931518555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,1,power_law_1.2,1.3093759536743164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,1,power_law_1.2,3.0324224472045898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,1,power_law_1.2,1.3709440231323242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,1,power_law_1.2,1.4201600074768066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,128,power_law_1.2,0.07662720084190369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,1,power_law_1.2,3.700377655029297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,1,power_law_1.2,1.4725888252258301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,128,power_law_1.2,0.0979968011379242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,32,power_law_1.2,0.1335744023323059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,1,power_law_1.2,1.4929535865783692
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,128,power_law_1.2,0.06080639958381653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,32,power_law_1.2,0.19615360498428344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,1,power_law_1.2,4.496787261962891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,32,power_law_1.2,0.14864640235900878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,1,power_law_1.2,1.5269887924194336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,128,power_law_1.2,0.06579840183258057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,32,power_law_1.2,0.09685760140419006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,1,power_law_1.2,1.642323112487793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,128,power_law_1.2,0.07235199809074402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,32,power_law_1.2,0.10172159671783447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,1,power_law_1.2,6.059584045410157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,128,power_law_1.2,0.07765759825706482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,32,power_law_1.2,0.11230720281600952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,1,power_law_1.2,1.6639488220214844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,32,power_law_1.2,0.12062079906463623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,128,power_law_1.2,0.07770879864692688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,1,power_law_1.2,1.6996288299560547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,32,power_law_1.2,0.12711679935455322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,128,power_law_1.2,0.0745024025440216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,32,power_law_1.2,0.13734400272369385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,1,power_law_1.2,1.814156723022461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,1,power_law_1.2,7.276729583740234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,128,power_law_1.2,0.07554559707641602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,32,power_law_1.2,0.14243199825286865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,1,power_law_1.2,2.03305606842041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,32,power_law_1.2,0.14042880535125732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,128,power_law_1.2,0.08007680177688599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,32,power_law_1.2,0.1444991946220398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,128,power_law_1.2,0.07795199751853943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,1,power_law_1.2,1.977734375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,32,power_law_1.2,0.14641280174255372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,128,power_law_1.2,0.08300160169601441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,2,1,power_law_1.2,8.730252838134765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,32,power_law_1.2,0.15599360466003417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,1,power_law_1.2,2.2496511459350588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,128,power_law_1.2,0.08702719807624817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,32,power_law_1.2,0.15564800500869752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,1,power_law_1.2,2.2644416809082033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,32,power_law_1.2,0.16053119897842408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,128,power_law_1.2,0.0870464026927948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,32,power_law_1.2,0.1667199969291687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,128,power_law_1.2,0.09619200229644775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,1,power_law_1.2,2.7396480560302736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,32,power_law_1.2,0.18805760145187378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,128,power_law_1.2,0.09645439982414246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,32,power_law_1.2,0.2065727949142456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,1,power_law_1.2,3.054252815246582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,2,1,power_law_1.2,13.66243896484375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,32,power_law_1.2,0.24457600116729736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,128,power_law_1.2,0.10673279762268066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,32,power_law_1.2,0.2766144037246704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,1,power_law_1.2,3.7195072174072266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,128,power_law_1.2,0.1255552053451538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,32,power_law_1.2,0.3480191946029663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,128,power_law_1.2,0.13504639863967896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,32,power_law_1.2,0.42621440887451173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,1,power_law_1.2,4.549190521240234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,32,power_law_1.2,0.5868159770965576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,128,power_law_1.2,0.15869439840316774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,32,power_law_1.2,0.6815231800079345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,128,power_law_1.2,0.17439359426498413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,1,power_law_1.2,6.166873550415039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,32,power_law_1.2,1.0047039985656738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,128,power_law_1.2,0.2528000116348267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,32,power_law_1.2,1.2512767791748047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,128,power_law_1.2,0.2497920036315918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,2,1,power_law_1.2,24.939768981933593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,32,power_law_1.2,1.4735360145568848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,1,power_law_1.2,7.058048248291016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,128,power_law_1.2,0.36224639415740967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,32,power_law_1.2,2.596601676940918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,128,power_law_1.2,0.45217280387878417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,1,1,power_law_1.2,8.394233703613281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,128,power_law_1.2,0.7751423835754394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,32,power_law_1.2,4.916716766357422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,128,power_law_1.2,0.7486720085144043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,1,1,balanced,0.07660800218582153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,1,1,balanced,0.13768000404040018
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,1,1,balanced,0.1616213321685791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,1,1,balanced,0.164000004529953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,1,1,balanced,0.16405866543451944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,1,1,balanced,0.16572266817092896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,1,128,power_law_1.2,1.0795071601867676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,1,1,power_law_1.2,13.073370361328125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,1,1,balanced,0.16554133097330728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,1,1,balanced,0.16662399967511496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,1,1,balanced,0.1693120002746582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,1,1,balanced,0.16910932461420694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,1,1,balanced,0.17114667097727457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,1,1,balanced,0.17324799299240112
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,1,1,balanced,0.17648533980051676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,1,1,balanced,0.18417600790659586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,1,1,balanced,0.1941386659940084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,1,1,balanced,0.19844265778859457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,1,1,balanced,0.21040532986323038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,1,1,balanced,0.2562826673189799
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,1,1,balanced,0.2746933301289876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,1,1,balanced,0.3494666814804077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,1,1,balanced,0.4306453466415405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,1,1,balanced,0.5841439962387085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,1,1,balanced,0.744213342666626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,1,1,balanced,1.0415039857228596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,1,1,balanced,1.3602612813313801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,1,1,balanced,1.971765359242757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,1,1,balanced,2.5423359870910645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,1,1,balanced,3.162421226501465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,1,128,power_law_1.2,1.4682368278503417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,1,1,balanced,4.927589416503906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,2,balanced,0.046181331078211464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,1,1,balanced,9.708058675130209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,2,balanced,0.06036800146102905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,2,balanced,0.0764160007238388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,2,balanced,0.10526933272679646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,2,balanced,0.1590720017751058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,2,balanced,0.18657066424687704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,2,balanced,0.18711467583974203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,2,balanced,0.19075733423233032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,2,balanced,0.1945013403892517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,2,balanced,0.19430400927861533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,2,balanced,0.19682133197784424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,2,balanced,0.2003893256187439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,2,balanced,0.20194133122762045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,2,balanced,0.2053333322207133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,2,balanced,0.21508800983428955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,2,balanced,0.22090667486190796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,2,balanced,0.22977066040039062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,2,balanced,0.25061333179473877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,2,balanced,0.27090134223302204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,2,balanced,0.3354719877243042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,2,balanced,0.36371731758117676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,2,balanced,0.5255893468856812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,2,balanced,0.5617226759592692
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,2,balanced,0.7932159900665283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,2,balanced,0.941648006439209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,2,balanced,1.4504586855570476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,2,balanced,1.7462186813354492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,2,balanced,2.0559520721435547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,2,balanced,3.239797274271647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,2,balanced,6.234890619913737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,1,1,power_law_1.2,23.853753662109376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,1,128,power_law_1.2,3.893804931640625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,64,power_law_1.01,0.03324800133705139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,64,power_law_1.01,0.029785600304603577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,64,power_law_1.01,0.02996479868888855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,64,power_law_1.01,0.030169600248336793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,64,power_law_1.01,0.03131519854068756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,64,power_law_1.01,0.031167998909950256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,64,power_law_1.01,0.030681601166725157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,64,power_law_1.01,0.031027200818061828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,128,balanced,0.05213333169619242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,128,balanced,0.05489066739877065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,128,balanced,0.04568000137805939
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,128,balanced,0.053344001372655235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,128,balanced,0.05400000015894572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,128,balanced,0.05374933282534281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,128,balanced,0.05614933371543884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,32,8,power_law_1.2,0.0369024008512497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,128,balanced,0.05594133337338766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,128,balanced,0.05406400064627329
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,128,balanced,0.054048001766204834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,128,balanced,0.05831466615200043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,64,power_law_1.01,0.031513598561286923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,128,balanced,0.058335999647776283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,128,balanced,0.057909334699312844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,128,balanced,0.05985599756240845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,128,balanced,0.0644160012404124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,128,balanced,0.06434666613737743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,128,balanced,0.06824000179767609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,128,balanced,0.08569600184758504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,64,power_law_1.01,0.03772799968719483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,128,balanced,0.08274133503437042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,32,8,power_law_1.2,0.030368000268936157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,128,balanced,0.11122666796048482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,128,balanced,0.12024000287055969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,128,balanced,0.1565546691417694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,128,balanced,0.15148267149925232
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,64,power_law_1.01,0.03163520097732544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,128,balanced,0.19106133778889975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,128,balanced,0.199562668800354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,128,balanced,0.2690453330675761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,32,8,power_law_1.2,0.02963840067386627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,64,power_law_1.01,0.03848319947719574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,128,balanced,0.305951992670695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,2,128,balanced,0.37220267454783124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,2,128,balanced,0.5231466690699259
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,64,power_law_1.01,0.03253119885921478
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,32,8,power_law_1.2,0.02980479896068573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,64,power_law_1.01,0.03395200073719025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,2,128,balanced,1.145248015721639
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,32,8,power_law_1.2,0.030700799822807313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,64,power_law_1.01,0.033228799700737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,64,power_law_1.01,0.03488639891147614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,32,8,power_law_1.2,0.03365119993686676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,64,power_law_1.01,0.034995201230049136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,64,power_law_1.01,0.033523198962211606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,32,8,power_law_1.2,0.03366400003433227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,64,power_law_1.01,0.03503359854221344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,32,8,power_law_1.2,0.033344000577926636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,64,power_law_1.01,0.0345984011888504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,64,power_law_1.01,0.035519999265670774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,32,8,power_law_1.2,0.034969601035118106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,64,power_law_1.01,0.037363201379776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,32,8,power_law_1.2,0.033990401029586795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,64,power_law_1.01,0.03578239977359772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,32,8,power_law_1.2,0.034841600060462954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,64,power_law_1.01,0.035308799147605895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,64,power_law_1.01,0.0382207989692688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,32,8,power_law_1.2,0.03640959858894348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,64,power_law_1.01,0.03562879860401154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,64,power_law_1.01,0.03964160084724426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,32,8,power_law_1.2,0.03707520067691803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,64,power_law_1.01,0.036627200245857236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,32,8,power_law_1.2,0.03893760144710541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,64,power_law_1.01,0.03738240003585815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,64,power_law_1.01,0.0442111998796463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,32,8,power_law_1.2,0.04160000085830688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,64,power_law_1.01,0.03841919898986816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,64,power_law_1.01,0.046412798762321475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,32,8,power_law_1.2,0.04384639859199524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,64,power_law_1.01,0.03924480080604553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,32,8,power_law_1.2,0.04715520143508911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,64,power_law_1.01,0.0526528000831604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,64,power_law_1.01,0.04213759899139404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,32,8,power_law_1.2,0.05457280278205871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,64,power_law_1.01,0.04280959963798523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,64,power_law_1.01,0.05400959849357605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,32,8,power_law_1.2,0.05916799902915955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,64,power_law_1.01,0.04597119987010956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,64,power_law_1.01,0.06811519861221313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,32,8,power_law_1.2,0.07443199753761291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,64,power_law_1.01,0.04772480130195618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,32,8,power_law_1.2,0.08306559920310974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,64,power_law_1.01,0.04885759949684143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,64,power_law_1.01,0.07613440155982971
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,32,8,power_law_1.2,0.10940159559249878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,64,power_law_1.01,0.055232000350952146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,64,power_law_1.01,0.11938560009002686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,32,8,power_law_1.2,0.1412608027458191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,64,power_law_1.01,0.06291199922561645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,32,8,power_law_1.2,0.203110408782959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,64,power_law_1.01,0.07949439883232116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,64,power_law_1.01,0.14261759519577027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,8,balanced,0.030554667115211487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,8,balanced,0.02962133288383484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,64,power_law_1.01,0.08977280259132385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,32,8,power_law_1.2,0.24661118984222413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,8,balanced,0.029989334444204967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,8,balanced,0.029904000461101532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,8,balanced,0.03373866776625315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,8,balanced,0.03573866685231527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,8,balanced,0.03365866591533025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,8,balanced,0.03501333296298981
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,8,balanced,0.03562666724125544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,8,balanced,0.03570666660865148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,64,power_law_1.01,0.21176960468292236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,8,balanced,0.03636800001064936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,8,balanced,0.03590933233499527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,8,balanced,0.0378506655494372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,8,balanced,0.03779733429352442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,8,balanced,0.04159466673930486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,8,balanced,0.042026668787002563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,64,power_law_1.01,0.12284159660339355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,8,balanced,0.04192000130812327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,8,balanced,0.04387733340263367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,8,balanced,0.052111998200416565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,8,balanced,0.0543146679798762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,32,8,power_law_1.2,0.37838718891143797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,8,balanced,0.05986666679382324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,8,balanced,0.0705813318490982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,8,balanced,0.08699199557304382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,8,balanced,0.12250133355458577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,8,balanced,0.16214932998021445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,8,balanced,0.21783467133839926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,8,balanced,0.274234672387441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,8,balanced,0.33378132184346515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,64,power_law_1.01,0.16476800441741943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,64,power_law_1.01,0.24865920543670655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,8,balanced,0.5112053155899048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,8,balanced,0.9988693396250407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,32,8,power_law_1.2,0.4979584217071533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,64,power_law_1.01,0.2051392078399658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,64,power_law_1.01,0.3003904104232788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,32,8,power_law_1.2,0.7510528087615966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,64,power_law_1.01,0.27645440101623536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,64,power_law_1.01,0.46723198890686035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,64,power_law_1.01,0.40010881423950195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,32,8,power_law_1.2,1.0212736129760742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,64,power_law_1.01,0.4994815826416016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,64,power_law_1.01,0.9223103523254395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,32,8,power_law_1.2,2.778860855102539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,64,power_law_1.01,1.124940776824951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,4,power_law_1.2,0.05092480182647705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,4,power_law_1.2,0.05234559774398804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,4,power_law_1.2,0.04353919923305512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,4,power_law_1.2,0.05729920268058777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,4,power_law_1.2,0.06665599942207337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,4,power_law_1.2,0.10113919973373413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,4,power_law_1.2,0.10403200387954711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,4,power_law_1.2,0.10883200168609619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,4,power_law_1.2,0.11183359622955322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,4,power_law_1.2,0.11308159828186035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,4,power_law_1.2,0.11673599481582642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,4,power_law_1.2,0.11772160530090332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,4,power_law_1.2,0.12111999988555908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,2,power_law_1.2,0.029600000381469725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,4,power_law_1.2,0.1212928056716919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,16,power_law_1.01,0.059436798095703125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,128,balanced,0.05413866539796194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,2,power_law_1.2,0.03141759932041168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,4,power_law_1.2,0.1318719983100891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,128,balanced,0.0561653325955073
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,128,balanced,0.05402133365472158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,128,balanced,0.05573866764704386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,128,balanced,0.05392533540725708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,128,balanced,0.060266668597857155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,128,balanced,0.060165335734685264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,128,balanced,0.060309335589408875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,128,balanced,0.06449066599210103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,2,power_law_1.2,0.035724800825119016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,128,balanced,0.06382399797439575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,4,power_law_1.2,0.12531839609146117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,128,balanced,0.06407466530799866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,128,balanced,0.06431999802589417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,128,balanced,0.0645546664794286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,128,balanced,0.06632533172766368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,128,balanced,0.06863999863465627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,128,balanced,0.07035199801127116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,128,balanced,0.07829333345095317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,128,balanced,0.08055466910203297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,2,power_law_1.2,0.042752000689506534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,128,balanced,0.0890933374563853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,4,power_law_1.2,0.14907519817352294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,128,balanced,0.1104213297367096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,128,balanced,0.13133866588274637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,128,balanced,0.16900267203648886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,16,power_law_1.01,0.05799679756164551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,128,balanced,0.19893866777420044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,128,balanced,0.26814399162928265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,2,power_law_1.2,0.04704639911651611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,128,balanced,0.33241599798202515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,4,power_law_1.2,0.17203840017318725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,128,balanced,0.4626986583073934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,128,balanced,0.5926293134689331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,2,power_law_1.2,0.06435840129852295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,128,balanced,0.7298666636149088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,4,power_law_1.2,0.16060800552368165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,128,balanced,1.1159199873606365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,2,power_law_1.2,0.06884480118751526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,4,power_law_1.2,0.1900928020477295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,128,balanced,2.196714719136556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,16,power_law_1.01,0.047193598747253415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,2,power_law_1.2,0.07052800059318542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,4,power_law_1.2,0.20799360275268555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,2,power_law_1.2,0.07226240038871765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,4,power_law_1.2,0.266975998878479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,2,power_law_1.2,0.07193599939346314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,16,power_law_1.01,0.04809600114822388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,2,power_law_1.2,0.07305600047111512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,4,power_law_1.2,0.27150719165802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,2,power_law_1.2,0.07722240090370178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,4,power_law_1.2,0.3264064073562622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,2,power_law_1.2,0.07857919931411743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,16,power_law_1.01,0.05146880149841308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,4,power_law_1.2,0.3955071926116943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,2,power_law_1.2,0.08117120265960694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,2,power_law_1.2,0.09038720130920411
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,4,power_law_1.2,0.5408512115478515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,2,power_law_1.2,0.09670400023460388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,16,power_law_1.01,0.05443840026855469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,2,power_law_1.2,0.09857919812202454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,4,power_law_1.2,0.7651391983032226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,2,power_law_1.2,0.10880000591278076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,16,power_law_1.01,0.05610880255699158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,2,power_law_1.2,0.11768319606781005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,1,4,power_law_1.2,0.8503487586975098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,2,power_law_1.2,0.1417407989501953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,2,power_law_1.2,0.14620800018310548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,16,power_law_1.01,0.05915520191192627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,1,4,power_law_1.2,1.2094079971313476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,2,power_law_1.2,0.17310080528259278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,2,power_law_1.2,0.19743360280990602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,16,power_law_1.01,0.05861759781837463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,2,power_law_1.2,0.28268160820007326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,1,4,power_law_1.2,2.475200080871582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,2,power_law_1.2,0.33562240600585935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,16,power_law_1.01,0.05960959792137146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,2,power_law_1.2,0.5103807926177979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,16,power_law_1.01,0.05801600217819214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,4,8,balanced,0.03514133393764496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,4,8,balanced,0.035717333356539406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,4,8,balanced,0.034160000582536064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,4,8,balanced,0.033813332517941795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,4,8,balanced,0.03387733300526937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,4,8,balanced,0.035717333356539406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,4,8,balanced,0.03465600063403448
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,2,power_law_1.2,0.5841023921966553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,4,8,balanced,0.03408533334732056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,4,8,balanced,0.03624533365170161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,4,8,balanced,0.03555200000603994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,4,8,balanced,0.03803733239571253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,4,8,balanced,0.03689600030581156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,4,8,balanced,0.03748800108830134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,4,8,balanced,0.037621334195137024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,4,8,balanced,0.0436160018046697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,4,8,balanced,0.044533332188924156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,4,8,balanced,0.042810668547948204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,4,8,balanced,0.05412266651789347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,4,8,balanced,0.052000001072883606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,4,8,balanced,0.0705973356962204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,4,8,balanced,0.07945066690444946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,16,power_law_1.01,0.060185599327087405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,4,8,balanced,0.1036959985891978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,4,8,balanced,0.11586667100588481
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,4,8,balanced,0.15549866358439127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,4,8,balanced,0.18754132588704428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,2688,1856,6,128,4,2,power_law_1.2,0.682425594329834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,4,8,balanced,0.2611520091692607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,4,8,balanced,0.3243733247121175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,4,8,balanced,0.39372265338897705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,4,8,balanced,0.5834559996922811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,4,8,balanced,1.1112159887949626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,16,power_law_1.01,0.061766397953033444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,2688,1856,6,128,4,2,power_law_1.2,1.082374382019043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,16,power_law_1.01,0.06480000019073487
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,16,power_law_1.01,0.07084159851074219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,2688,1856,6,128,4,2,power_law_1.2,1.974073600769043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,16,power_law_1.01,0.07276160120964051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,16,power_law_1.01,0.07608320116996765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,16,power_law_1.01,0.08458240032196045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,16,power_law_1.01,0.09285119771957398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,16,power_law_1.01,0.110534405708313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,1,1,power_law_1.01,0.12506879568099977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,16,power_law_1.01,0.13392000198364257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,1,1,power_law_1.01,0.17890559434890746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,1,1,power_law_1.01,0.19533439874649047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,1,1,power_law_1.01,0.37411839962005616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,1,1,power_law_1.01,0.38184959888458253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,1,1,power_law_1.01,0.3823872089385986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,16,power_law_1.01,0.16678400039672853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,1,1,power_law_1.01,0.3817087888717651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,1,1,power_law_1.01,0.39921278953552247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,1,1,power_law_1.01,0.4096127986907959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,1,1,power_law_1.01,0.41938557624816897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,1,1,power_law_1.01,0.4029183864593506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,1,1,power_law_1.01,0.4248064041137695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,16,power_law_1.01,0.19534720182418824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,1,1,power_law_1.01,0.42332801818847654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,1,1,power_law_1.01,0.47379841804504397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,1,1,power_law_1.01,0.5314943790435791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,1,1,power_law_1.01,0.5651135921478272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,128,power_law_1.2,0.03827199935913086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,1,1,power_law_1.01,0.5559487819671631
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,16,power_law_1.01,0.25541760921478274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,1,1,power_law_1.01,0.6604991912841797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,128,power_law_1.2,0.037747201323509214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,1,1,power_law_1.01,0.6868351936340332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,1,1,power_law_1.01,0.8494848251342774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,128,power_law_1.2,0.03831680119037628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,1,1,power_law_1.01,0.9869248390197753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,16,power_law_1.01,0.30892798900604246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,128,power_law_1.2,0.038790398836135866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,1,1,power_law_1.01,1.3172543525695801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,128,power_law_1.2,0.0383296012878418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,1,1,power_law_1.01,1.6472639083862304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,128,power_law_1.2,0.039743998646736146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,8,power_law_1.01,0.09023360013961793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,16,power_law_1.01,0.43909759521484376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,1,1,power_law_1.01,2.3168319702148437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,128,power_law_1.2,0.03937279880046844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,8,power_law_1.01,0.0828607976436615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,128,power_law_1.2,0.03882879912853241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,1,1,power_law_1.01,2.995987129211426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,8,power_law_1.01,0.07140480279922486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,128,power_law_1.2,0.03968639969825745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,16,power_law_1.01,0.5757696151733398
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,8,power_law_1.01,0.08526080250740051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,128,power_law_1.2,0.03960959911346436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,1,1,power_law_1.01,4.43070068359375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,128,power_law_1.2,0.04035840034484863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,8,power_law_1.01,0.09671679735183716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,128,power_law_1.2,0.040243199467658995
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,8,power_law_1.01,0.11921919584274292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,1,1,power_law_1.01,5.791334533691407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,16,power_law_1.01,0.7207104206085205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,128,power_law_1.2,0.04056319892406464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,8,power_law_1.01,0.12424319982528687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,128,power_law_1.2,0.042828801274299624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,8,power_law_1.01,0.12458879947662353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,128,power_law_1.2,0.04714879989624023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,8,power_law_1.01,0.1277184009552002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,1,1,power_law_1.01,7.274547576904297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,128,power_law_1.2,0.04618239998817444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,8,power_law_1.01,0.12833919525146484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,128,power_law_1.2,0.04940800070762634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,16,power_law_1.01,1.1026240348815919
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,8,power_law_1.01,0.12743040323257446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,128,power_law_1.2,0.05724160075187683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,8,power_law_1.01,0.1331264019012451
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,128,power_law_1.2,0.0633791983127594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,128,power_law_1.2,0.07631999850273133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,8,power_law_1.01,0.13365119695663452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,1,1,power_law_1.01,11.735462188720703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,128,power_law_1.2,0.09134079813957215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,8,power_law_1.01,0.1374079942703247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,128,power_law_1.2,0.13223040103912354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,8,power_law_1.01,0.14715520143508912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,16,power_law_1.01,2.1857791900634767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,128,power_law_1.2,0.16458879709243773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,8,power_law_1.01,0.1512768030166626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,8,power_law_1.01,0.15665279626846312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,128,power_law_1.2,0.2391103982925415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,8,power_law_1.01,0.17802879810333253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,128,power_law_1.2,0.30047359466552737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,8,power_law_1.01,0.19202560186386108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,128,power_law_1.2,0.45212159156799314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,8,power_law_1.01,0.22778239250183105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,8,power_law_1.01,0.25109119415283204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,128,power_law_1.2,0.5934336185455322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,1,1,power_law_1.01,22.236697387695312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,8,power_law_1.01,0.3335423946380615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,128,power_law_1.2,0.7287424087524415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,8,power_law_1.01,0.38214399814605715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,8,power_law_1.01,0.4961855888366699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,128,power_law_1.2,1.1409279823303222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,8,power_law_1.01,0.596992015838623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,8,power_law_1.01,0.895680046081543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,128,power_law_1.2,2.244646453857422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,8,power_law_1.01,1.0294336318969726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,8,power_law_1.01,1.2964032173156739
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,64,balanced,0.05570133527119955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,64,balanced,0.057589332262674965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,64,balanced,0.05754133562246958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,64,balanced,0.0566293348868688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,64,balanced,0.06286400059858958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,64,balanced,0.07010666529337566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,64,balanced,0.07141866783301036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,64,balanced,0.07131200035413106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,64,balanced,0.07339199880758922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,64,balanced,0.07210666437943776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,64,balanced,0.0734506646792094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,64,balanced,0.07525333265463512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,64,balanced,0.07550933460394542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,64,balanced,0.07444266478220622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,8,power_law_1.01,1.9650432586669921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,64,balanced,0.08214933176835378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,64,balanced,0.08993066350618999
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,64,balanced,0.08876267075538635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,64,balanced,0.09943999846776326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,64,balanced,0.10777067144711812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,64,balanced,0.12542399764060974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,64,balanced,0.14198933045069376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,64,balanced,0.17859200636545816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,64,balanced,0.20962133010228476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,64,balanced,0.27974933385849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,64,balanced,0.34485868612925213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,64,balanced,0.4901119867960612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,64,balanced,0.6198346614837646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,64,balanced,0.7632959683736166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,64,balanced,1.1709226767222087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,64,balanced,2.29367462793986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,8,power_law_1.01,3.856371307373047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,32,1,power_law_1.2,0.07771520018577575
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,32,1,power_law_1.2,0.08108800053596496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,32,1,power_law_1.2,0.08894720077514648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,32,1,power_law_1.2,0.10643839836120605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,32,1,power_law_1.2,0.12894079685211182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,32,1,power_law_1.2,0.15036799907684326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,32,power_law_1.2,0.056518399715423585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,32,1,power_law_1.2,0.19145599603652955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,32,1,power_law_1.2,0.20373120307922363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,32,power_law_1.2,0.05790079832077026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,32,1,power_law_1.2,0.2085184097290039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,32,1,power_law_1.2,0.2162303924560547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,32,power_law_1.2,0.048332801461219786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,32,1,power_law_1.2,0.2251584053039551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,256,balanced,0.05668266614278158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,256,balanced,0.06136533121267954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,256,balanced,0.058378666639328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,256,balanced,0.06638399759928386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,256,balanced,0.06259733438491821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,256,balanced,0.06495466828346252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,256,balanced,0.068122665087382
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,256,balanced,0.06702400247255962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,32,1,power_law_1.2,0.23364479541778566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,256,balanced,0.06654933094978333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,256,balanced,0.06628799935181935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,256,balanced,0.06917333106199901
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,32,power_law_1.2,0.050963199138641356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,256,balanced,0.07077333331108093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,8,4,power_law_1.2,0.04031359851360321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,256,balanced,0.0706879993279775
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,256,balanced,0.07373333474000295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,8,4,power_law_1.2,0.03395200073719025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,256,balanced,0.07497600217660268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,256,balanced,0.07582933207352956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,8,4,power_law_1.2,0.03209599852561951
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,256,balanced,0.0812960018714269
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,256,balanced,0.09317866961161296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,8,4,power_law_1.2,0.0326335996389389
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,32,1,power_law_1.2,0.23912959098815917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,256,balanced,0.09748799602190654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,256,balanced,0.11802132924397786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,8,4,power_law_1.2,0.03276160061359405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,256,balanced,0.14217066764831543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,8,4,power_law_1.2,0.033913600444793704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,256,balanced,0.19430400927861533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,8,4,power_law_1.2,0.041196799278259276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,256,balanced,0.20217067003250122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,32,1,power_law_1.2,0.2556351900100708
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,8,4,power_law_1.2,0.04320000112056732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,256,balanced,0.2635040084520976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,32,power_law_1.2,0.053523200750350955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,8,4,power_law_1.2,0.043884798884391785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,256,balanced,0.31320534149805707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,8,4,power_law_1.2,0.04604800045490265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,256,balanced,0.43221867084503174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,8,4,power_law_1.2,0.04656639993190766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,256,balanced,0.5327680110931396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,8,4,power_law_1.2,0.04945279955863953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,32,1,power_law_1.2,0.26017279624938966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,8,4,power_law_1.2,0.048902401328086854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,1,256,balanced,0.6320373217264811
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,8,4,power_law_1.2,0.05196800231933594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,1,256,balanced,0.9564639727274576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,8,4,power_law_1.2,0.059552001953125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,8,4,power_law_1.2,0.06266239881515503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,8,4,power_law_1.2,0.07264639735221863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,32,1,power_law_1.2,0.273305606842041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,32,power_law_1.2,0.056678402423858645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,1,256,balanced,2.008277257283529
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,8,4,power_law_1.2,0.08302080035209655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,8,4,power_law_1.2,0.0855296015739441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,8,4,power_law_1.2,0.10603519678115844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,8,4,power_law_1.2,0.13765759468078614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,8,4,power_law_1.2,0.16165759563446044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,32,1,power_law_1.2,0.28874878883361815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,8,4,power_law_1.2,0.17608959674835206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,8,4,power_law_1.2,0.31242239475250244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,32,power_law_1.2,0.05850239992141724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,8,4,power_law_1.2,0.39983999729156494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,32,1,power_law_1.2,0.34031360149383544
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,8,4,power_law_1.2,0.6391744136810302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,8,4,power_law_1.2,0.5944255828857422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,32,1,power_law_1.2,0.3783807992935181
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,8,4,power_law_1.2,0.8723711967468262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,32,power_law_1.2,0.05976960062980652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,8,4,power_law_1.2,1.0545151710510254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,32,1,power_law_1.2,0.46871042251586914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,8,4,power_law_1.2,2.103206443786621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,32,power_law_1.2,0.05802879929542541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,32,1,power_law_1.2,0.5507840156555176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,32,1,power_law_1.2,0.7229887962341308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,32,power_law_1.2,0.059308797121047974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,32,1,power_law_1.2,0.8920831680297852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,32,power_law_1.2,0.0638592004776001
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,32,1,power_law_1.2,1.227187156677246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,32,power_law_1.2,0.06694399714469909
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,32,1,power_law_1.2,1.5633279800415039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,32,power_law_1.2,0.06789759993553161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,32,1,power_law_1.2,2.2427520751953125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,32,power_law_1.2,0.07322239875793457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,32,1,power_law_1.2,3.002252769470215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,32,power_law_1.2,0.07836160063743591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,32,1,power_law_1.2,3.6829246520996093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,1,power_law_1.2,0.08513919711112976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,32,power_law_1.2,0.08181120157241821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,1,balanced,0.0974773367245992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,1,balanced,0.1617759962876638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,1,balanced,0.1931040088335673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,1,balanced,0.3187999924023946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,1,balanced,0.5752266645431519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,1,balanced,0.7062346935272217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,1,balanced,0.7133599917093912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,1,balanced,0.7160373528798422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,1,balanced,0.7187893390655518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,1,balanced,0.7201066811879476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,1,balanced,0.7283039887746176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,32,power_law_1.2,0.08394240140914917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,1,balanced,0.7329066594441732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,1,balanced,0.7363146940867106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,1,balanced,0.7469333012898763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,1,balanced,0.7591893672943115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,1,balanced,0.7718773682912191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,32,1,power_law_1.2,5.629753494262696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,1,balanced,0.7907466888427734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,1,balanced,0.8308959801991781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,1,power_law_1.2,0.0982591986656189
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,1,balanced,0.8713706334431967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,1,balanced,1.143280029296875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,32,power_law_1.2,0.10239360332489014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,1,balanced,1.0560959974924724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,1,balanced,1.4172213872273762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,1,balanced,1.4528320630391438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,1,balanced,2.1533759435017905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,1,balanced,2.4870079358418784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,32,power_law_1.2,0.10743039846420288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,1,balanced,3.755530675252279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,1,power_law_1.2,0.14455039501190187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,1,balanced,4.513087908426921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,32,power_law_1.2,0.1329792022705078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,1,balanced,5.264533360799153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,32,1,power_law_1.2,11.020256042480469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,1,power_law_1.2,0.22016000747680664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,32,power_law_1.2,0.1626688003540039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,1,balanced,8.624607721964518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,32,power_law_1.2,0.20578560829162598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,1,power_law_1.2,0.31667840480804443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,1,balanced,16.360490163167317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,32,power_law_1.2,0.22067201137542725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,1,power_law_1.2,0.5939263820648193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,32,power_law_1.2,0.3071295976638794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,1,power_law_1.2,0.655673599243164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,32,power_law_1.2,0.35538558959960936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,1,power_law_1.2,0.6926271915435791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,32,power_law_1.2,0.5243391990661621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,1,power_law_1.2,0.7139135837554932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,1,power_law_1.2,0.052223998308181765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,32,power_law_1.2,0.725542402267456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,1,power_law_1.2,0.7364287853240967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,1,power_law_1.2,0.05623040199279785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,8,power_law_1.01,0.046393600106239316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,1,power_law_1.2,0.07034239768981934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,1,power_law_1.2,0.7834432125091553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,8,power_law_1.01,0.05006719827651977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,1,power_law_1.2,0.08799999952316284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,8,power_law_1.01,0.04350079894065857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,8,32,power_law_1.2,0.7617023944854736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,8,power_law_1.01,0.04550400078296661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,1,power_law_1.2,0.1292799949645996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,8,power_law_1.01,0.04904319941997528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,1,power_law_1.2,0.8008831977844239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,8,power_law_1.01,0.05265920162200928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,1,power_law_1.2,0.16264959573745727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,8,power_law_1.01,0.05368959903717041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,8,power_law_1.01,0.055852800607681274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,1,power_law_1.2,0.19573760032653809
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,1,power_law_1.2,0.8140607833862304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,8,power_law_1.01,0.05549439787864685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,8,32,power_law_1.2,1.24268798828125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,8,power_law_1.01,0.05694079995155334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,1,power_law_1.2,0.261843204498291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,8,power_law_1.01,0.0608959972858429
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,8,power_law_1.01,0.06130560040473938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,1,power_law_1.2,0.8632384300231933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,1,power_law_1.2,0.2789056062698364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,8,power_law_1.01,0.06347519755363465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,8,power_law_1.01,0.06531839966773986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,1,power_law_1.2,0.27985920906066897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,8,power_law_1.01,0.07285119891166687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,8,power_law_1.01,0.07772160172462464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,1,power_law_1.2,0.9019136428833008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,1,power_law_1.2,0.30399360656738283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,8,power_law_1.01,0.07848320007324219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,8,power_law_1.01,0.09511680006980897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,1,power_law_1.2,0.32255361080169676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,8,power_law_1.01,0.10857599973678589
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,1,power_law_1.2,0.9253503799438476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,8,power_law_1.01,0.14412800073623658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,1,power_law_1.2,0.32829439640045166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,8,32,power_law_1.2,2.916223907470703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,8,power_law_1.01,0.16679680347442627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,1,power_law_1.2,0.343174409866333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,8,power_law_1.01,0.21860480308532715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,8,power_law_1.01,0.2712064027786255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,1,power_law_1.2,0.9788736343383789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,1,power_law_1.2,0.3679359912872314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,8,power_law_1.01,0.3840575933456421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,1,power_law_1.2,0.38156158924102784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,8,power_law_1.01,0.447603178024292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,1,power_law_1.2,1.1116991996765138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,8,power_law_1.01,0.6435711860656739
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,1,power_law_1.2,0.3985088109970093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,8,power_law_1.01,0.8350591659545898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,1,power_law_1.2,0.45728001594543455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,1,power_law_1.2,1.1242560386657714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,1,power_law_1.2,0.5148352146148681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,8,power_law_1.01,1.0802047729492188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,1,power_law_1.2,0.6134463787078858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,1,power_law_1.2,1.3200448036193848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,8,power_law_1.01,1.6004671096801757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,1,power_law_1.2,0.682150411605835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,1,power_law_1.2,0.8625344276428223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,1,power_law_1.2,1.3593088150024415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,32,balanced,0.03675200045108795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,32,balanced,0.03422933320204417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,32,balanced,0.03012266755104065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,32,balanced,0.03346133232116699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,32,balanced,0.03552533437808355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,32,balanced,0.035962666074434914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,32,balanced,0.03573333223660787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,32,balanced,0.03581333408753077
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,8,power_law_1.01,3.1447423934936523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,32,balanced,0.03570666660865148
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,32,balanced,0.03629866739114126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,32,balanced,0.03678400069475174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,32,balanced,0.035936000446478523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,32,balanced,0.036288000643253326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,32,balanced,0.037685332198937736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,32,balanced,0.039877332746982574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,32,balanced,0.04163199911514918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,32,balanced,0.04178133110205332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,32,balanced,0.04422933359940847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,32,balanced,0.04737600187460581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,32,balanced,0.052111998200416565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,32,balanced,0.05890666445096334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,32,balanced,0.06665066878000896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,32,balanced,0.07358933488527934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,1,power_law_1.2,1.0618368148803712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,32,balanced,0.08505599697430928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,32,balanced,0.09752000371615092
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,32,balanced,0.1170186698436737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,32,balanced,0.1532746652762095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,32,balanced,0.17960000038146973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,32,balanced,0.2367146611213684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,32,balanced,0.427072008450826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,1,power_law_1.2,1.6524799346923829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,1,power_law_1.2,1.3771648406982422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,1,power_law_1.2,1.8253440856933594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,1,power_law_1.2,1.917919921875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,1,power_law_1.2,2.4797887802124023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,1,power_law_1.2,2.5002815246582033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,1,power_law_1.2,2.9493696212768556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,1,power_law_1.2,3.2781566619873046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,1,power_law_1.2,3.590367889404297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,16,power_law_1.01,0.038150399923324585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,4,1,balanced,0.040149333576361336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,4,1,balanced,0.04889066517353058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,16,power_law_1.01,0.03377279937267304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,4,1,balanced,0.06850133339564006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,4,1,balanced,0.06899199883143108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,4,1,balanced,0.06940799951553345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,4,1,balanced,0.07116800049940745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,4,1,balanced,0.0712960014740626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,4,1,balanced,0.07049599786599477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,4,1,balanced,0.07003200054168701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,4,1,balanced,0.07137066622575124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,4,1,balanced,0.07293333113193512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,4,1,balanced,0.07418133318424225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,16,power_law_1.01,0.030700799822807313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,4,1,balanced,0.0749013324578603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,4,1,balanced,0.08329600095748901
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,1,power_law_1.2,5.647776031494141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,4,1,balanced,0.09014399846394856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,4,1,balanced,0.0930560032526652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,4,1,balanced,0.09131733576456706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,4,1,balanced,0.11262933413187663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,4,1,balanced,0.12146666646003723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,4,1,balanced,0.15132799744606018
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,16,power_law_1.01,0.03078399896621704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,4,1,balanced,0.1795413295427958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,4,1,balanced,0.23757867018381754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,1,power_law_1.2,4.446758270263672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,4,1,balanced,0.281823992729187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,4,1,balanced,0.3945173422495524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,4,1,balanced,0.4980586767196655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,16,power_law_1.01,0.03271040022373199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,4,1,balanced,0.7095253467559814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,4,1,balanced,0.9128106435139974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,16,power_law_1.01,0.03407360017299652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,4,1,balanced,1.1297866503397624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,4,1,balanced,1.7461919784545898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,16,power_law_1.01,0.03454720079898834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,4,1,balanced,3.47379207611084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,16,power_law_1.01,0.03496319949626923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,16,power_law_1.01,0.035417601466178894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,16,power_law_1.01,0.036396801471710205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,16,power_law_1.01,0.03743999898433685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,16,power_law_1.01,0.03845759928226471
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,1,power_law_1.2,5.500223922729492
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,1,power_law_1.2,10.656896209716797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,16,power_law_1.01,0.03954559862613678
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,16,power_law_1.01,0.042905598878860474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,16,power_law_1.01,0.04787839949131012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,16,power_law_1.01,0.05041279792785645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,16,power_law_1.01,0.05063679814338684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,16,power_law_1.01,0.054841601848602296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,16,power_law_1.01,0.06261759996414185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,16,power_law_1.01,0.07861120104789734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,16,power_law_1.01,0.0917952001094818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,4,1,power_law_1.2,6.587091064453125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,16,power_law_1.01,0.12229119539260865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,16,power_law_1.01,0.15166720151901245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,16,power_law_1.01,0.19166719913482666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,16,power_law_1.01,0.25408639907836916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,16,power_law_1.01,0.34391040802001954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,16,power_law_1.01,0.4579455852508545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,16,power_law_1.01,0.5737343788146972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,16,power_law_1.2,0.08297600150108338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,4,1,power_law_1.2,10.1123779296875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,16,power_law_1.01,0.9086976051330566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,4,power_law_1.01,0.07229440212249756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,16,power_law_1.2,0.07159039974212647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,4,power_law_1.01,0.0906175971031189
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,4,power_law_1.01,0.07369599938392639
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,16,power_law_1.2,0.0572160005569458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,4,power_law_1.01,0.09223039746284485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,16,power_law_1.01,1.776838493347168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,4,power_law_1.01,0.11550719738006592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,4,power_law_1.01,0.14967039823532105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,16,power_law_1.2,0.06603519916534424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,4,power_law_1.01,0.15879039764404296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,4,power_law_1.01,0.16359039545059204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,4,power_law_1.01,0.1652351975440979
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,16,power_law_1.2,0.07452800273895263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,4,power_law_1.01,0.16594560146331788
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,4,power_law_1.01,0.174399995803833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,16,power_law_1.2,0.09198079705238342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,4,power_law_1.01,0.17997440099716186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,4,power_law_1.01,0.17763839960098265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,4,power_law_1.01,0.1843775987625122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,16,power_law_1.2,0.09594240188598632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,4,power_law_1.01,0.19753600358963014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,4,power_law_1.01,0.20894079208374022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,4,1,power_law_1.2,19.41863098144531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,16,power_law_1.2,0.09787520170211791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,4,power_law_1.01,0.21571199893951415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,4,power_law_1.01,0.2344896078109741
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,16,power_law_1.2,0.09670400023460388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,4,power_law_1.01,0.25920639038085935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,4,power_law_1.01,0.31651840209960935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,16,power_law_1.2,0.09949439764022827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,4,power_law_1.01,0.3439552068710327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,2,balanced,0.04161066561937332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,2,balanced,0.043925335009892784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,4,power_law_1.01,0.4700607776641846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,2,balanced,0.047744000951449074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,2,balanced,0.06020266811052958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,2,balanced,0.08705600102742513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,2,balanced,0.08896533648173015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,2,balanced,0.09092799822489421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,2,balanced,0.09120532870292664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,2,balanced,0.0912000040213267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,2,balanced,0.09569600224494934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,2,balanced,0.09206400314966838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,2,balanced,0.09704533219337463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,2,balanced,0.09795733292897542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,2,balanced,0.09928533434867859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,2,balanced,0.10621333122253418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,2,balanced,0.10778666536013286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,16,power_law_1.2,0.10291839838027954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,2,balanced,0.11506133278210957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,4,power_law_1.01,0.5232448101043701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,2,balanced,0.12738133470217386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,2,balanced,0.1434719959894816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,2,balanced,0.17292799552281699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,2,balanced,0.19167466958363852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,2,balanced,0.2750506599744161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,2,balanced,0.328277329603831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,2,balanced,0.47140800952911377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,2,balanced,0.5997173388799032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,2,balanced,0.872058629989624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,4,power_law_1.01,0.718067216873169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,2,balanced,1.1337813536326091
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,2,balanced,1.395861307779948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,16,power_law_1.2,0.10734720230102539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,2,balanced,2.156714598337809
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,4,power_law_1.01,0.8799551963806153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,2,balanced,4.203664143880208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,16,power_law_1.2,0.10492160320281982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,4,power_law_1.01,1.219315242767334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,4,power_law_1.01,1.5353216171264648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,16,power_law_1.2,0.10814080238342286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,4,power_law_1.01,1.8705152511596679
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,16,power_law_1.2,0.11743359565734864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,4,balanced,0.0699893335501353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,4,balanced,0.0939573347568512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,4,balanced,0.11782933274904887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,4,balanced,0.16522666811943054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,4,balanced,0.2521119912465413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,4,balanced,0.3235573371251424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,4,balanced,0.32814399401346844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,4,balanced,0.33582401275634766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,4,balanced,0.33526933193206787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,4,balanced,0.34120531876881915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,4,balanced,0.34016533692677814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,16,power_law_1.2,0.12030719518661499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,4,power_law_1.01,2.9107072830200194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,4,balanced,0.3440320094426473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,4,balanced,0.3476639986038208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,4,balanced,0.35125335057576496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,4,balanced,0.3612266778945923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,4,balanced,0.3662293354670207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,4,balanced,0.3784373203913371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,4,balanced,0.4076106548309326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,4,balanced,0.42801066239674884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,4,balanced,0.5656799872716268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,4,balanced,0.5445333321889242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,4,balanced,0.7960853576660156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,16,power_law_1.2,0.12681599855422973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,4,balanced,0.7914400100708008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,4,balanced,1.1453973452250164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,4,balanced,1.2969600359598796
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,4,balanced,1.9016106923421223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,4,balanced,2.2836106618245444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,4,power_law_1.01,5.239616012573242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,16,power_law_1.2,0.1492735981941223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,1,4,balanced,2.814629236857096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,1,4,balanced,4.474778811136882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,16,power_law_1.2,0.15688320398330688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,1,4,balanced,8.553962707519531
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,2,power_law_1.01,0.14398080110549927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,16,power_law_1.2,0.2038719892501831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,2,power_law_1.01,0.23826560974121094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,16,power_law_1.2,0.22390398979187012
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,2,power_law_1.01,0.4333951950073242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,16,power_law_1.2,0.30078079700469973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,2,power_law_1.01,0.3105664014816284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,2,power_law_1.01,0.655289602279663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,16,power_law_1.2,0.31969280242919923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,2,power_law_1.01,0.9419584274291992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,16,power_law_1.2,0.42232961654663087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,2,power_law_1.01,1.1511167526245116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,1,power_law_1.2,0.03557760119438171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,1,power_law_1.2,0.054150402545928955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,1,power_law_1.2,0.0623744010925293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,2,power_law_1.01,1.5047295570373536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,16,power_law_1.2,0.5344384193420411
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,1,power_law_1.2,0.08011519908905029
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,1,power_law_1.2,0.10147199630737305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,1,power_law_1.2,0.17375359535217286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,2,power_law_1.01,1.525177574157715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,1,power_law_1.2,0.18807679414749146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,1,power_law_1.2,0.19642239809036255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,1,power_law_1.2,0.20355839729309083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,16,power_law_1.2,0.722047996520996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,2,power_law_1.01,1.5771519660949707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,1,power_law_1.2,0.20342400074005126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,1,power_law_1.2,0.2117055892944336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,1,power_law_1.2,0.21244800090789795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,2,power_law_1.01,1.688960075378418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,16,power_law_1.2,0.06959360241889953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,1,power_law_1.2,0.22198400497436524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,1,power_law_1.2,0.22279040813446044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,16,power_law_1.2,0.8930111885070801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,2,power_law_1.01,1.7334976196289062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,1,power_law_1.2,0.23398399353027344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,1,power_law_1.2,0.23935360908508302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,1,power_law_1.2,0.2535936117172241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,2,power_law_1.01,1.7861951828002929
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,1,power_law_1.2,0.29468159675598143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,16,power_law_1.2,0.06272640228271484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,1,power_law_1.2,0.3087807893753052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,1,power_law_1.2,0.3745408058166504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,2,power_law_1.01,1.9009599685668945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,4,16,power_law_1.2,1.2017600059509277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,1,power_law_1.2,0.4280704021453857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,1,power_law_1.2,0.5357376098632812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,2,power_law_1.01,1.90631046295166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,16,power_law_1.2,0.055769598484039305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,1,power_law_1.2,0.6452479839324952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,1,power_law_1.2,0.8484031677246093
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,2,power_law_1.01,1.9777791976928711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,1,power_law_1.2,1.0734592437744142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,16,power_law_1.2,0.05845119953155518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,1,power_law_1.2,1.5302911758422852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,4,16,power_law_1.2,1.6541631698608399
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,2,power_law_1.01,2.0358463287353517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,1,power_law_1.2,1.8084287643432617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,2,power_law_1.01,2.1844480514526365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,16,power_law_1.2,0.061926400661468504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,1,power_law_1.2,2.231110382080078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,2,power_law_1.01,2.3375295639038085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,16,power_law_1.2,0.07201279997825623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,1,power_law_1.2,3.4520702362060547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,2,power_law_1.01,2.5905536651611327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,2,power_law_1.01,2.6331903457641603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,16,power_law_1.2,0.07587839961051941
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,4,16,power_law_1.2,3.893695831298828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,1,power_law_1.2,6.643724822998047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,2,power_law_1.01,3.15347843170166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,16,power_law_1.2,0.07854080200195312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,2,power_law_1.01,2.9834943771362306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,16,power_law_1.2,0.07871999740600585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,2,power_law_1.01,3.4897727966308594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,16,power_law_1.2,0.07971199750900268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,2,power_law_1.01,4.125049591064453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,16,power_law_1.2,0.08203520178794861
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,2,power_law_1.01,5.1290241241455075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,16,power_law_1.2,0.08482559919357299
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,16,power_law_1.2,0.08678399920463561
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,32,8,balanced,0.05225066840648651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,32,8,balanced,0.054757331808408104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,2,power_law_1.01,6.1135295867919925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,32,8,balanced,0.05253333350022634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,32,8,balanced,0.05689600110054016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,32,8,balanced,0.06267199913660686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,32,8,balanced,0.07260799904664357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,32,8,balanced,0.07332266867160797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,32,8,balanced,0.0746506651242574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,32,8,balanced,0.07328533132870992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,32,8,balanced,0.07512533167997996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,32,8,balanced,0.07665066421031952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,32,8,balanced,0.07754666606585185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,32,8,balanced,0.07708266874154408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,32,8,balanced,0.07879466811815898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,32,8,balanced,0.0851093331972758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,32,8,balanced,0.08704533179601033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,32,8,balanced,0.09077866872151692
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,32,8,balanced,0.10087466239929199
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,32,8,balanced,0.10828266541163127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,32,8,balanced,0.12744533022244772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,32,8,balanced,0.1540000041325887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,32,8,balanced,0.18600533405939737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,2,power_law_1.01,0.06231039762496948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,32,8,balanced,0.21618133783340454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,32,8,balanced,0.3022186756134033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,16,power_law_1.2,0.09019520282745361
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,32,8,balanced,0.36353600025177
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,32,8,balanced,0.521717349688212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,32,8,balanced,0.6690719922383627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,32,8,balanced,0.8241546948750814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,32,8,balanced,1.273045301437378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,2,power_law_1.01,0.07292159795761108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,32,8,balanced,2.499226729075114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,2,power_law_1.01,7.143475341796875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,2,power_law_1.01,0.07324159741401673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,16,power_law_1.2,0.09468799829483032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,2,power_law_1.01,0.08956159949302674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,2,power_law_1.01,0.11573120355606079
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,16,power_law_1.2,0.10065280199050904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,2,power_law_1.01,0.13897600173950195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,2,power_law_1.01,9.123359680175781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,2,power_law_1.01,0.16891520023345946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,2,power_law_1.01,0.17891839742660523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,16,power_law_1.2,0.10734080076217652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,2,power_law_1.01,0.18540159463882447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,2,power_law_1.01,0.1950271964073181
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,16,power_law_1.2,0.11819519996643066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,2,power_law_1.01,0.19757440090179443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,2,power_law_1.01,0.2082495927810669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,16,power_law_1.2,0.13115520477294923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,2,power_law_1.01,0.21191680431365967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,2,power_law_1.01,0.22005760669708252
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,2,power_law_1.01,18.557286071777344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,2,power_law_1.01,0.22760961055755616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,16,power_law_1.2,0.1641088008880615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,4,power_law_1.01,0.03587839901447296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,2,power_law_1.01,0.23450241088867188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,4,power_law_1.01,0.0301503986120224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,4,power_law_1.01,0.029625600576400755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,2,power_law_1.01,0.24710400104522706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,4,power_law_1.01,0.031865599751472476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,16,power_law_1.2,0.19771519899368287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,2,power_law_1.01,0.26831998825073244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,4,power_law_1.01,0.03328000009059906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,4,power_law_1.01,0.036032000184059144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,2,power_law_1.01,0.2967744112014771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,4,power_law_1.01,0.03619840145111084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,4,power_law_1.01,0.03713920116424561
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,2,power_law_1.01,0.34401280879974366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,16,power_law_1.2,0.26248319149017335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,4,power_law_1.01,0.03708159923553467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,4,power_law_1.01,0.03776000142097473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,2,power_law_1.01,0.39687678813934324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,4,power_law_1.01,0.03966720104217529
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,4,power_law_1.01,0.04190720021724701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,2,power_law_1.01,0.4904319763183594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,4,power_law_1.01,0.04360319972038269
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,16,power_law_1.2,0.27777919769287107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,4,power_law_1.01,0.04492799937725067
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,2,power_law_1.01,0.6139008045196533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,4,power_law_1.01,0.05064319968223572
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,4,power_law_1.01,0.04929920136928558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,2,power_law_1.01,0.8271488189697266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,4,power_law_1.01,0.05514879822731018
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,16,power_law_1.2,0.39210240840911864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,4,power_law_1.01,0.06193280220031738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,2,power_law_1.01,1.0286335945129395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,4,power_law_1.01,0.07185919880867005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,4,power_law_1.01,0.08097280263900757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,32,balanced,0.054229333996772766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,32,balanced,0.04788800080617269
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,32,balanced,0.04553066690762838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,32,balanced,0.051445335149765015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,4,power_law_1.01,0.0952127993106842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,32,balanced,0.06932266553243001
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,32,balanced,0.09710400303204854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,2,power_law_1.01,1.4735103607177735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,32,balanced,0.09755200147628784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,32,balanced,0.09795199831326802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,32,balanced,0.09694400429725647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,32,balanced,0.09945066769917806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,32,balanced,0.09902933239936829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,32,balanced,0.10100799798965454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,32,balanced,0.10068800052007039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,32,balanced,0.10271466771761577
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,32,balanced,0.10683199763298035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,32,balanced,0.10700266559918721
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,4,power_law_1.01,0.1241536021232605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,32,balanced,0.10931199789047241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,32,balanced,0.11560533444086711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,32,balanced,0.11768533786137898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,32,balanced,0.12944533427556357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,32,balanced,0.13900799552599588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,32,balanced,0.15807466705640158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,16,power_law_1.2,0.4438528060913086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,32,balanced,0.1771626671155294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,32,balanced,0.21449067195256552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,32,balanced,0.2563946644465129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,4,power_law_1.01,0.17263360023498536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,32,balanced,0.34107200304667157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,32,balanced,0.4243786732355754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,32,balanced,0.5123146772384644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,2,power_law_1.01,1.8405183792114257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,32,balanced,0.7283360163370768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,4,power_law_1.01,0.23301119804382325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,32,balanced,1.3782240549723308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,4,power_law_1.01,0.2772608041763306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,2,power_law_1.01,2.2819583892822264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,4,power_law_1.01,0.4241983890533447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,16,power_law_1.2,0.66212477684021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,4,power_law_1.01,0.5447423934936524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,2,power_law_1.01,3.5012351989746096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,4,power_law_1.01,0.6000256061553955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,16,power_law_1.2,0.8176511764526367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,4,power_law_1.01,0.9937151908874512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,2,power_law_1.01,7.005222320556641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,4,power_law_1.01,1.9830720901489258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,8,16,power_law_1.2,0.9895039558410644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,64,balanced,0.08472533027331035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,64,balanced,0.08734400073687236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,64,balanced,0.0851146678129832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,64,balanced,0.08506133159001668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,64,balanced,0.09718400239944458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,64,balanced,0.12731200456619263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,64,balanced,0.16941332817077637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,64,balanced,0.16521066427230835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,64,balanced,0.16292267044385275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,64,balanced,0.17228800058364868
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,64,balanced,0.17506666978200278
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,64,balanced,0.17499732971191406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,64,balanced,0.16941332817077637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,64,balanced,0.17897067467371622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,64,balanced,0.17471466461817423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,64,balanced,0.17228267590204874
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,64,balanced,0.18743999799092612
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,64,balanced,0.18175466855367026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,64,balanced,0.19364800055821738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,64,balanced,0.20582934220631918
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,64,balanced,0.21123733123143515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,64,balanced,0.2360746661822001
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,64,balanced,0.25840532779693604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,64,balanced,0.3107146620750427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,64,balanced,0.4095733165740967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,64,balanced,0.48394668102264404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,8,16,power_law_1.2,1.6143936157226562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,64,balanced,0.6201866865158081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,64,balanced,0.8210773468017578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,64,balanced,1.1928213437398274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,64,balanced,2.1764639218648276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,16,8,balanced,0.038880000511805214
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,16,8,balanced,0.03779733429352442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,16,8,balanced,0.03543466577927271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,16,8,balanced,0.03467733412981033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,16,8,balanced,0.03544000039498011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,16,8,balanced,0.0356480007370313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,16,8,balanced,0.037530665596326195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,16,8,balanced,0.03689600030581156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,16,8,balanced,0.035631999373435974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,16,8,balanced,0.03755199909210205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,16,8,balanced,0.03762666632731756
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,16,8,balanced,0.03807466725508372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,16,8,balanced,0.03759466608365377
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,16,8,balanced,0.037632000943024956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,16,8,balanced,0.043696001172065735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,16,8,balanced,0.04277333120505015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,16,8,balanced,0.043749332427978516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,16,8,balanced,0.047557334105173744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,16,8,balanced,0.04976533353328705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,16,8,balanced,0.05291733145713806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,16,8,balanced,0.0620959997177124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,8,16,power_law_1.2,3.6607742309570312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,16,8,balanced,0.0828959991534551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,16,8,balanced,0.08933867017428081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,16,8,balanced,0.11556266744931538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,16,8,balanced,0.1327786644299825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,16,8,balanced,0.18144534031550089
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,16,8,balanced,0.22869332631429037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,16,8,balanced,0.2725653251012166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,16,8,balanced,0.3986453215281169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,2,power_law_1.2,0.06922879815101624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,16,8,balanced,0.7470080057779948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,2,power_law_1.2,0.0759168028831482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,2,power_law_1.2,0.0729856014251709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,2,power_law_1.2,0.09844480156898498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,4,power_law_1.01,0.0586624026298523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,4,power_law_1.01,0.06993280053138733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,2,power_law_1.2,0.12821120023727417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,4,power_law_1.01,0.09592959880828858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,2,power_law_1.2,0.1932479977607727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,4,power_law_1.01,0.07127040028572082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,2,power_law_1.2,0.20665600299835205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,4,power_law_1.01,0.09339519739151
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,4,8,power_law_1.2,0.031699201464653014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,4,power_law_1.01,0.10978560447692871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,2,power_law_1.2,0.2189568042755127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,4,8,power_law_1.2,0.03178240060806274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,4,8,power_law_1.2,0.031033599376678468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,4,8,power_law_1.2,0.030803200602531434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,4,8,power_law_1.2,0.031251201033592226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,4,8,power_law_1.2,0.03288320004940033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,4,8,power_law_1.2,0.03754880130290985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,4,8,power_law_1.2,0.03604480028152466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,4,power_law_1.01,0.11794559955596924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,4,8,power_law_1.2,0.04081279933452606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,2,power_law_1.2,0.2292543888092041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,4,8,power_law_1.2,0.04032639861106872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,4,8,power_law_1.2,0.03835520148277283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,4,8,power_law_1.2,0.04337919950485229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,4,8,power_law_1.2,0.04447360038757324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,4,8,power_law_1.2,0.04523519873619079
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,4,8,power_law_1.2,0.054092800617218016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,4,power_law_1.01,0.14220160245895386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,64,power_law_1.01,0.06382719874382019
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,4,8,power_law_1.2,0.05857279896736145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,4,8,power_law_1.2,0.06094719767570496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,4,8,power_law_1.2,0.07607679963111877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,2,power_law_1.2,0.2340991973876953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,4,8,power_law_1.2,0.08097280263900757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,4,8,power_law_1.2,0.10437120199203491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,4,8,power_law_1.2,0.13020800352096557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,4,power_law_1.01,0.1433087944984436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,4,8,power_law_1.2,0.17063039541244507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,4,8,power_law_1.2,0.20325760841369628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,4,8,power_law_1.2,0.2851263999938965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,4,8,power_law_1.2,0.3405951976776123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,2,power_law_1.2,0.2447808027267456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,4,power_law_1.01,0.1480191946029663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,4,8,power_law_1.2,0.5081600189208985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,4,8,power_law_1.2,0.677126407623291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,4,8,power_law_1.2,0.8282815933227539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,64,power_law_1.01,0.0692799985408783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,4,power_law_1.01,0.15141760110855101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,2,power_law_1.2,0.24977281093597412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,4,8,power_law_1.2,1.245139217376709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,4,power_law_1.01,0.16098560094833375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,4,8,power_law_1.2,2.4153024673461916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,2,power_law_1.2,0.2618367910385132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,64,power_law_1.01,0.055193597078323366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,4,power_law_1.01,0.1657855987548828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,2,power_law_1.2,0.2720832109451294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,4,power_law_1.01,0.16714240312576295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,2,power_law_1.2,0.28513920307159424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,64,power_law_1.01,0.05663999915122986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,4,power_law_1.01,0.1736191987991333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,4,power_law_1.01,0.1795904040336609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,2,power_law_1.2,0.29786880016326905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,64,power_law_1.01,0.05860480070114136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,4,power_law_1.01,0.19324159622192383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,2,power_law_1.2,0.3082495927810669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,4,power_law_1.01,0.20595839023590087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,64,power_law_1.01,0.06112639904022217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,4,power_law_1.01,0.2298815965652466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,2,power_law_1.2,0.3608000040054321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,4,power_law_1.01,0.2658368110656738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,2,power_law_1.2,0.39661440849304197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,64,power_law_1.01,0.06181120276451111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,4,power_law_1.01,0.3047679901123047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,1,8,power_law_1.2,0.02807680070400238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,2,power_law_1.2,0.4896063804626465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,1,8,power_law_1.2,0.02810240089893341
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,4,power_law_1.01,0.376531195640564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,64,power_law_1.01,0.061343997716903687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,1,8,power_law_1.2,0.030086401104927062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,1,8,power_law_1.2,0.0332863986492157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,4,power_law_1.01,0.44249601364135743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,2,power_law_1.2,0.5999616146087646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,1,8,power_law_1.2,0.03438720107078552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,1,8,power_law_1.2,0.04679040014743805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,1,8,power_law_1.2,0.04856959879398346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,64,power_law_1.01,0.061440002918243405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,4,power_law_1.01,0.5928063869476319
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,1,8,power_law_1.2,0.04748800098896026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,2,power_law_1.2,0.779040002822876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,1,8,power_law_1.2,0.05026559829711914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,1,8,power_law_1.2,0.05149440169334411
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,4,power_law_1.01,0.7627071857452392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,1,8,power_law_1.2,0.05221760272979736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,64,power_law_1.01,0.062015998363494876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,1,8,power_law_1.2,0.05186560153961182
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,2,power_law_1.2,0.9130240440368652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,1,8,power_law_1.2,0.05381119847297668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,1,8,power_law_1.2,0.056601601839065555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,4,power_law_1.01,1.0721216201782227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,1,8,power_law_1.2,0.06268799901008607
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,64,power_law_1.01,0.06353920102119445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,1,8,power_law_1.2,0.06382719874382019
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,1,8,power_law_1.2,0.06496000289916992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,2,power_law_1.2,1.2240896224975586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,4,power_law_1.01,1.3574527740478515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,1,8,power_law_1.2,0.07457919716835022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,8,power_law_1.2,0.07820159792900086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,64,power_law_1.01,0.06639360189437866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,8,power_law_1.2,0.08479359745979309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,8,power_law_1.2,0.09631360173225403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,4,power_law_1.01,1.631667137145996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,2,power_law_1.2,1.4834176063537599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,8,power_law_1.2,0.11564799547195434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,8,power_law_1.2,0.14725760221481324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,64,power_law_1.01,0.06722559928894042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,8,power_law_1.2,0.19995520114898682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,8,power_law_1.2,0.23247361183166504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,4,power_law_1.01,2.6714111328125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,2,power_law_1.2,2.142304039001465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,64,power_law_1.01,0.06706560254096985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,8,power_law_1.2,0.40746240615844725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,8,power_law_1.2,0.4652671813964844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,64,power_law_1.01,0.0720575988292694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,8,power_law_1.2,0.5418560028076171
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,2,power_law_1.2,2.8201791763305666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,4,power_law_1.01,4.958214569091797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,8,power_law_1.2,0.9477631568908691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,64,power_law_1.01,0.07425919771194459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,64,power_law_1.01,0.0804095983505249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,8,power_law_1.2,2.0740928649902344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,8,2,power_law_1.2,3.255206298828125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,64,power_law_1.01,0.09477120041847228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,64,power_law_1.01,0.10654720067977905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,8,2,power_law_1.2,5.2734016418457035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,64,power_law_1.01,0.13325439691543578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,64,power_law_1.01,0.16293760538101196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,16,balanced,0.03213333338499069
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,16,balanced,0.03182933231194814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,16,balanced,0.0320266659061114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,16,balanced,0.03379199902216593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,16,balanced,0.03369600077470144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,16,balanced,0.03583999971548716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,16,balanced,0.036415999134381614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,16,balanced,0.036714665591716766
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,16,balanced,0.038346665600935616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,16,balanced,0.035936000446478523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,16,balanced,0.03775466730197271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,16,balanced,0.037733333806196846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,16,balanced,0.03814399987459183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,16,balanced,0.03979733337958654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,16,balanced,0.04055999964475632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,16,balanced,0.04404266675313314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,16,balanced,0.04390933116277059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,16,balanced,0.04610666632652283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,16,balanced,0.0498933345079422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,16,balanced,0.056554665168126426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,16,balanced,0.062496001521746315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,16,balanced,0.08265066643555959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,16,balanced,0.10570133725802104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,16,balanced,0.14300266901652017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,16,balanced,0.18118933836619058
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,16,balanced,0.24426132440567017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,64,power_law_1.01,0.21524479389190673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,16,balanced,0.30794666210810345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,16,balanced,0.3736639817555745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,16,balanced,0.5657653411229452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,16,balanced,1.0893920262654622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,8,power_law_1.01,0.05114240050315857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,8,power_law_1.01,0.06440320014953613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,64,power_law_1.01,0.22860159873962402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,8,power_law_1.01,0.049753600358963014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,8,power_law_1.01,0.053855997323989865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,8,2,power_law_1.2,10.095993804931641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,8,power_law_1.01,0.058297598361968996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,8,power_law_1.01,0.07217919826507568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,64,power_law_1.01,0.3023423910140991
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,8,power_law_1.01,0.0790719985961914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,8,power_law_1.01,0.07824640274047852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,8,power_law_1.01,0.08104959726333619
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,8,power_law_1.01,0.07996159791946411
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,8,power_law_1.01,0.08039039969444275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,64,power_law_1.01,0.3713023900985718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,8,power_law_1.01,0.0861952006816864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,8,power_law_1.01,0.08526080250740051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,8,power_law_1.01,0.08910080194473266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,8,power_law_1.01,0.09454079866409301
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,8,power_law_1.01,0.09969919919967651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,64,power_law_1.01,0.5332736015319824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,8,power_law_1.01,0.10316799879074097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,8,power_law_1.01,0.11448320150375366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,8,power_law_1.01,0.12475520372390747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,8,power_law_1.01,0.1538815975189209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,256,power_law_1.01,0.06241919994354248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,8,power_law_1.01,0.16977920532226562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,64,power_law_1.01,0.6514815807342529
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,256,power_law_1.01,0.06311039924621582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,8,power_law_1.01,0.21298561096191407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,8,power_law_1.01,0.2543872117996216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,256,power_law_1.01,0.06308479905128479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,1,power_law_1.01,0.049516800045967105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,8,power_law_1.01,0.34092159271240235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,256,power_law_1.01,0.06264960169792175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,1,power_law_1.01,0.054764801263809205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,8,power_law_1.01,0.42749438285827634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,256,power_law_1.01,0.061267197132110596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,1,power_law_1.01,0.0645632028579712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,4,64,power_law_1.01,0.8519040107727051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,8,power_law_1.01,0.5406720161437988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,256,power_law_1.01,0.06252800226211548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,1,power_law_1.01,0.0945792019367218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,256,power_law_1.01,0.06416000127792358
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,8,power_law_1.01,0.7030655860900878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,1,power_law_1.01,0.12861440181732178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,256,power_law_1.01,0.06306560039520263
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,8,power_law_1.01,0.8976127624511718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,1,power_law_1.01,0.15783040523529052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,256,power_law_1.01,0.06454399824142457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,1,power_law_1.01,0.21175680160522461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,256,power_law_1.01,0.06550400257110596
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,8,power_law_1.01,1.4851712226867675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,1,power_law_1.01,0.22641279697418212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,256,power_law_1.01,0.06476160287857055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,4,64,power_law_1.01,1.2614912033081054
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,1,power_law_1.01,0.24246399402618407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,256,power_law_1.01,0.06603519916534424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,1,power_law_1.01,0.24483840465545653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,256,power_law_1.01,0.06607360243797303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,1,power_law_1.01,0.25437440872192385
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,8,power_law_1.01,2.5007808685302733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,256,power_law_1.01,0.06564480066299438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,1,power_law_1.01,0.2658112049102783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,256,power_law_1.01,0.06818559765815735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,1,power_law_1.01,0.2693567991256714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,256,power_law_1.01,0.06880639791488648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,1,power_law_1.01,0.2853568077087402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,256,power_law_1.01,0.07068799734115601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,256,power_law_1.01,0.08584960103034973
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,1,power_law_1.01,0.2985536098480225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,256,power_law_1.01,0.09390079975128174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,1,power_law_1.01,0.30335359573364257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,4,64,power_law_1.01,2.778169631958008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,256,power_law_1.01,0.11930240392684936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,1,power_law_1.01,0.3286272048950195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,256,power_law_1.01,0.14268800020217895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,1,power_law_1.01,0.380511999130249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,1,power_law_1.01,0.4119296073913574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,256,power_law_1.01,0.19007999897003175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,1,power_law_1.01,0.4988096237182617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,256,power_law_1.01,0.2344575881958008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,1,power_law_1.01,0.5756095886230469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,256,power_law_1.01,0.3172991991043091
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,1,power_law_1.01,0.731001615524292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,256,power_law_1.01,0.3988735914230347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,1,power_law_1.01,0.9077247619628906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,256,power_law_1.01,0.583244800567627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,16,power_law_1.01,0.07461760044097901
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,16,power_law_1.01,0.08228480219841003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,1,power_law_1.01,1.2043840408325195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,16,power_law_1.01,0.052825599908828735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,256,power_law_1.01,0.7569856166839599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,16,power_law_1.01,0.05985280275344849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,1,power_law_1.01,1.4969471931457519
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,16,power_law_1.01,0.06464639902114869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,256,power_law_1.01,0.9148799896240234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,16,power_law_1.01,0.07634559869766236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,16,power_law_1.01,0.07985280156135559
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,1,power_law_1.01,2.0985343933105467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,16,power_law_1.01,0.07848960161209106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,256,power_law_1.01,1.4424960136413574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,16,power_law_1.01,0.08030080199241638
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,16,power_law_1.01,0.08325759768486023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,1,power_law_1.01,2.713702392578125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,16,power_law_1.01,0.08284159898757934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,16,power_law_1.01,0.0851967990398407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,2,power_law_1.2,0.05025920271873474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,16,power_law_1.01,0.08687360286712646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,1,power_law_1.01,3.29290885925293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,16,power_law_1.01,0.09008640050888062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,256,power_law_1.01,2.998931121826172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,16,power_law_1.01,0.09909120202064514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,2,power_law_1.2,0.07155839800834655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,16,power_law_1.01,0.10227839946746826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,16,power_law_1.01,0.11042560338973999
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,1,power_law_1.01,5.080588912963867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,16,power_law_1.01,0.12846720218658447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,2,power_law_1.2,0.061504000425338747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,16,power_law_1.01,0.1292799949645996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,16,power_law_1.01,0.15843839645385743
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,2,power_law_1.2,0.08163840174674988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,16,power_law_1.01,0.18035199642181396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,16,power_law_1.01,0.23661439418792723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,2,power_law_1.2,0.10311039686203002
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,16,power_law_1.01,0.3129791975021362
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,16,power_law_1.01,0.39246718883514403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,1,power_law_1.01,9.720819091796875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,2,power_law_1.2,0.1317247986793518
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,16,power_law_1.01,0.4725056171417236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,2,power_law_1.2,0.1861567974090576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,16,power_law_1.01,0.6395008087158203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,16,power_law_1.01,0.8615936279296875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,2,power_law_1.2,0.19882880449295043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,16,power_law_1.01,1.0496383666992188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,2,power_law_1.2,0.20309760570526122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,1,4,power_law_1.01,0.11908479928970336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,1,4,power_law_1.01,0.13967360258102418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,1,4,power_law_1.01,0.12012799978256225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,1,4,power_law_1.01,0.12414079904556274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,1,4,power_law_1.01,0.12583039999008178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,2,power_law_1.2,0.2084223985671997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,1,4,power_law_1.01,0.12869759798049926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,1,4,power_law_1.01,0.13437440395355224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,16,power_law_1.01,1.678816032409668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,1,4,power_law_1.01,0.1344704031944275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,1,4,power_law_1.01,0.14156800508499146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,1,4,power_law_1.01,0.15871360301971435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,1,4,power_law_1.01,0.15646079778671265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,1,4,power_law_1.01,0.15773439407348633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,2,power_law_1.2,0.21386239528656006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,1,4,power_law_1.01,0.16064000129699707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,1,4,power_law_1.01,0.193996798992157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,1,4,power_law_1.01,0.19251199960708618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,1,4,power_law_1.01,0.22865281105041504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,1,4,power_law_1.01,0.228108811378479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,1,4,power_law_1.01,0.2571903944015503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,2,power_law_1.2,0.2246783971786499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,1,4,power_law_1.01,0.33731200695037844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,1,4,power_law_1.01,0.4883903980255127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,1,4,power_law_1.01,0.56179838180542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,16,power_law_1.01,3.020044708251953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,1,4,power_law_1.01,0.749132776260376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,2,power_law_1.2,0.2351680040359497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,1,4,power_law_1.01,0.9117055892944336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,1,4,power_law_1.01,1.3050559997558593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,2,power_law_1.2,0.25063679218292234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,1,4,power_law_1.01,2.4811775207519533
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,16,2,balanced,0.031850665807724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,16,2,balanced,0.03331200033426285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,16,2,balanced,0.03329066683848699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,16,2,balanced,0.03335466732581457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,16,2,balanced,0.0334346666932106
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,16,2,balanced,0.03551999976237615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,16,2,balanced,0.035589332381884255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,16,2,balanced,0.03366400053103765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,16,2,balanced,0.035717333356539406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,16,2,balanced,0.035487999518712364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,16,2,balanced,0.037845333417256675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,16,2,balanced,0.037477334340413414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,1,4,power_law_1.01,2.956870460510254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,16,2,balanced,0.03782933453718821
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,16,2,balanced,0.0382080003619194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,2,power_law_1.2,0.2657023906707764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,16,2,balanced,0.04752533137798309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,16,2,balanced,0.0476746658484141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,16,2,balanced,0.04709866642951965
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,16,2,balanced,0.06002666552861532
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,16,2,balanced,0.06406933565934499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,16,2,balanced,0.08201600114504497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,16,2,balanced,0.09758399923642476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,16,2,balanced,0.13489066561063132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,16,2,balanced,0.15497600038846335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,16,2,balanced,0.20800000429153442
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,16,2,balanced,0.24230400721232095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,16,2,balanced,0.35550399621327716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,16,2,balanced,0.46166400114695233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,1,4,power_law_1.01,3.979788970947266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,16,2,balanced,0.5724266767501831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,2,power_law_1.2,0.2761087894439697
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,16,2,balanced,0.8672107060750326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,16,2,balanced,1.675264040629069
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,1,4,power_law_1.01,4.953030395507812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,2,power_law_1.2,0.28819200992584226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,2,power_law_1.2,0.3216896057128906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,1,4,power_law_1.01,8.253421020507812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,2,power_law_1.2,0.3460416078567505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,2,power_law_1.2,0.3986560106277466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,2,power_law_1.2,0.47530241012573243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,1,4,power_law_1.01,15.914009094238281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,2,power_law_1.2,0.5718719959259033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,2,power_law_1.2,0.6933887958526611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,64,power_law_1.2,0.10032639503479004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,64,power_law_1.2,0.09595519900321961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,2,power_law_1.2,0.9227840423583984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,64,power_law_1.2,0.08814719915390015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,2,power_law_1.2,1.12161283493042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,16,power_law_1.2,0.06707199811935424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,64,power_law_1.2,0.06735360026359558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,16,power_law_1.2,0.06308479905128479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,2,power_law_1.2,1.5074048042297363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,16,power_law_1.2,0.052716797590255736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,64,power_law_1.2,0.07756159901618957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,16,power_law_1.2,0.055776000022888184
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,16,power_law_1.2,0.05715199708938599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,64,power_law_1.2,0.07770879864692688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,2,power_law_1.2,1.8276479721069336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,16,power_law_1.2,0.06832640171051026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,16,power_law_1.2,0.06945279836654664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,64,power_law_1.2,0.07948160171508789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,16,power_law_1.2,0.07317119836807251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,16,power_law_1.2,0.07477759718894958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,2,power_law_1.2,2.202662467956543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,64,power_law_1.2,0.08936960101127625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,16,power_law_1.2,0.07448959946632386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,16,power_law_1.2,0.07749760150909424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,64,power_law_1.2,0.0986303985118866
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,1,power_law_1.01,0.07573760151863099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,16,power_law_1.2,0.07728639841079712
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,16,power_law_1.2,0.07839999794960022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,64,power_law_1.2,0.09422720074653626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,2,power_law_1.2,3.3272640228271486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,1,power_law_1.01,0.10870399475097656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,16,power_law_1.2,0.08145920038223267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,1,power_law_1.01,0.10848640203475952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,16,power_law_1.2,0.08625919818878174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,64,power_law_1.2,0.09352959990501404
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,16,power_law_1.2,0.0887935996055603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,1,power_law_1.01,0.16751999855041505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,64,power_law_1.2,0.09838079810142517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,16,power_law_1.2,0.09450240135192871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,1,power_law_1.01,0.233459210395813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,16,power_law_1.2,0.1081279993057251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,64,power_law_1.2,0.0944703996181488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,1,power_law_1.01,0.30451838970184325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,16,power_law_1.2,0.12025599479675293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,2,power_law_1.2,6.276076889038086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,16,power_law_1.2,0.14825600385665894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,1,power_law_1.01,0.418233585357666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,64,power_law_1.2,0.09236479997634887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,16,power_law_1.2,0.1538815975189209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,1,power_law_1.01,0.43736958503723145
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,16,power_law_1.2,0.20993919372558595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,64,power_law_1.2,0.10524159669876099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,1,power_law_1.01,0.44879999160766604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,16,power_law_1.2,0.24766719341278076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,64,power_law_1.2,0.10472960472106933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,16,power_law_1.2,0.31541121006011963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,1,power_law_1.01,0.4689727783203125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,16,power_law_1.2,0.46362881660461425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,1,power_law_1.01,0.48732800483703614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,64,power_law_1.2,0.10660480260848999
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,1,power_law_1.01,0.5028863906860351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,16,power_law_1.2,0.5610943794250488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,64,power_law_1.2,0.12001919746398926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,1,power_law_1.01,0.505299186706543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,16,power_law_1.2,0.7467455863952637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,64,power_law_1.2,0.12879999876022338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,1,power_law_1.01,0.5345215797424316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,16,power_law_1.2,1.1353792190551757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,1,power_law_1.01,0.5550015926361084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,64,power_law_1.2,0.14785280227661132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,1,power_law_1.01,0.5739136219024659
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,64,power_law_1.2,0.18691840171813964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,16,power_law_1.2,1.59683198928833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,1,power_law_1.01,0.6104703903198242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,64,power_law_1.2,0.23236479759216308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,1,power_law_1.01,0.6880767822265625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,1,power_law_1.01,0.7538112163543701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,64,power_law_1.2,0.2796096086502075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,16,power_law_1.2,3.197433662414551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,1,power_law_1.01,0.8871487617492676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,8,2,power_law_1.2,0.04347519874572754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,64,power_law_1.2,0.3669696092605591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,1,power_law_1.01,0.916256046295166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,8,2,power_law_1.2,0.04297600090503693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,64,power_law_1.2,0.46638078689575196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,1,power_law_1.01,1.122201633453369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,8,2,power_law_1.2,0.04602240025997162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,1,power_law_1.01,1.3636544227600098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,8,2,power_law_1.2,0.05173119902610779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,64,power_law_1.2,0.7104703903198242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,8,2,power_law_1.2,0.05610880255699158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,1,power_law_1.01,1.7673343658447265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,8,2,power_law_1.2,0.07149440050125122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,64,power_law_1.2,0.8846783638000488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,8,2,power_law_1.2,0.07376639842987061
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,1,power_law_1.01,2.1732351303100588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,8,2,power_law_1.2,0.07553920149803162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,64,power_law_1.2,1.1758336067199706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,8,2,power_law_1.2,0.07719039916992188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,1,power_law_1.01,2.969759941101074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,8,2,power_law_1.2,0.07740799784660339
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,8,2,power_law_1.2,0.07985280156135559
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,1,power_law_1.01,3.8757503509521483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,64,power_law_1.2,1.8296512603759765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,8,2,power_law_1.2,0.0843775987625122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,8,2,power_law_1.2,0.08787840008735656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,8,2,power_law_1.2,0.08731520175933838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,1,power_law_1.01,4.6835071563720705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,8,2,power_law_1.2,0.09610880017280579
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,8,2,power_law_1.2,0.10332800149917602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,2,power_law_1.2,0.12088960409164429
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,64,power_law_1.2,3.938822555541992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,8,2,power_law_1.2,0.11173759698867798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,8,2,power_law_1.2,0.12770559787750244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,1,power_law_1.01,6.961151885986328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,8,2,power_law_1.2,0.14457600116729735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,2,power_law_1.2,0.12369920015335083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,8,2,power_law_1.2,0.17278720140457154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,8,2,power_law_1.2,0.21650559902191163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,2,power_law_1.2,0.14624639749526977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,8,2,power_law_1.2,0.28008959293365476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,8,2,power_law_1.2,0.35526399612426757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,2,power_law_1.2,0.2354111909866333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,8,2,power_law_1.2,0.5049407958984375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,1,power_law_1.01,13.450189208984375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,2,power_law_1.2,0.32214400768280027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,8,2,power_law_1.2,0.6795839786529541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,8,2,power_law_1.2,0.9904191970825196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,2,power_law_1.2,0.5844223976135254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,8,2,power_law_1.2,1.2895039558410644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,2,power_law_1.2,0.6363135814666748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,8,2,power_law_1.2,1.60199031829834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,32,power_law_1.2,0.11915520429611207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,2,power_law_1.2,0.6842175960540772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,32,power_law_1.2,0.15201280117034913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,8,2,power_law_1.2,2.8202943801879883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,2,power_law_1.2,0.7084735870361328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,32,power_law_1.2,0.12551039457321167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,32,power_law_1.2,0.07232000231742859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,2,power_law_1.2,0.7243264198303223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,32,power_law_1.2,0.08608639836311341
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,8,2,power_law_1.2,4.9079231262207035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,2,power_law_1.2,0.7607744216918946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,4,4,power_law_1.01,0.04222719967365265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,32,power_law_1.2,0.09667840003967285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,4,4,power_law_1.01,0.03689599931240082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,2,power_law_1.2,0.7894591808319091
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,4,4,power_law_1.01,0.03534719944000244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,32,power_law_1.2,0.1097983956336975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,4,4,power_law_1.01,0.03820799887180328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,4,4,power_law_1.01,0.04028159976005554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,32,power_law_1.2,0.12054400444030762
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,2,power_law_1.2,0.8169535636901856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,4,4,power_law_1.01,0.0435263991355896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,4,4,power_law_1.01,0.04537599980831146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,4,4,power_law_1.01,0.045433598756790164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,32,power_law_1.2,0.11429120302200317
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,4,4,power_law_1.01,0.0455808013677597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,2,power_law_1.2,0.845849609375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,4,4,power_law_1.01,0.04584319889545441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,32,power_law_1.2,0.11016319990158081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,4,4,power_law_1.01,0.04693120121955872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,4,4,power_law_1.01,0.04962559938430786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,2,power_law_1.2,0.8678655624389648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,32,power_law_1.2,0.1192639946937561
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,4,4,power_law_1.01,0.04903680086135864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,4,4,power_law_1.01,0.0538752019405365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,4,4,power_law_1.01,0.05750399827957153
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,32,power_law_1.2,0.12302720546722412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,4,4,power_law_1.01,0.05961599946022034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,2,power_law_1.2,0.8769599914550781
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,4,4,power_law_1.01,0.06004480123519897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,32,power_law_1.2,0.12609280347824098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,4,4,power_law_1.01,0.06898559927940369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,4,4,power_law_1.01,0.07804160118103028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,2,power_law_1.2,0.9399168014526367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,32,power_law_1.2,0.12180479764938354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,4,4,power_law_1.01,0.09573760032653808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,32,1,balanced,0.033728001018365227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,32,1,balanced,0.032730666299661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,32,1,balanced,0.03669866671164831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,4,4,power_law_1.01,0.10728960037231446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,32,1,balanced,0.03557866563399633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,32,1,balanced,0.036943999429543815
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,32,1,balanced,0.037045332292715706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,32,1,balanced,0.038362666964530945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,32,1,balanced,0.03724266588687897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,32,1,balanced,0.038634667793909706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,32,power_law_1.2,0.13171839714050293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,32,1,balanced,0.03957333415746689
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,4,4,power_law_1.01,0.13843200206756592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,32,1,balanced,0.03982933362325033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,32,1,balanced,0.04073066761096319
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,32,1,balanced,0.040448000033696495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,32,1,balanced,0.04493333399295807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,32,1,balanced,0.05287999908129374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,32,1,balanced,0.051632001996040344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,32,1,balanced,0.05483733117580414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,2,power_law_1.2,1.0701696395874023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,4,4,power_law_1.01,0.1676800012588501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,32,1,balanced,0.06957333286603291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,32,1,balanced,0.07541333138942719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,32,1,balanced,0.10132267077763875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,32,1,balanced,0.1184266706307729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,32,1,balanced,0.16005333264668783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,32,power_law_1.2,0.1314239978790283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,32,1,balanced,0.1846346656481425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,32,1,balanced,0.2643360098203023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,4,4,power_law_1.01,0.23356161117553711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,32,1,balanced,0.3411253293355306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,32,1,balanced,0.48949865500132245
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,32,1,balanced,0.6344319979349772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,32,1,balanced,0.7885013421376547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,4,4,power_law_1.01,0.3070080041885376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,32,1,balanced,1.2673439979553223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,32,power_law_1.2,0.14011520147323608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,2,power_law_1.2,1.059539222717285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,32,1,balanced,2.478837331136068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,4,4,power_law_1.01,0.40621438026428225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,32,power_law_1.2,0.15372159481048583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,4,4,power_law_1.01,0.5807104110717773
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,2,power_law_1.2,1.2004672050476075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,32,power_law_1.2,0.1636415958404541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,4,4,power_law_1.01,0.7520063877105713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,32,power_law_1.2,0.1966591954231262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,2,power_law_1.2,1.3197248458862305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,4,4,power_law_1.01,1.2454336166381836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,32,power_law_1.2,0.2222912073135376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,2,power_law_1.2,1.5543231964111328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,32,power_law_1.2,0.28556160926818847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,4,4,power_law_1.01,2.401375961303711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,32,power_law_1.2,0.3398911952972412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,2,power_law_1.2,1.6455808639526368
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,32,power_law_1.2,0.46512641906738283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,32,power_law_1.2,0.5644480228424072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,2,power_law_1.2,2.008870315551758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,64,balanced,0.037802666425704956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,64,balanced,0.03942933430274328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,64,balanced,0.035605333745479584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,64,balanced,0.03431999931732813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,64,balanced,0.036474667489528656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,64,balanced,0.036917333801587425
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,64,balanced,0.038389332592487335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,64,balanced,0.03665599972009659
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,32,power_law_1.2,0.7706175804138183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,64,balanced,0.036517334481080375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,64,balanced,0.03677866607904434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,64,balanced,0.037045332292715706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,64,balanced,0.03682133307059606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,64,balanced,0.03809600075085958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,64,balanced,0.03941333293914795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,64,balanced,0.04233600199222565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,64,balanced,0.040965333580970764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,64,balanced,0.04118400067090988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,64,balanced,0.04389866689840952
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,64,balanced,0.04446400205294291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,64,balanced,0.050479998191197716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,64,balanced,0.06118933359781901
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,2,power_law_1.2,2.504416084289551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,64,balanced,0.06498133142789204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,64,balanced,0.06868800024191539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,64,balanced,0.07730666796366374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,64,balanced,0.08867200215657552
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,64,balanced,0.1029866635799408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,64,balanced,0.13276799519856772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,64,balanced,0.1560479998588562
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,64,balanced,0.2034239967664083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,32,power_law_1.2,1.1370176315307616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,64,balanced,0.34619732697804767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,32,power_law_1.2,1.3638208389282227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,2,power_law_1.2,3.584492874145508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,32,4,balanced,0.029919999341169994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,32,4,balanced,0.0322026660044988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,32,4,balanced,0.03188266605138779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,32,4,balanced,0.03587199995915095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,32,4,balanced,0.03825599948565165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,32,4,balanced,0.038389332592487335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,32,4,balanced,0.038149334490299225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,32,4,balanced,0.03972800076007843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,32,4,balanced,0.039642666776975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,32,4,balanced,0.039749334255854286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,32,4,balanced,0.040133332212766014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,32,power_law_1.2,1.98602237701416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,32,4,balanced,0.040207999447981514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,32,4,balanced,0.04167466859022776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,32,4,balanced,0.04377066592375437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,32,4,balanced,0.0455626646677653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,32,4,balanced,0.04614933331807455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,32,4,balanced,0.05228800078233083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,32,4,balanced,0.056186666091283165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,32,4,balanced,0.06070933242638906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,32,4,balanced,0.07835733393828075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,32,4,balanced,0.0883733332157135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,32,4,balanced,0.11458667119344075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,32,4,balanced,0.15660267074902853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,32,4,balanced,0.22928533951441446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,32,4,balanced,0.26607465744018555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,32,4,balanced,0.40194133917490643
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,32,4,balanced,0.4970613320668538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,32,4,balanced,0.6088800032933553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,32,4,balanced,0.9216266473134359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,2,power_law_1.2,4.061920166015625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,32,4,balanced,1.7596747080485027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,32,power_law_1.2,4.31910400390625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,1,2,power_law_1.2,4.629312133789062
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,1,balanced,0.05508266886075338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,1,balanced,0.09055466453234355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,1,balanced,0.10460266470909119
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,1,balanced,0.15754666924476624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,1,balanced,0.26121066013971966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,1,balanced,0.31772265831629437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,1,balanced,0.32182933886845905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,1,balanced,0.3204960028330485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,1,balanced,0.324069341023763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,1,balanced,0.32727466026941937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,1,balanced,0.32892266909281415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,1,balanced,0.3323306639989217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,1,balanced,0.3385973374048869
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,1,balanced,0.34419198830922443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,1,balanced,0.3535573482513428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,1,balanced,0.3614240090052287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,1,balanced,0.38213332494099933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,1,balanced,0.4215039809544881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,1,balanced,0.43702399730682373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,1,balanced,0.5423253377278646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,1,balanced,0.5680799881617228
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,1,balanced,0.8256586392720541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,1,balanced,0.8563199838002523
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,1,balanced,1.3941653569539387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,1,balanced,1.4566240310668945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,1,balanced,2.338101387023926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,1,balanced,2.7053066889444985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,1,2,power_law_1.2,7.5328125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,1,balanced,3.151797294616699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,1,balanced,5.069471995035808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,1,balanced,9.569061279296875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,8,power_law_1.2,0.09591680169105529
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,8,power_law_1.2,0.14467840194702147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,2,power_law_1.01,0.049542400240898135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,8,power_law_1.2,0.14633599519729615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,2,power_law_1.01,0.08515840172767639
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,8,power_law_1.2,0.0880895972251892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,2,power_law_1.01,0.10055040121078491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,8,power_law_1.2,0.13654400110244752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,1,2,power_law_1.2,14.339897155761719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,2,power_law_1.01,0.08490239977836608
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,8,power_law_1.2,0.1749824047088623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,2,power_law_1.01,0.12400000095367432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,8,power_law_1.2,0.1740991950035095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,2,power_law_1.01,0.16915199756622315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,8,power_law_1.2,0.24028160572052001
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,2,power_law_1.01,0.20371840000152588
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,8,power_law_1.2,0.25547521114349364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,4,1,power_law_1.2,0.055904000997543335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,2,power_law_1.01,0.2408128023147583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,4,1,power_law_1.2,0.06813439726829529
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,8,power_law_1.2,0.24555521011352538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,4,1,power_law_1.2,0.07351040244102477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,4,1,power_law_1.2,0.1174015998840332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,2,power_law_1.01,0.26408960819244387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,4,1,power_law_1.2,0.12085119485855103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,8,power_law_1.2,0.2659199953079224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,4,1,power_law_1.2,0.12175999879837036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,4,1,power_law_1.2,0.12557439804077147
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,2,power_law_1.01,0.2634943962097168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,4,1,power_law_1.2,0.12473599910736084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,4,1,power_law_1.2,0.1279680013656616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,8,power_law_1.2,0.271724796295166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,4,1,power_law_1.2,0.13050240278244019
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,4,1,power_law_1.2,0.13022719621658324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,2,power_law_1.01,0.28012158870697024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,4,1,power_law_1.2,0.13175040483474731
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,4,1,power_law_1.2,0.13430399894714357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,8,power_law_1.2,0.2797760009765625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,4,1,power_law_1.2,0.14328320026397706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,4,1,power_law_1.2,0.15780479907989503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,2,power_law_1.01,0.29431040287017823
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,4,1,power_law_1.2,0.16643199920654297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,4,1,power_law_1.2,0.17089279890060424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,8,power_law_1.2,0.28635520935058595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,4,1,power_law_1.2,0.19748480319976808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,4,1,power_law_1.2,0.2195199966430664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,2,power_law_1.01,0.29229440689086916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,8,4,power_law_1.2,0.0328575998544693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,4,1,power_law_1.2,0.27550721168518066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,8,4,power_law_1.2,0.033395200967788696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,8,power_law_1.2,0.29005439281463624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,4,1,power_law_1.2,0.322380805015564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,8,4,power_law_1.2,0.03150080144405365
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,8,4,power_law_1.2,0.03152639865875244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,4,1,power_law_1.2,0.42099838256835936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,8,4,power_law_1.2,0.03189760148525238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,8,4,power_law_1.2,0.03261440098285675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,2,power_law_1.01,0.31087360382080076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,4,1,power_law_1.2,0.5152063846588135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,8,4,power_law_1.2,0.034220799803733826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,8,4,power_law_1.2,0.03489919900894165
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,8,4,power_law_1.2,0.03612799942493439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,8,power_law_1.2,0.3146687984466553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,4,1,power_law_1.2,0.7212224006652832
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,8,4,power_law_1.2,0.03671680092811584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,8,4,power_law_1.2,0.036051198840141296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,4,1,power_law_1.2,0.9142784118652344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,8,4,power_law_1.2,0.038431999087333676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,8,4,power_law_1.2,0.04012160003185272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,2,power_law_1.01,0.33258240222930907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,8,4,power_law_1.2,0.041740798950195314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,8,4,power_law_1.2,0.05355520248413086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,4,1,power_law_1.2,1.3148159980773926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,8,power_law_1.2,0.3207231998443604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,8,4,power_law_1.2,0.054099202156066895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,8,4,power_law_1.2,0.05713919997215271
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,8,4,power_law_1.2,0.07454079985618592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,8,4,power_law_1.2,0.0828927993774414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,4,1,power_law_1.2,1.6939775466918945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,2,power_law_1.01,0.33509120941162107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,8,4,power_law_1.2,0.10665600299835205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,8,4,power_law_1.2,0.12584320306777955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,8,4,power_law_1.2,0.17182719707489014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,8,power_law_1.2,0.3606911897659302
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,8,4,power_law_1.2,0.18784639835357667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,4,1,power_law_1.2,2.0751935958862306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,8,4,power_law_1.2,0.2846143960952759
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,2,power_law_1.01,0.35358080863952634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,8,4,power_law_1.2,0.34275839328765867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,8,4,power_law_1.2,0.5773056030273438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,8,power_law_1.2,0.38546559810638426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,4,1,power_law_1.2,3.239244842529297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,8,4,power_law_1.2,0.5646912097930908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,2,power_law_1.01,0.38990719318389894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,8,4,power_law_1.2,0.8381952285766602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,8,power_law_1.2,0.44525442123413084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,8,4,power_law_1.2,1.432755184173584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,2,power_law_1.01,0.42254080772399905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,4,1,power_law_1.2,6.348774337768555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,8,power_law_1.2,0.472870397567749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,8,4,power_law_1.2,2.6059776306152345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,2,power_law_1.01,0.490828800201416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,8,power_law_1.2,0.5455808162689209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,2,power_law_1.01,0.5509888172149658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,8,power_law_1.2,0.6340352058410644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,2,power_law_1.01,0.6956352233886719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,8,power_law_1.2,0.8400064468383789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,2,power_law_1.01,0.8781696319580078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,8,power_law_1.2,0.9085568428039551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,2,power_law_1.01,1.1408448219299316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,8,power_law_1.2,1.1891136169433594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,2,power_law_1.01,1.3250368118286133
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,8,power_law_1.2,1.5088064193725585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,2,power_law_1.01,1.848409652709961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,2,balanced,0.05173333485921224
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,2,balanced,0.05734399954477946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,16,4,balanced,0.043605332573254905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,2,balanced,0.07391466697057088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,16,4,balanced,0.03365866591533025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,2,balanced,0.10333333412806193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,16,4,balanced,0.0353973334034284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,16,4,balanced,0.03537066777547201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,2,balanced,0.15491200486818948
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,16,4,balanced,0.035429333647092186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,16,4,balanced,0.03570133447647095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,16,4,balanced,0.035445332527160645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,2,balanced,0.25045865774154663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,16,4,balanced,0.03575466573238373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,16,4,balanced,0.0352960005402565
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,2,balanced,0.2548853357632955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,16,4,balanced,0.03571200122435888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,16,4,balanced,0.03957333415746689
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,8,power_law_1.2,2.009337615966797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,16,4,balanced,0.03737066686153412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,2,balanced,0.2586933374404907
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,16,4,balanced,0.03728000074625015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,2,power_law_1.01,2.3611072540283202
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,16,4,balanced,0.03975466638803482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,16,4,balanced,0.050016000866889954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,2,balanced,0.25704000393549603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,16,4,balanced,0.05004799862702688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,16,4,balanced,0.04970133304595947
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,16,4,balanced,0.05835199852784475
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,2,balanced,0.25782932837804157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,16,4,balanced,0.0654666672150294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,16,4,balanced,0.0698773314555486
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,2,balanced,0.2623093326886495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,16,4,balanced,0.08021866778532664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,16,4,balanced,0.10154666503270467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,2,balanced,0.2635466655095418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,16,4,balanced,0.11926399668057759
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,16,4,balanced,0.15129066507021585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,2,balanced,0.2669493357340495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,16,4,balanced,0.17929067214330038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,2,balanced,0.27242134014765423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,16,4,balanced,0.23975465695063272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,2,balanced,0.2841920057932536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,16,4,balanced,0.30137600501378375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,16,4,balanced,0.3630186716715495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,2,balanced,0.2870826721191406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,2,balanced,0.297818660736084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,16,4,balanced,0.5211679935455322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,2,balanced,0.3189653356870015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,16,4,balanced,0.9897387027740479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,2,balanced,0.3340799808502197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,2,balanced,0.38236268361409503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,2,balanced,0.41443200906117755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,2,balanced,0.518336017926534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,2,balanced,0.5916586716969808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,2,balanced,0.8118027051289877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,2,power_law_1.01,2.8604480743408205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,2,balanced,0.9552373091379801
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,8,power_law_1.2,2.650150489807129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,2,balanced,1.3184160391489665
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,2,balanced,1.6487253506978352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,2,balanced,2.045413335164388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,2,balanced,3.043813387552897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,2,balanced,5.799317042032878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,2,power_law_1.01,4.346905517578125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,8,power_law_1.2,5.019782257080078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,2,power_law_1.01,8.194451141357423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,4,power_law_1.2,0.047302401065826415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,4,power_law_1.2,0.0536191999912262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,4,power_law_1.2,0.04764800071716309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,4,power_law_1.2,0.05416319966316223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,2,power_law_1.2,0.07713919878005981
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,4,power_law_1.2,0.05616639852523804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,2,power_law_1.2,0.06395519971847534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,16,power_law_1.01,0.06779519915580749
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,4,power_law_1.2,0.06342399716377259
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,16,power_law_1.01,0.11742719411849975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,2,power_law_1.2,0.08381440043449402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,4,power_law_1.2,0.07751039862632751
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,16,power_law_1.01,0.05930240154266357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,2,power_law_1.2,0.11678080558776856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,4,power_law_1.2,0.08314239978790283
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,16,power_law_1.01,0.06076800227165222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,128,power_law_1.2,0.046163201332092285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,2,power_law_1.2,0.1471295952796936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,4,power_law_1.2,0.0828224003314972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,16,power_law_1.01,0.0664192020893097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,128,power_law_1.2,0.04684799909591675
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,2,power_law_1.2,0.2385024070739746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,128,power_law_1.2,0.04528000056743622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,4,power_law_1.2,0.08650239706039428
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,16,power_law_1.01,0.07856000065803528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,128,power_law_1.2,0.04615040123462677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,2,power_law_1.2,0.26236159801483155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,4,power_law_1.2,0.0923520028591156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,16,power_law_1.01,0.08673920035362244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,128,power_law_1.2,0.044736000895500186
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,2,power_law_1.2,0.266592001914978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,128,power_law_1.2,0.047276800870895384
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,4,power_law_1.2,0.09620479941368103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,16,power_law_1.01,0.09077759981155395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,128,power_law_1.2,0.04635519981384277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,2,power_law_1.2,0.2849087953567505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,4,power_law_1.2,0.09966719746589661
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,128,power_law_1.2,0.045612800121307376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,16,power_law_1.01,0.08823680281639099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,128,power_law_1.2,0.046003198623657225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,4,power_law_1.2,0.10884480476379395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,2,power_law_1.2,0.28548479080200195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,16,power_law_1.01,0.09047679901123047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,128,power_law_1.2,0.04456959962844849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,4,power_law_1.2,0.11967359781265259
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,128,power_law_1.2,0.04257920086383819
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,2,power_law_1.2,0.3009536027908325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,16,power_law_1.01,0.09055359959602356
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,128,power_law_1.2,0.04283519983291626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,4,power_law_1.2,0.1277567982673645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,16,power_law_1.01,0.0943615972995758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,2,power_law_1.2,0.3138047933578491
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,128,power_law_1.2,0.04257920086383819
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,128,power_law_1.2,0.044588801264762876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,4,power_law_1.2,0.1296895980834961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,16,power_law_1.01,0.09827200174331666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,2,power_law_1.2,0.30977280139923097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,128,power_law_1.2,0.04893440008163452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,128,power_law_1.2,0.047443199157714847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,16,power_law_1.01,0.09740800261497498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,4,power_law_1.2,0.14807679653167724
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,2,power_law_1.2,0.3279616117477417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,128,power_law_1.2,0.05223039984703064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,16,power_law_1.01,0.10194560289382934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,128,power_law_1.2,0.06152960062026978
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,4,power_law_1.2,0.17040640115737915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,2,power_law_1.2,0.3307775974273682
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,128,power_law_1.2,0.06951040029525757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,16,power_law_1.01,0.10599679946899414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,4,power_law_1.2,0.2076416015625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,128,power_law_1.2,0.08872960209846496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,2,power_law_1.2,0.3488895893096924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,128,power_law_1.2,0.09299200177192687
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,16,power_law_1.01,0.107532799243927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,4,power_law_1.2,0.23415679931640626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,128,power_law_1.2,0.12570879459381104
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,2,power_law_1.2,0.3568511962890625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,16,power_law_1.01,0.1255679965019226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,128,power_law_1.2,0.16725120544433594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,4,power_law_1.2,0.3122175931930542
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,2,power_law_1.2,0.3972287893295288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,16,power_law_1.01,0.12789759635925294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,128,power_law_1.2,0.24560000896453857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,4,power_law_1.2,0.4006336212158203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,2,power_law_1.2,0.42175998687744143
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,16,power_law_1.01,0.15852160453796388
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,128,power_law_1.2,0.314137601852417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,16,power_law_1.01,0.1742527961730957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,2,power_law_1.2,0.5137343883514405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,4,power_law_1.2,0.5468160152435303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,128,power_law_1.2,0.4481152057647705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,16,power_law_1.01,0.20704638957977295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,2,power_law_1.2,0.5598336219787597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,128,power_law_1.2,0.6117184162139893
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,4,power_law_1.2,0.6639488220214844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,16,power_law_1.01,0.2545023918151855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,2,power_law_1.2,0.6863167762756348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,128,power_law_1.2,0.7432831764221192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,4,power_law_1.2,0.9277119636535645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,16,power_law_1.01,0.3180351972579956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,2,power_law_1.2,0.7987135887145996
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,128,power_law_1.2,1.188492774963379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,16,power_law_1.01,0.4342463970184326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,4,power_law_1.2,1.3757568359375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,2,power_law_1.2,1.0146112442016602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,16,power_law_1.01,0.5462399959564209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,2,power_law_1.2,1.1960512161254884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,4,power_law_1.2,1.6027776718139648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,128,power_law_1.2,2.35150089263916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,16,power_law_1.01,0.6770495891571044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,2,power_law_1.2,1.6702848434448243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,4,power_law_1.2,2.55930233001709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,16,power_law_1.01,0.8240703582763672
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,2,power_law_1.2,2.199673652648926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,16,power_law_1.01,1.2145600318908691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,2,power_law_1.2,2.8775039672851563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,4,power_law_1.2,5.02476806640625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,2,power_law_1.2,3.83441276550293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,16,power_law_1.01,2.3275136947631836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,2,power_law_1.2,7.296550750732422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,256,balanced,0.06668266654014587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,256,balanced,0.0719893326361974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,256,balanced,0.0685280015071233
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,256,balanced,0.07624533275763194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,256,balanced,0.0743999977906545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,256,balanced,0.07470933099587758
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,256,balanced,0.0763626645008723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,256,balanced,0.0784853349129359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,256,balanced,0.0811466674009959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,256,balanced,0.08099199831485748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,256,balanced,0.07885333398977916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,256,balanced,0.07660266757011414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,256,balanced,0.08050666749477386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,256,balanced,0.07958933214346568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,256,balanced,0.08563199639320374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,256,balanced,0.08618666728337605
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,256,balanced,0.08948799967765808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,256,balanced,0.10122666756312053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,256,balanced,0.10552533467610677
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,256,balanced,0.12377599875132243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,256,balanced,0.14006933569908142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,256,balanced,0.18122132619222006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,256,balanced,0.176581343015035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,256,balanced,0.23207465807596842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,256,balanced,0.258245329062144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,256,balanced,0.3593440055847168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,256,balanced,0.4255359967549642
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,1,256,balanced,0.4943679968516032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,1,256,balanced,0.7376159826914469
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,1,256,balanced,1.586714744567871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,128,power_law_1.01,0.03296000063419342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,128,power_law_1.01,0.0335999995470047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,2,balanced,0.06903466582298279
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,2,balanced,0.0771679977575938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,128,power_law_1.01,0.033055999875068666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,2,balanced,0.09830933809280396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,2,balanced,0.132341335217158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,2,balanced,0.19340799252192178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,2,balanced,0.31959466139475506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,2,balanced,0.44414401054382324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,2,balanced,0.44761065642038983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,128,power_law_1.01,0.03335680067539215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,2,balanced,0.4513973395029704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,2,balanced,0.4498186508814494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,2,balanced,0.4549280007680257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,2,balanced,0.45899732907613117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,128,power_law_1.01,0.03411200046539307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,2,balanced,0.4645013411839803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,2,balanced,0.468453327814738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,2,balanced,0.4754879872004191
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,2,balanced,0.480618675549825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,2,balanced,0.5036853154500326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,128,power_law_1.01,0.034835198521614076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,2,balanced,0.5215253432591757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,2,balanced,0.5557493368784586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,2,balanced,0.6107840140660604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,2,balanced,0.6520373423894247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,128,power_law_1.01,0.034796801209449765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,2,balanced,0.7497066656748453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,2,balanced,0.850757360458374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,128,power_law_1.01,0.03561600148677826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,2,balanced,1.0391626358032227
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,2,balanced,1.3232853412628174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,128,power_law_1.01,0.03529599905014038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,2,balanced,1.7042826016743977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,128,power_law_1.01,0.035596799850463864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,2,balanced,2.3251147270202637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,16,power_law_1.2,0.09178239703178406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,128,power_law_1.01,0.0361407995223999
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,2,balanced,2.946533203125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,128,power_law_1.01,0.036601600050926206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,2,balanced,4.168176015218099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,16,power_law_1.2,0.06547840237617493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,128,power_law_1.01,0.03709439933300018
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,128,power_law_1.01,0.037574398517608645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,16,power_law_1.2,0.06064640283584595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,128,power_law_1.01,0.040345600247383116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,2,balanced,7.90178108215332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,4,power_law_1.01,0.041145598888397215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,128,power_law_1.01,0.04074240028858185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,16,power_law_1.2,0.06762239933013917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,4,power_law_1.01,0.03412480056285858
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,128,power_law_1.01,0.04273920059204102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,4,power_law_1.01,0.035206401348114015
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,128,power_law_1.01,0.04255360066890716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,16,power_law_1.2,0.07354239821434021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,4,power_law_1.01,0.04180479943752289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,128,power_law_1.01,0.04589439928531647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,4,power_law_1.01,0.041740798950195314
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,128,power_law_1.01,0.0484607994556427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,16,power_law_1.2,0.0852735996246338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,4,power_law_1.01,0.048511999845504764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,128,power_law_1.01,0.052095997333526614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,4,power_law_1.01,0.049369600415229795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,16,power_law_1.2,0.09057919979095459
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,128,power_law_1.01,0.0621504008769989
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,4,power_law_1.01,0.0494271993637085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,128,power_law_1.01,0.07634559869766236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,4,power_law_1.01,0.050387197732925416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,16,power_law_1.2,0.09082880020141601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,4,power_law_1.01,0.050860798358917235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,128,power_law_1.01,0.10670720338821411
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,16,power_law_1.2,0.09289600253105164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,4,power_law_1.01,0.05221760272979736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,128,power_law_1.01,0.13461120128631593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,4,power_law_1.01,0.05560960173606873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,16,power_law_1.2,0.0914687991142273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,4,power_law_1.01,0.05487359762191772
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,128,power_law_1.01,0.189683198928833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,4,power_law_1.01,0.05694079995155334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,16,power_law_1.2,0.09593600034713745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,4,power_law_1.01,0.06264320015907288
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,128,power_law_1.01,0.24186239242553711
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,4,power_law_1.01,0.06394879817962647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,16,power_law_1.2,0.09715840220451355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,2,128,power_law_1.01,0.292409610748291
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,4,power_law_1.01,0.06894720196723939
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,4,power_law_1.01,0.07964159846305847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,16,power_law_1.2,0.09883520007133484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,4,power_law_1.01,0.08654720187187195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,2,128,power_law_1.01,0.45034241676330566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,16,power_law_1.2,0.09897599816322326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,4,power_law_1.01,0.09920639991760254
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,4,power_law_1.01,0.11561599969863892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,16,power_law_1.2,0.10618879795074462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,2,balanced,0.049642667174339294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,2,balanced,0.051957334081331887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,2,balanced,0.056885331869125366
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,2,balanced,0.07821866869926453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,4,power_law_1.01,0.14876799583435057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,2,balanced,0.11126933495203654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,2,balanced,0.16794133186340332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,2,balanced,0.1709280014038086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,2,balanced,0.17454934120178223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,2,balanced,0.1769226590792338
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,2,128,power_law_1.01,0.9022080421447753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,2,balanced,0.17883199453353882
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,2,balanced,0.18168532848358154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,16,power_law_1.2,0.10685440301895141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,2,balanced,0.18367467323939005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,2,balanced,0.18417600790659586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,2,balanced,0.18531199296315512
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,4,power_law_1.01,0.20213758945465088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,2,balanced,0.19261332352956137
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,2,balanced,0.19356266657511392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,2,balanced,0.20946133136749268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,2,balanced,0.23250667254130045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,2,balanced,0.250602662563324
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,2,balanced,0.2908160090446472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,2,balanced,0.3207733432451884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,2,balanced,0.4320853153864543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,4,power_law_1.01,0.25420799255371096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,16,power_law_1.2,0.11612160205841064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,2,balanced,0.4863946835199992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,2,balanced,0.6909546852111816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,2,balanced,0.8324800332387289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,2,balanced,1.1974559624989827
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,4,power_law_1.01,0.35997440814971926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,16,power_law_1.2,0.1341312050819397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,2,balanced,1.5317120552062988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,2,balanced,1.8665439287821453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,4,power_law_1.01,0.48731517791748047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,16,power_law_1.2,0.14895360469818114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,2,balanced,2.8268585205078125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,16,power_law_1.2,0.1770367980003357
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,4,power_law_1.01,0.5682688236236573
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,2,balanced,5.452517191569011
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,16,power_law_1.2,0.20397439002990722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,4,power_law_1.01,0.7362815856933593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,16,power_law_1.2,0.24753279685974122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,16,power_law_1.2,0.30472960472106936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,4,power_law_1.01,1.186950397491455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,16,power_law_1.2,0.39152638912200927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,16,power_law_1.2,0.5394752025604248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,4,power_law_1.01,2.348620796203613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,16,power_law_1.2,0.847532844543457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,16,power_law_1.2,1.156454372406006
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,2,power_law_1.01,0.032953599095344545
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,16,power_law_1.2,1.1777088165283203
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,2,power_law_1.01,0.03468160033226013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,2,power_law_1.01,0.039052799344062805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,2,power_law_1.01,0.04467200040817261
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,2,power_law_1.01,0.05651199817657471
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,16,power_law_1.2,2.429113578796387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,2,power_law_1.01,0.07579519748687744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,2,power_law_1.01,0.07865599989891052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,2,power_law_1.01,0.08117759823799134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,1,power_law_1.2,0.07503359913825988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,2,power_law_1.01,0.08357759714126586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,2,power_law_1.01,0.08216959834098816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,1,power_law_1.2,0.07839360237121581
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,2,power_law_1.01,0.0868287980556488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,16,power_law_1.2,4.354054260253906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,1,power_law_1.2,0.12052479982376099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,2,power_law_1.01,0.09000319838523865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,1,power_law_1.2,0.16462719440460205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,2,power_law_1.01,0.09205120205879211
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,1,power_law_1.2,0.23662080764770507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,2,power_law_1.01,0.09149439930915833
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,2,power_law_1.01,0.10139520168304443
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,1,power_law_1.2,0.4397759914398193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,2,2,power_law_1.2,0.06452479958534241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,2,power_law_1.01,0.10432640314102173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,2,2,power_law_1.2,0.07215359807014465
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,1,power_law_1.2,0.476364803314209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,2,2,power_law_1.2,0.07338240146636962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,2,power_law_1.01,0.1273408055305481
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,2,2,power_law_1.2,0.09265919923782348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,2,2,power_law_1.2,0.0936959981918335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,1,power_law_1.2,0.5038591861724854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,2,2,power_law_1.2,0.09402239918708802
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,2,power_law_1.01,0.14771840572357178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,2,2,power_law_1.2,0.09423999786376953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,2,2,power_law_1.2,0.09468799829483032
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,1,power_law_1.2,0.5179327964782715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,2,2,power_law_1.2,0.09715200066566468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,2,power_law_1.01,0.1402943968772888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,2,2,power_law_1.2,0.09727360010147094
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,2,2,power_law_1.2,0.09884160161018371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,2,2,power_law_1.2,0.10000640153884888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,2,power_law_1.01,0.1710528016090393
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,1,power_law_1.2,0.5418240070343018
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,2,2,power_law_1.2,0.10088319778442383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,2,2,power_law_1.2,0.10954240560531617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,2,2,power_law_1.2,0.12729599475860595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,2,power_law_1.01,0.18974080085754394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,2,2,power_law_1.2,0.13365119695663452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,1,power_law_1.2,0.5493311882019043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,2,2,power_law_1.2,0.1415552020072937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,2,2,power_law_1.2,0.16721919775009156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,2,power_law_1.01,0.25596160888671876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,2,2,power_law_1.2,0.1976639986038208
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,1,power_law_1.2,0.5725632190704346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,2,2,power_law_1.2,0.22684800624847412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,2,2,power_law_1.2,0.2623424053192139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,2,power_law_1.01,0.315283203125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,2,2,power_law_1.2,0.3952255964279175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,1,power_law_1.2,0.5889599800109864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,2,2,power_law_1.2,0.4656320095062256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,2,power_law_1.01,0.41969919204711914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,2,2,power_law_1.2,0.6845823764801026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,1,power_law_1.2,0.592851209640503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,2,2,power_law_1.2,0.9151871681213379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,2,power_law_1.01,0.5659584045410156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,1,power_law_1.2,0.626585578918457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,2,2,power_law_1.2,1.1235903739929198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,2,2,power_law_1.2,1.6775167465209961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,1,power_law_1.2,0.6447487831115722
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,2,power_law_1.01,0.7596992015838623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,2,2,power_law_1.2,1.9413440704345704
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,1,power_law_1.2,0.6669439792633056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,2,power_law_1.01,0.9742527961730957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,1,power_law_1.2,0.7369088172912598
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,2,2,power_law_1.2,3.4068672180175783
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,1,power_law_1.2,0.7806719779968262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,4096,14336,2,8,16,1,balanced,0.03369600077470144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,4096,14336,2,8,16,1,balanced,0.035258665680885315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,4096,14336,2,8,16,1,balanced,0.037578667203585304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,2,power_law_1.01,1.1680191993713378
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,4096,14336,2,8,16,1,balanced,0.039818666875362396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,4096,14336,2,8,16,1,balanced,0.04053866614898046
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,4096,14336,2,8,16,1,balanced,0.039264000952243805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,4096,14336,2,8,16,1,balanced,0.03913066784540812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,1,power_law_1.2,0.9374912261962891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,4096,14336,2,8,16,1,balanced,0.04106666644414266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,4096,14336,2,8,16,1,balanced,0.04182399809360504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,4096,14336,2,8,16,1,balanced,0.04246933261553446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,4096,14336,2,8,16,1,balanced,0.04365866879622141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,4096,14336,2,8,16,1,balanced,0.04452799757321676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,4096,14336,2,8,16,1,balanced,0.04689066608746847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,4096,14336,2,8,16,1,balanced,0.050016000866889954
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,4096,14336,2,8,16,1,balanced,0.06292266647020976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,2,2,power_law_1.2,6.798944091796875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,4096,14336,2,8,16,1,balanced,0.0621066689491272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,4096,14336,2,8,16,1,balanced,0.07197333375612895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,1,power_law_1.2,1.023583984375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,4096,14336,2,8,16,1,balanced,0.09005332986513774
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,16,1,balanced,0.09268800417582194
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,16,1,balanced,0.11925333738327026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,16,1,balanced,0.12979732950528464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,16,1,balanced,0.17347200711568198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,2,power_law_1.01,1.7806207656860351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,16,1,balanced,0.20542933543523154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,16,1,balanced,0.29426666100819904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,16,1,balanced,0.3747040033340454
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,1,power_law_1.2,1.2817983627319336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,16,1,balanced,0.5424853165944418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,16,1,balanced,0.7005226612091064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,16,1,balanced,0.8608106772104899
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,16,1,balanced,1.342682679494222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,1,power_law_1.2,1.3945856094360352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,16,1,balanced,2.623077392578125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,1,power_law_1.2,1.811347198486328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,2,power_law_1.01,3.5560062408447264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,1,power_law_1.2,2.5101247787475587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,1,power_law_1.2,3.2764225006103516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,1,power_law_1.2,4.033567810058594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,1,power_law_1.2,4.870367813110351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,64,power_law_1.2,0.08763520121574402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,1,power_law_1.2,6.937510681152344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,64,power_law_1.2,0.055795198678970336
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,64,power_law_1.2,0.050783997774124144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,64,power_law_1.2,0.05162879824638367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,64,power_law_1.2,0.05240960121154785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,256,32,1,power_law_1.2,0.04895359873771667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,4,balanced,0.03514666606982549
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,4,balanced,0.032933334509531655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,4,balanced,0.03479466587305069
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,4,balanced,0.039221333960692085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,4,balanced,0.05788266658782959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,4,balanced,0.05961599946022034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,64,power_law_1.2,0.05485439896583557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,4,balanced,0.059802666306495667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,4,balanced,0.061573331554730736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,256,32,1,power_law_1.2,0.05586559772491455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,4,balanced,0.06117333471775055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,4,balanced,0.0611413319905599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,4,balanced,0.061808000008265175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,4,balanced,0.06230400005976359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,4,balanced,0.06423999865849812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,4,balanced,0.0658133327960968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,4,balanced,0.07364266614119212
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,4,balanced,0.07196799914042155
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,64,power_law_1.2,0.05667200088500977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,4,balanced,0.07426133255163829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,4,balanced,0.08141333361466725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,4,balanced,0.09270399808883667
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,1,power_law_1.2,13.270719909667969
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,4,balanced,0.10774933298428853
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,4,balanced,0.11210133632024129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,4,balanced,0.14442132910092673
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,256,32,1,power_law_1.2,0.062015998363494876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,4,balanced,0.1673706571261088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,4,balanced,0.23361066977183023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,4,balanced,0.277946670850118
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,64,power_law_1.2,0.05868800282478333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,4,balanced,0.4029386838277181
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,4,balanced,0.5134880145390829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,4,balanced,0.6309173504511515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,4,balanced,0.9701066811879476
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,256,32,1,power_law_1.2,0.07621759772300721
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,64,power_law_1.2,0.058220797777175905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,4,balanced,1.8768107096354167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,64,power_law_1.2,0.06056960225105286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,256,32,1,power_law_1.2,0.09760640263557434
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,64,power_law_1.2,0.060102397203445436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,256,32,1,power_law_1.2,0.12365440130233765
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,64,power_law_1.2,0.06093440055847168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,256,32,1,power_law_1.2,0.1624575972557068
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,64,power_law_1.2,0.061382400989532473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,256,32,1,power_law_1.2,0.16792320013046264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,64,power_law_1.2,0.06845440268516541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,256,32,1,power_law_1.2,0.17608319520950316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,64,power_law_1.2,0.07854080200195312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,256,32,1,power_law_1.2,0.18246400356292725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,64,power_law_1.2,0.07913600206375122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,256,32,1,power_law_1.2,0.1919935941696167
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,64,power_law_1.2,0.08303359746932984
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,256,32,1,power_law_1.2,0.20833919048309327
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,64,power_law_1.2,0.09467520117759705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,256,32,1,power_law_1.2,0.209881591796875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,64,power_law_1.2,0.10685440301895141
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,256,32,1,power_law_1.2,0.23311359882354737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,64,power_law_1.2,0.12342400550842285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,256,32,1,power_law_1.2,0.24108800888061524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,64,power_law_1.2,0.15192320346832275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,4,16,power_law_1.01,0.03929600119590759
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,256,32,1,power_law_1.2,0.25214719772338867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,4,16,power_law_1.01,0.036108800768852235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,64,power_law_1.2,0.19352960586547852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,256,32,1,power_law_1.2,0.2559999942779541
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,1,8,power_law_1.01,0.08135039806365967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,4,16,power_law_1.01,0.03335680067539215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,1,8,power_law_1.01,0.08371840119361877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,4,16,power_law_1.01,0.03288320004940033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,1,8,power_law_1.01,0.08184959888458251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,64,power_law_1.2,0.23717761039733887
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,256,32,1,power_law_1.2,0.30458879470825195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,1,8,power_law_1.01,0.08318079710006714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,4,16,power_law_1.01,0.03379839956760407
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,1,8,power_law_1.01,0.08280320167541504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,1,8,power_law_1.01,0.08674560189247131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,4,16,power_law_1.01,0.03516800105571747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,64,power_law_1.2,0.30726399421691897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,256,32,1,power_law_1.2,0.3714688062667847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,1,8,power_law_1.01,0.09368320107460022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,4,16,power_law_1.01,0.03542400002479553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,1,8,power_law_1.01,0.0930239975452423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,1,8,power_law_1.01,0.09380480051040649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,4,16,power_law_1.01,0.03537920117378235
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,1,8,power_law_1.01,0.09750400185585022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,64,power_law_1.2,0.4131648063659668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,256,32,1,power_law_1.2,0.44219517707824707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,4,16,power_law_1.01,0.035718399286270144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,1,8,power_law_1.01,0.09751039743423462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,1,8,power_law_1.01,0.10775680541992187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,4,16,power_law_1.01,0.03607040047645569
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,1,8,power_law_1.01,0.10674560070037842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,256,32,1,power_law_1.2,0.48694400787353515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,64,power_law_1.2,0.5459775924682617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,1,8,power_law_1.01,0.13898240327835082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,4,16,power_law_1.01,0.036831998825073244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,1,8,power_law_1.01,0.15831680297851564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,4,16,power_law_1.01,0.0374208003282547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,1,8,power_law_1.01,0.1729856014251709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,256,32,1,power_law_1.2,0.6372799873352051
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,1,8,power_law_1.01,0.15482879877090455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,4,16,power_law_1.01,0.03792639970779419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,1,8,power_law_1.01,0.20481278896331787
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,64,power_law_1.2,0.7521024227142334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,1,8,power_law_1.01,0.23418879508972168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,4,16,power_law_1.01,0.03912320137023926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,256,32,1,power_law_1.2,0.8034496307373047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,1,8,power_law_1.01,0.36165120601654055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,4,16,power_law_1.01,0.04348160028457641
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,1,8,power_law_1.01,0.4089087963104248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,4,16,power_law_1.01,0.04444159865379334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,1,8,power_law_1.01,0.5537792205810547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,64,power_law_1.2,0.8954815864562988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,256,32,1,power_law_1.2,1.0843584060668945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,1,8,power_law_1.01,0.7437183856964111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,4,16,power_law_1.01,0.046540799736976626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,1,8,power_law_1.01,1.1675583839416503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,4,16,power_law_1.01,0.04915199875831604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,4,16,power_law_1.01,0.05204480290412903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,1,8,power_law_1.01,1.4860608100891113
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,256,32,1,power_law_1.2,1.3594112396240234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,64,power_law_1.2,1.3965632438659668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,4,16,power_law_1.01,0.05729280114173889
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,1,8,power_law_1.01,2.2066112518310548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,4,16,power_law_1.01,0.0692351996898651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,256,32,1,power_law_1.2,1.9584127426147462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,4,16,power_law_1.01,0.08894720077514648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,1,8,power_law_1.01,2.778816032409668
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,4,16,power_law_1.01,0.09214720129966736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,1,8,power_law_1.01,3.7822078704833983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,4,16,power_law_1.01,0.13177599906921386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,256,32,1,power_law_1.2,2.54453125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,64,power_law_1.2,2.8014144897460938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,4,16,power_law_1.01,0.15723520517349243
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,1,8,power_law_1.01,5.953145599365234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,4,16,power_law_1.01,0.2459264039993286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,256,32,1,power_law_1.2,3.115622329711914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,4,16,power_law_1.01,0.28958079814910886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,4,16,power_law_1.01,0.32624640464782717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,1,8,power_law_1.01,12.01370849609375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,256,32,1,power_law_1.2,4.868716812133789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,4,16,power_law_1.01,0.49064321517944337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,4,16,power_law_1.01,1.0066368103027343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,256,32,1,power_law_1.2,9.463660430908202
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,16,power_law_1.01,0.059648001194000246
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,16,power_law_1.01,0.06824960112571717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,16,power_law_1.01,0.06707839965820313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,16,power_law_1.01,0.03628160059452057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,16,power_law_1.01,0.04746240079402923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,16,power_law_1.01,0.03338879942893982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,16,power_law_1.01,0.029286399483680725
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,16,power_law_1.01,0.05177599787712097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,16,power_law_1.01,0.029657599329948426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,16,power_law_1.01,0.031174400448799135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,16,power_law_1.01,0.05556480288505554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,16,power_law_1.01,0.032678401470184325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,16,power_law_1.01,0.03232640027999878
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,16,power_law_1.01,0.05726079940795899
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,2,power_law_1.2,0.1732800006866455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,16,power_law_1.01,0.03219200074672699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,14336,2,8,4,8,power_law_1.2,0.04252800047397613
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,14336,2,8,4,8,power_law_1.2,0.04140160083770752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,16,power_law_1.01,0.03459199965000152
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,16,power_law_1.01,0.06175360083580017
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,2,power_law_1.2,0.1635967969894409
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,14336,2,8,4,8,power_law_1.2,0.042335999011993405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,14336,2,8,4,8,power_law_1.2,0.04399999976158142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,14336,2,8,4,8,power_law_1.2,0.04318720102310181
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,14336,2,8,4,8,power_law_1.2,0.047788798809051514
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,2,power_law_1.2,0.2122431993484497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,14336,2,8,4,8,power_law_1.2,0.05475839972496033
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,14336,2,8,4,8,power_law_1.2,0.04503040015697479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,14336,2,8,4,8,power_law_1.2,0.04901120066642761
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,16,power_law_1.01,0.06466559767723083
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,16,power_law_1.01,0.034220799803733826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,14336,2,8,4,8,power_law_1.2,0.051148802042007446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,2,power_law_1.2,0.3068608045578003
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,14336,2,8,4,8,power_law_1.2,0.045311999320983884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,14336,2,8,4,8,power_law_1.2,0.051526397466659546
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,14336,2,8,4,8,power_law_1.2,0.05726720094680786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,14336,2,8,4,8,power_law_1.2,0.05820159912109375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,2,power_law_1.2,0.4224063873291016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,14336,2,8,4,8,power_law_1.2,0.06995840072631836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,16,power_law_1.01,0.06430720090866089
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,16,power_law_1.01,0.03463039994239807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,14336,2,8,4,8,power_law_1.2,0.06872320175170898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,14336,2,8,4,8,power_law_1.2,0.07416319847106934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,14336,2,8,4,8,power_law_1.2,0.09290239810943604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,2,power_law_1.2,0.7842624187469482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,14336,2,8,4,8,power_law_1.2,0.09468160271644592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,14336,2,8,4,8,power_law_1.2,0.12280319929122925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,14336,2,8,4,8,power_law_1.2,0.1450111985206604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,16,power_law_1.01,0.035385599732398985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,16,power_law_1.01,0.06607999801635742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,2,power_law_1.2,0.8935487747192383
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,14336,2,8,4,8,power_law_1.2,0.17853440046310426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,14336,2,8,4,8,power_law_1.2,0.22883200645446777
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,14336,2,8,4,8,power_law_1.2,0.349068808555603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,2,power_law_1.2,0.9294272422790527
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,16,power_law_1.01,0.03553279936313629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,14336,2,8,4,8,power_law_1.2,0.3904383897781372
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,16,power_law_1.01,0.06874240040779114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,14336,2,8,4,8,power_law_1.2,0.5921472072601318
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,2,power_law_1.2,0.9721920013427734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,14336,2,8,4,8,power_law_1.2,0.8613568305969238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,16,power_law_1.01,0.03749119937419891
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,14336,2,8,4,8,power_law_1.2,1.0999872207641601
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,2,power_law_1.2,0.9808575630187988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,16,power_law_1.01,0.06719359755516052
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,16,power_law_1.01,0.04084480106830597
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,14336,2,8,4,8,power_law_1.2,1.5535872459411622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,2,power_law_1.2,1.0357888221740723
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,16,power_law_1.01,0.041817599534988405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,16,power_law_1.01,0.07349759936332703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,2,power_law_1.2,1.0406784057617187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,14336,2,8,4,8,power_law_1.2,3.040115165710449
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,16,power_law_1.01,0.04475519955158234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,2,power_law_1.2,1.0699968338012695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,16,power_law_1.01,0.08044800162315369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,16,power_law_1.01,0.04792320132255554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,2,power_law_1.2,1.1056639671325683
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,16,power_law_1.01,0.0566976010799408
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,16,power_law_1.01,0.08131840229034423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,2,power_law_1.2,1.1730560302734374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,16,power_law_1.01,0.06566399931907654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,16,power_law_1.01,0.08862079977989197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,2,power_law_1.2,1.174124813079834
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,16,power_law_1.01,0.07493119835853576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,2,power_law_1.2,1.1680319786071778
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,16,power_law_1.01,0.10030080080032348
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,16,power_law_1.01,0.08980479836463928
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,2,power_law_1.2,1.3403840065002441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,16,power_law_1.01,0.1023743987083435
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,16,power_law_1.01,0.10890239477157593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,2,power_law_1.2,1.4136192321777343
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,16,power_law_1.01,0.1688447952270508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,16,power_law_1.01,0.1271999955177307
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,2,power_law_1.2,1.6960384368896484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,16,power_law_1.01,0.1938815951347351
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,2,power_law_1.2,1.7520959854125977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,16,power_law_1.01,0.1422271966934204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,16,power_law_1.01,0.25679359436035154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,2,power_law_1.2,2.1735551834106444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,16,power_law_1.01,0.18205440044403076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,16,power_law_1.01,0.32026240825653074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,2,power_law_1.2,2.1407487869262694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,16,power_law_1.01,0.22176640033721923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,16,power_law_1.01,0.36272640228271485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,2,power_law_1.2,2.6399999618530274
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,16,power_law_1.01,0.2858367919921875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,2,power_law_1.2,3.0260543823242188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,16,power_law_1.01,0.622214412689209
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,16,power_law_1.2,0.08573439717292786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,16,power_law_1.01,0.35917439460754397
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,2,power_law_1.2,4.077401733398437
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,16,power_law_1.01,0.5061183929443359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,16,power_law_1.2,0.07279360294342041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,16,power_law_1.01,1.1713855743408204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,2,power_law_1.2,5.48535041809082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,16,power_law_1.01,0.6338751792907715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,16,power_law_1.2,0.05920000076293945
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,16,power_law_1.2,0.0655232012271881
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,16,power_law_1.01,0.7784895896911621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,2,power_law_1.2,6.317971038818359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,16,power_law_1.2,0.07308160066604615
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,16,power_law_1.01,1.2450240135192872
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,16,power_law_1.2,0.0898688018321991
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,2,power_law_1.2,9.163609313964844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,16,power_law_1.2,0.09728639721870422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,16,power_law_1.01,2.386969566345215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,16,power_law_1.2,0.09696000218391418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,16,power_law_1.2,0.09866880178451538
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,32,2,power_law_1.2,0.03307519853115082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,32,2,power_law_1.2,0.03221119940280914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,32,2,power_law_1.2,0.03342719972133636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,2,power_law_1.2,16.62401885986328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,16,power_law_1.2,0.10154240131378174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,32,2,power_law_1.2,0.034355199337005614
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,32,2,power_law_1.2,0.03678080141544342
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,32,2,power_law_1.2,0.035667198896408084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,32,2,power_law_1.2,0.036831998825073244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,32,2,power_law_1.2,0.03711999952793121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,32,2,power_law_1.2,0.040268799662590025
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,16,power_law_1.2,0.10576640367507935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,32,2,power_law_1.2,0.041715198755264284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,32,2,power_law_1.2,0.04387840032577515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,32,2,power_law_1.2,0.044435200095176694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,32,2,power_law_1.2,0.0436928004026413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,16,power_law_1.2,0.1077888011932373
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,32,2,power_law_1.2,0.04853120148181915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,32,2,power_law_1.2,0.05697280168533325
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,32,2,power_law_1.2,0.05878400206565857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,32,2,power_law_1.2,0.06440960168838501
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,32,2,power_law_1.2,0.07541760206222534
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,16,power_law_1.2,0.10958080291748047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,32,2,power_law_1.2,0.08269439935684204
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,32,2,power_law_1.2,0.09132159948348999
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,32,2,power_law_1.2,0.12459520101547242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,32,2,power_law_1.2,0.16840319633483886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,32,2,power_law_1.2,0.18825600147247315
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,16,power_law_1.2,0.1126207947731018
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,32,2,power_law_1.2,0.2805567979812622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,32,2,power_law_1.2,0.33868160247802737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,32,2,power_law_1.2,0.4519680023193359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,32,2,power_law_1.2,0.5959360122680664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,16,power_law_1.2,0.11851520538330078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,32,2,power_law_1.2,0.6382463932037353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,32,2,power_law_1.2,1.037504005432129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,1,power_law_1.01,0.07341439723968506
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,16,power_law_1.2,0.12649600505828856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,1,power_law_1.01,0.0944320023059845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,32,2,power_law_1.2,2.1437376022338865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,1,power_law_1.01,0.12496639490127563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,16,power_law_1.2,0.13278080224990846
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,1,power_law_1.01,0.18922879695892333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,64,balanced,0.03989866624275843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,64,balanced,0.036090667049090065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,64,balanced,0.035904000202814736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,64,balanced,0.0366239994764328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,64,balanced,0.03984533250331879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,64,balanced,0.03777066618204117
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,64,balanced,0.0378560001651446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,64,balanced,0.03793599953254064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,64,balanced,0.03813866774241129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,64,balanced,0.03949866692225138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,1,power_law_1.01,0.2568831920623779
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,64,balanced,0.038015998899936676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,64,balanced,0.0397119993964831
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,64,balanced,0.039861333866914116
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,64,balanced,0.03965333352486292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,64,balanced,0.04402133325735728
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,64,balanced,0.0422026664018631
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,64,balanced,0.04598933458328247
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,64,balanced,0.046021332343419395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,64,balanced,0.04786133269468943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,64,balanced,0.05026666820049286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,64,balanced,0.056277334690093994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,64,balanced,0.06496533254782359
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,1,power_law_1.01,0.43024640083312987
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,64,balanced,0.07251200079917908
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,64,balanced,0.0904853343963623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,64,balanced,0.10365866621335347
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,16,power_law_1.2,0.1527232050895691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,64,balanced,0.13056533535321554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,64,balanced,0.16077333688735962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,64,balanced,0.18970666329065958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,64,balanced,0.27053866783777875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,1,power_law_1.01,0.472544002532959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,64,balanced,0.4888906478881836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,1,power_law_1.01,0.4947199821472168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,16,power_law_1.2,0.16367360353469848
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,1,power_law_1.01,0.49073281288146975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,1,power_law_1.01,0.5026559829711914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,1,power_law_1.01,0.5356031894683838
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,16,power_law_1.2,0.20357120037078857
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,1,power_law_1.01,0.5421440124511718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,1,power_law_1.01,0.5382336139678955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,16,power_law_1.2,0.23016960620880128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,1,power_law_1.01,0.5674880027770997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,1,power_law_1.01,0.5884607791900635
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,1,power_law_1.01,0.5966976165771485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,16,power_law_1.2,0.3028863906860352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,1,power_law_1.01,0.6319808006286621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,1,power_law_1.01,0.6748991966247558
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,1,power_law_1.01,0.7220992088317871
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,16,power_law_1.2,0.3244352102279663
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,1,power_law_1.01,0.823692798614502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,4,power_law_1.01,0.10242559909820556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,1,power_law_1.01,0.8783743858337403
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,16,power_law_1.2,0.4375999927520752
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,1,power_law_1.01,1.0479104042053222
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,4,power_law_1.01,0.1148800015449524
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,1,power_law_1.01,1.2323519706726074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,4,power_law_1.01,0.09615359902381897
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,2,power_law_1.2,0.0709119975566864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,1,power_law_1.01,1.5791168212890625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,16,power_law_1.2,0.5681024074554444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,4,power_law_1.01,0.14191360473632814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,2,power_law_1.2,0.12172160148620606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,1,power_law_1.01,2.0615680694580076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,4,power_law_1.01,0.16239999532699584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,2,power_law_1.2,0.10888320207595825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,16,power_law_1.2,0.7776000022888183
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,1,power_law_1.01,2.7080255508422852
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,4,power_law_1.01,0.26245119571685793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,2,power_law_1.2,0.12169599533081055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,4,power_law_1.01,0.27047679424285886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,1,power_law_1.01,3.5465728759765627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,2,power_law_1.2,0.1766975998878479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,4,power_law_1.01,0.27575039863586426
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,16,power_law_1.2,0.9646592140197754
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,2,power_law_1.2,0.24414079189300536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,4,power_law_1.01,0.27779839038848875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,1,power_law_1.01,4.223961639404297
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,4,power_law_1.01,0.290393590927124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,2,power_law_1.2,0.29577600955963135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,4,power_law_1.01,0.2902656078338623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,2,power_law_1.2,0.4108863830566406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,4,power_law_1.01,0.3074944019317627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,1,power_law_1.01,5.990899276733399
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,2,16,power_law_1.2,1.2641023635864257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,4,power_law_1.01,0.2990272045135498
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,2,power_law_1.2,0.4212800025939941
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,4,power_law_1.01,0.3122368097305298
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,2,power_law_1.2,0.43303680419921875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,4,power_law_1.01,0.3272831916809082
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,2,power_law_1.2,0.45365118980407715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,4,power_law_1.01,0.3403520107269287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,2,16,power_law_1.2,1.8697919845581055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,1,power_law_1.01,11.36714859008789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,4,power_law_1.01,0.35329279899597166
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,2,power_law_1.2,0.47178239822387696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,4,power_law_1.01,0.3829119920730591
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,2,power_law_1.2,0.49822077751159666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,4,power_law_1.01,0.41818881034851074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,2,power_law_1.2,0.5076735973358154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,4,power_law_1.01,0.508019208908081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,4,power_law_1.01,0.5833856105804444
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,2,power_law_1.2,0.5559807777404785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,4,power_law_1.01,0.6805952072143555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,2,power_law_1.2,0.5664896011352539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,4,power_law_1.01,0.7505087852478027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,2,power_law_1.2,0.6040128231048584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,2,16,power_law_1.2,3.9008319854736326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,4,power_law_1.01,0.9365504264831543
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,2,power_law_1.2,0.6648960113525391
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,4,power_law_1.01,1.1829119682312013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,2,power_law_1.2,0.6806272029876709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,4,power_law_1.01,1.6634559631347656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,2,power_law_1.2,0.7868800163269043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,4,power_law_1.01,2.197305679321289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,32,power_law_1.01,0.0389631986618042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,2,power_law_1.2,0.8748096466064453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,32,power_law_1.01,0.03550719916820526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,32,power_law_1.01,0.028966400027275085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,32,power_law_1.01,0.029478400945663452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,4,power_law_1.01,2.368454360961914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,2,power_law_1.2,1.055788803100586
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,32,power_law_1.01,0.03025279939174652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,32,power_law_1.01,0.031327998638153075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,32,power_law_1.01,0.03195520043373108
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,2,power_law_1.2,1.2276479721069335
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,32,power_law_1.01,0.03255040049552917
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,32,power_law_1.01,0.03270399868488312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,4,power_law_1.01,3.623155212402344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,32,power_law_1.01,0.03351680040359497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,32,power_law_1.01,0.03441280126571655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,2,power_law_1.2,1.5876416206359862
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,32,power_law_1.01,0.03529599905014038
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,32,power_law_1.01,0.03763200044631958
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,32,power_law_1.01,0.03535360097885132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,2,power_law_1.2,1.8224191665649414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,32,power_law_1.01,0.03982079923152924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,16,power_law_1.01,0.04348799884319306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,32,power_law_1.01,0.04033919870853424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,16,power_law_1.01,0.03578880131244659
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,32,power_law_1.01,0.04481280148029328
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,4,power_law_1.01,7.089107513427734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,16,power_law_1.01,0.033497598767280576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,32,power_law_1.01,0.0494592010974884
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,2,power_law_1.2,2.405171203613281
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,32,power_law_1.01,0.05231999754905701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,16,power_law_1.01,0.03313280045986176
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,32,power_law_1.01,0.05854079723358154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,16,power_law_1.01,0.034867200255393985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,32,power_law_1.01,0.06442239880561829
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,16,power_law_1.01,0.03585279881954193
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,32,power_law_1.01,0.08001279830932617
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,16,power_law_1.01,0.03591040074825287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,2,power_law_1.2,2.8939840316772463
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,32,power_law_1.01,0.1021888017654419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,16,power_law_1.01,0.036025598645210266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,16,power_law_1.01,0.03580799996852875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,32,power_law_1.01,0.13162239789962768
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,16,power_law_1.01,0.03651840090751648
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,32,power_law_1.01,0.15907200574874877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,16,power_law_1.01,0.03681280016899109
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,32,power_law_1.01,0.20269439220428467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,2,power_law_1.2,3.369087982177734
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,16,power_law_1.01,0.03768959939479828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,16,power_law_1.01,0.03848319947719574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,32,power_law_1.01,0.2595455884933472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,16,power_law_1.01,0.04010879993438721
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,16,power_law_1.01,0.0436928004026413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,2688,1856,6,128,1,32,power_law_1.01,0.31376640796661376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,16,power_law_1.01,0.044896000623703004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,16,power_law_1.01,0.04748159945011139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,2688,1856,6,128,1,32,power_law_1.01,0.46705918312072753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,2,power_law_1.2,5.291187286376953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,16,power_law_1.01,0.052691197395324706
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,16,power_law_1.01,0.05459200143814087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,16,power_law_1.01,0.06119040250778198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,16,power_law_1.01,0.07215999960899352
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,2688,1856,6,128,1,32,power_law_1.01,0.9942975997924804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,16,power_law_1.01,0.08469759821891784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,16,power_law_1.01,0.09570559859275818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,16,power_law_1.01,0.13826559782028197
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,1,power_law_1.01,0.03464959859848023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,16,power_law_1.01,0.17147519588470458
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,2,power_law_1.2,10.255052947998047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,1,power_law_1.01,0.03943040072917938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,16,power_law_1.01,0.23559041023254396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,1,power_law_1.01,0.05025920271873474
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,16,power_law_1.01,0.34128000736236574
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,1,power_law_1.01,0.06600959897041321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,16,power_law_1.01,0.3810751914978027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,1,power_law_1.01,0.07960960268974304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,1,power_law_1.01,0.11407359838485717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,16,power_law_1.01,0.6306496143341065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,1,power_law_1.01,0.12036479711532592
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,1,power_law_1.01,0.1250815987586975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,1,power_law_1.01,0.12662400007247926
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,16,power_law_1.01,1.13307523727417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,1,power_law_1.01,0.12759679555892944
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,1,power_law_1.01,0.134662401676178
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,1,power_law_1.01,0.14309120178222656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,16,2,power_law_1.2,0.031488001346588135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,16,2,power_law_1.2,0.03199360072612763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,16,2,power_law_1.2,0.031123200058937074
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,1,power_law_1.01,0.14339200258255005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,16,2,power_law_1.2,0.033107200264930726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,16,2,power_law_1.2,0.033497598767280576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,16,2,power_law_1.2,0.03470079898834229
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,16,2,power_law_1.2,0.03883520066738129
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,1,power_law_1.01,0.14984960556030275
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,16,2,power_law_1.2,0.03903999924659729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,16,2,power_law_1.2,0.04116480052471161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,16,2,power_law_1.2,0.042387199401855466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,16,2,power_law_1.2,0.04386560022830963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,16,2,power_law_1.2,0.047007998824119566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,1,power_law_1.01,0.1588544011116028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,16,2,power_law_1.2,0.04789760112762451
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,16,2,power_law_1.2,0.050393599271774295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,16,2,power_law_1.2,0.06155520081520081
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,16,2,power_law_1.2,0.06598399877548218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,1,power_law_1.01,0.16647679805755616
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,32,balanced,0.048207998275756836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,16,2,power_law_1.2,0.07025279998779296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,32,balanced,0.05613866448402405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,32,balanced,0.06033066908518473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,16,2,power_law_1.2,0.08951680064201355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,32,balanced,0.08866666754086812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,32,balanced,0.10656533638636272
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,32,balanced,0.10751466949780782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,32,balanced,0.10719466209411621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,16,2,power_law_1.2,0.10172159671783447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,32,balanced,0.10687999924023946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,32,balanced,0.11002666751543681
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,1,power_law_1.01,0.17970559597015381
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,32,balanced,0.10890133182207744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,16,2,power_law_1.2,0.12596479654312134
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,32,balanced,0.11001066366831462
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,32,balanced,0.11201600233713786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,32,balanced,0.11154133081436157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,16,2,power_law_1.2,0.147161602973938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,32,balanced,0.11494400103886922
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,32,balanced,0.11840533216794331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,32,balanced,0.11948800086975098
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,32,balanced,0.12506666779518127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,16,2,power_law_1.2,0.19459199905395508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,32,balanced,0.14518400033315024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,32,balanced,0.14499200383822122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,32,balanced,0.17890665928522745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,16,2,power_law_1.2,0.2328255891799927
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,1,power_law_1.01,0.20655999183654786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,32,balanced,0.18588266770044962
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,32,balanced,0.2639946738878886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,32,balanced,0.24119466543197632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,16,2,power_law_1.2,0.3513535976409912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,32,balanced,0.31547733147939044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,32,balanced,0.3563466469446818
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,16,2,power_law_1.2,0.4494016170501709
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,32,balanced,0.4975626468658447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,1,power_law_1.01,0.2285696029663086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,32,balanced,0.5827680031458536
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,16,2,power_law_1.2,0.6535168170928956
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,1,32,balanced,0.7188533147176107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,1,32,balanced,1.0448266665140789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,16,2,power_law_1.2,0.8477631568908691
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,1,power_law_1.01,0.28273279666900636
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,1,32,balanced,2.1455839474995932
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,16,2,power_law_1.2,0.9639103889465332
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,1,power_law_1.01,0.3223680019378662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,16,2,power_law_1.2,1.5660991668701172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,1,power_law_1.01,0.42538881301879883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,16,2,power_law_1.2,3.3550273895263674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,1,power_law_1.01,0.5154111862182618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,8,balanced,0.05007466673851013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,8,balanced,0.051872000098228455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,1,power_law_1.01,0.7087103843688964
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,8,balanced,0.0539626677831014
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,8,balanced,0.07778666913509369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,8,balanced,0.09406399726867676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,8,balanced,0.10818666219711304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,8,balanced,0.10936533411343892
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,8,balanced,0.11152533690134685
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,8,balanced,0.11124799648920695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,8,balanced,0.11292800307273865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,8,balanced,0.11292800307273865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,8,balanced,0.11391466856002808
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,8,balanced,0.11552000045776367
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,8,balanced,0.11531733473141988
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,8,balanced,0.12098667025566101
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,8,balanced,0.12481066584587097
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,8,balanced,0.12878400087356567
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,8,balanced,0.16022400061289468
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,1,power_law_1.01,0.8930047988891602
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,8,balanced,0.15621866782506308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,8,balanced,0.2020906607309977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,8,balanced,0.2160266637802124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,8,balanced,0.29812800884246826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,8,balanced,0.29547733068466187
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,8,balanced,0.4241280158360799
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,8,balanced,0.4718986749649048
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,8,balanced,0.6775733629862467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,1,power_law_1.01,1.262502384185791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,8,balanced,0.8152053356170654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,8,8,balanced,1.03002134958903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,8,8,balanced,1.5079785982767742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,8,8,balanced,3.0750293731689453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,1,power_law_1.01,1.6436479568481446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,1,power_law_1.01,2.0227840423583983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,1,power_law_1.01,3.0595584869384767
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,4,balanced,0.05299200117588043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,4,balanced,0.05444266895453135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,4,balanced,0.06676800052324931
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,4,balanced,0.09101866682370503
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,4,balanced,0.1264586647351583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,4,balanced,0.14013866583506265
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,4,balanced,0.14006933569908142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,4,balanced,0.14417599638303122
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,4,balanced,0.14361600081125894
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,4,balanced,0.14245866735776266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,4,balanced,0.1451359987258911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,4,balanced,0.1479520003000895
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,4,balanced,0.1473973294099172
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,4,balanced,0.15246933698654175
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,4,balanced,0.1581653356552124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,4,balanced,0.1611199975013733
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,4,balanced,0.1673226753870646
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,4,balanced,0.17750400304794312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,4,balanced,0.1907306710879008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,4,balanced,0.2521119912465413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,1,power_law_1.01,5.900364685058594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,4,balanced,0.24049067497253418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,4,balanced,0.3228106697400411
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,4,balanced,0.3484213352203369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,4,balanced,0.48369065920511883
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,4,balanced,0.5844800074895223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,4,balanced,0.8080639839172363
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,4,balanced,1.049280007680257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,4,balanced,1.2449920177459717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,4,balanced,1.9593706130981445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,1,power_law_1.2,0.09941120147705078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,4,balanced,3.732224146525065
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,1,power_law_1.2,0.14168319702148438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,1,power_law_1.2,0.1453503966331482
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,16,2,power_law_1.2,0.03632639944553375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,16,2,power_law_1.2,0.036422398686408994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,1,power_law_1.2,0.23488640785217285
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,16,2,power_law_1.2,0.03627519905567169
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,16,2,power_law_1.2,0.03717760145664215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,16,2,power_law_1.2,0.037567999958992
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,16,2,power_law_1.2,0.03849599957466125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,16,2,power_law_1.2,0.03866240084171295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,1,power_law_1.2,0.31420159339904785
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,16,2,power_law_1.2,0.03938559889793396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,16,2,power_law_1.2,0.03984639942646027
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,16,2,power_law_1.2,0.040863999724388124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,16,2,power_law_1.2,0.040191999077796935
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,16,2,power_law_1.2,0.043059200048446655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,16,2,power_law_1.2,0.04407680034637451
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,1,power_law_1.2,0.44471039772033694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,16,2,power_law_1.2,0.04826239943504333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,16,2,power_law_1.2,0.053964799642562865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,16,2,power_law_1.2,0.057734400033950806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,16,2,power_law_1.2,0.05867519974708557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,16,2,power_law_1.2,0.07243520021438599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,1,power_law_1.2,0.6992576122283936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,16,2,power_law_1.2,0.07942399978637696
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,16,2,power_law_1.2,0.10502400398254394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,16,2,power_law_1.2,0.11605119705200195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,16,2,power_law_1.2,0.172326397895813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,1,power_law_1.2,0.7420415878295898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,16,2,power_law_1.2,0.18458240032196044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,16,2,power_law_1.2,0.2504767894744873
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,16,2,power_law_1.2,0.293286395072937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,1,power_law_1.2,0.7770880222320556
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,16,2,power_law_1.2,0.4949952125549316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,16,2,power_law_1.2,0.5363967895507813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,16,2,power_law_1.2,0.7963583946228028
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,1,power_law_1.2,0.7986048221588135
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,16,2,power_law_1.2,1.2591103553771972
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,1,power_law_1.2,0.8422271728515625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,2048,768,8,128,2,64,power_law_1.2,0.02311680018901825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,16,2,power_law_1.2,2.1508224487304686
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,2048,768,8,128,2,64,power_law_1.2,0.02401279956102371
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,1,power_law_1.2,0.8799039840698242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,2048,768,8,128,2,64,power_law_1.2,0.02462079972028732
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,1,power_law_1.2,0.8932031631469727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,2048,768,8,128,2,64,power_law_1.2,0.024851199984550477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,4,4,power_law_1.2,0.042310398817062375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,4,4,power_law_1.2,0.03592320084571839
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,2048,768,8,128,2,64,power_law_1.2,0.02513279914855957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,1,power_law_1.2,0.933619213104248
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,4,4,power_law_1.2,0.035385599732398985
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,4,4,power_law_1.2,0.03849599957466125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,1,power_law_1.2,0.9706432342529296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,2048,768,8,128,2,64,power_law_1.2,0.026502400636672974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,4,4,power_law_1.2,0.03982079923152924
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,4,4,power_law_1.2,0.044249600172042845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,1,power_law_1.2,0.9857088088989258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,2048,768,8,128,2,64,power_law_1.2,0.026291200518608095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,4,4,power_law_1.2,0.045228800177574156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,4,4,power_law_1.2,0.04493440091609955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,1,power_law_1.2,1.0151040077209472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,2048,768,8,128,2,64,power_law_1.2,0.029151999950408937
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,4,4,power_law_1.2,0.04625279903411865
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,4,4,power_law_1.2,0.046028798818588255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,1,power_law_1.2,1.1370752334594727
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,2048,768,8,128,2,64,power_law_1.2,0.03049600124359131
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,4,4,power_law_1.2,0.0472896009683609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,4,4,power_law_1.2,0.049107199907302855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,2048,768,8,128,2,64,power_law_1.2,0.03027839958667755
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,1,power_law_1.2,1.1970368385314942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,4,4,power_law_1.2,0.04995200037956238
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,2048,768,8,128,2,64,power_law_1.2,0.030527999997138976
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,4,4,power_law_1.2,0.05275520086288452
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,1,power_law_1.2,1.3729087829589843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,4,4,power_law_1.2,0.060083198547363284
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,2048,768,8,128,2,64,power_law_1.2,0.030720001459121703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,4,4,power_law_1.2,0.062412798404693604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,1,power_law_1.2,1.4511615753173828
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,2048,768,8,128,2,64,power_law_1.2,0.03128960132598877
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,4,4,power_law_1.2,0.06220160126686096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,4,4,power_law_1.2,0.06937599778175355
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,2048,768,8,128,2,64,power_law_1.2,0.03052160143852234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,32,balanced,0.053599998354911804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,4,4,power_law_1.2,0.0849407970905304
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,32,balanced,0.053898667295773826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,1,power_law_1.2,1.6906496047973634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,32,balanced,0.05622399846712748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,32,balanced,0.06222933530807495
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,32,balanced,0.07083199918270111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,32,balanced,0.07277333239714305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,32,balanced,0.07196266452471416
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,32,balanced,0.07166400055090587
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,32,balanced,0.07438399891058604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,2048,768,8,128,2,64,power_law_1.2,0.03351039886474609
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,32,balanced,0.07270933190981548
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,4,4,power_law_1.2,0.096288001537323
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,32,balanced,0.07506133119265239
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,32,balanced,0.07843199868996938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,32,balanced,0.0777759999036789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,32,balanced,0.081194669008255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,32,balanced,0.08543999989827473
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,32,balanced,0.08719467123349507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,32,balanced,0.09340266386667888
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,32,balanced,0.10672533512115479
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,32,balanced,0.11475200454394023
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,4,4,power_law_1.2,0.11173759698867798
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,32,balanced,0.14221866925557455
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,32,balanced,0.16190933187802634
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,2048,768,8,128,2,64,power_law_1.2,0.03424000144004822
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,32,balanced,0.21422400077184042
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,32,balanced,0.2049600084622701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,1,power_law_1.2,1.9650495529174805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,32,balanced,0.2633919914563497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,32,balanced,0.30471465984980267
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,4,4,power_law_1.2,0.14176000356674195
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,32,balanced,0.41703999042510986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,32,balanced,0.4996853272120158
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,2048,768,8,128,2,64,power_law_1.2,0.035411199927330016
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,20480,4096,2688,22,512,2,32,balanced,0.6092480023701986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,4,4,power_law_1.2,0.18896000385284423
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,32768,4096,2688,22,512,2,32,balanced,0.9091359774271647
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,2048,768,8,128,2,64,power_law_1.2,0.038150399923324585
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,1,power_law_1.2,2.5421695709228516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,nvfp4,65536,4096,2688,22,512,2,32,balanced,1.895898660024007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,4,4,power_law_1.2,0.2415424108505249
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,64,power_law_1.2,0.04524160027503967
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,4,4,power_law_1.2,0.33597440719604493
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,64,power_law_1.2,0.04826239943504333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,1,power_law_1.2,3.0107328414916994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,4,4,power_law_1.2,0.4744448184967041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,64,power_law_1.2,0.05336959958076477
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,4,4,power_law_1.2,0.6907519817352294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,64,power_law_1.2,0.06218879818916321
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,1,power_law_1.2,4.062995147705078
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,64,power_law_1.2,0.07352319955825806
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,4,4,power_law_1.2,0.9071616172790528
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,64,power_law_1.2,0.09751679897308349
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,4,4,power_law_1.2,1.290060806274414
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,64,power_law_1.2,0.1356608033180237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,1,power_law_1.2,5.071091079711914
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,64,power_law_1.2,0.223091197013855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,64,power_law_1.2,0.24808320999145508
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,4,4,power_law_1.2,2.8042688369750977
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,1,power_law_1.2,6.0980224609375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,64,power_law_1.2,0.27682559490203856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,64,power_law_1.2,0.6053055763244629
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,8,2,balanced,0.03355200091997782
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,8,2,balanced,0.039664000272750854
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,8,2,balanced,0.05994133154551188
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,8,2,balanced,0.055311997731526695
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,8,2,balanced,0.05663999915122986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,8,2,balanced,0.05584533512592316
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,8,2,balanced,0.058320000767707825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,8,2,balanced,0.05788800120353699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,8,2,balanced,0.056890666484832764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,8,2,balanced,0.0589279979467392
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,8,2,balanced,0.060032000144322716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,8,2,balanced,0.06309866905212402
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,1,power_law_1.2,9.016806030273438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,8,2,balanced,0.06280000011126201
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,8,2,balanced,0.06494933366775513
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,8,2,balanced,0.0765173335870107
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,8,2,balanced,0.0766186664501826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,8,2,balanced,0.08148266871770223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,8,2,balanced,0.08966933687527974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,8,2,balanced,0.09963732957839966
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,8,2,balanced,0.11974400281906128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,8,2,balanced,0.15330666303634644
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,8,2,balanced,0.19553067286809286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,8,2,balanced,0.23468265930811563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,8,2,balanced,0.31856000423431396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,8,2,balanced,0.39558398723602295
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,8,2,balanced,0.5685493151346842
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,8,2,balanced,0.7275520165761312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,8,2,balanced,0.8908533255259196
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,64,power_law_1.2,1.0209792137145997
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,8,2,balanced,1.379733403523763
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,8,2,balanced,2.686234792073568
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,1,power_law_1.2,16.843423461914064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,2,1,balanced,0.11959999799728394
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,2,1,balanced,0.2751680016517639
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,2,1,balanced,0.38653333981831867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,2,1,balanced,0.3803146680196126
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,2,1,balanced,0.381061315536499
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,2,1,balanced,0.37811732292175293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,2,1,balanced,0.38014400005340576
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,2,1,balanced,0.38018667697906494
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,2,1,balanced,0.38174935181935626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,2,1,balanced,0.3831839958826701
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,2,1,balanced,0.3903199831644694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,2,1,balanced,0.3922453324000041
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,2,1,balanced,0.3932960033416748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,2,1,balanced,0.4070880015691121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,2,1,balanced,0.43427733580271405
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,2,1,balanced,0.4376320044199626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,2,1,balanced,0.4376000165939331
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,2,1,balanced,0.5560586849848429
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,2,1,balanced,0.5386346578598022
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,2,1,balanced,0.7136693000793457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,2,1,balanced,0.7453227043151855
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,2,1,balanced,0.9931519826253256
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,2,1,balanced,1.261093298594157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,2,1,balanced,1.8447413444519043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,2,1,balanced,2.408474604288737
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,2,1,balanced,3.570687929789225
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,32,balanced,0.053930665055910744
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,32,balanced,0.05022400120894114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,32,balanced,0.04837333162625631
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,32,balanced,0.051632001996040344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,32,balanced,0.05259199937184652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,32,balanced,0.05444799860318502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,32,balanced,0.05669333537419637
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,2,1,balanced,4.785898526509603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,32,balanced,0.05604266623655955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,32,balanced,0.05622399846712748
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,32,balanced,0.056608001391092934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,32,balanced,0.05856533348560333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,32,balanced,0.05830933153629303
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,32,balanced,0.05990933378537496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,32,balanced,0.060271998246510826
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,32,balanced,0.06457599997520447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,32,balanced,0.06713599960009257
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,32,balanced,0.06697600086530049
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,32,balanced,0.07479466497898102
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,32,balanced,0.07962133487065633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,32,balanced,0.09098133444786072
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,2,1,balanced,5.962064107259114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,32,balanced,0.10538132985432942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,32,balanced,0.13111467162768045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,32,balanced,0.15060266852378845
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,16,power_law_1.2,0.10257279872894287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,32,balanced,0.19891732931137085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,32,balanced,0.2445653279622396
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,32,balanced,0.33312533299128216
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,32,balanced,0.42003198464711505
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,16,power_law_1.2,0.05957760214805603
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,32,balanced,0.5167680184046427
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,32,balanced,0.7627786795298258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,2,1,balanced,9.686794916788736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,16,power_law_1.2,0.0654528021812439
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,32,balanced,1.4986079533894856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,16,power_law_1.2,0.07971199750900268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,16,power_law_1.2,0.07813760042190551
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,4,balanced,0.03150933235883713
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,4,balanced,0.031386665999889374
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,4,balanced,0.031397332747777305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,4,balanced,0.031680000325044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,4,balanced,0.037445334096749626
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,16,power_law_1.2,0.10646400451660157
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,4,balanced,0.038047999143600464
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,4,balanced,0.03752533346414566
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,4,balanced,0.03788266579310099
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,4,balanced,0.03755199909210205
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,4,balanced,0.03955733279387156
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,4,balanced,0.039594667653242745
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,4,balanced,0.0397173340121905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,4,balanced,0.04186133543650309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,4,balanced,0.041802664597829185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,4,balanced,0.050053333242734276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,4,balanced,0.04801600178082784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,4,balanced,0.04832000037034353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,4,balanced,0.05603733162085215
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,4,balanced,0.0602453351020813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,16,power_law_1.2,0.10955519676208496
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,4,balanced,0.07221866647402446
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,4,balanced,0.08072533210118611
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,4,balanced,0.10995200276374817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,4,balanced,0.1395146648089091
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,4,balanced,0.1894986629486084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,4,balanced,0.2309653361638387
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,2,1,balanced,18.988975524902344
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,4,balanced,0.32850666840871173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,4,balanced,0.4296160141626994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,16,power_law_1.2,0.10703359842300415
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,4,balanced,0.5337333281834921
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,4,balanced,0.8377493222554525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,4,balanced,1.6686293284098308
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,16,power_law_1.2,0.11263999938964844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,16,power_law_1.2,0.1141055941581726
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,16,power_law_1.2,0.11638400554656983
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,16,power_law_1.2,0.12153600454330445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,16,power_law_1.2,0.12062079906463623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,16,power_law_1.2,0.1275264024734497
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,16,power_law_1.2,0.13929599523544312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,16,power_law_1.2,0.1414720058441162
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,16,power_law_1.2,0.15905920267105103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,16,power_law_1.2,0.19044480323791504
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,16,power_law_1.2,0.210534405708313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,16,power_law_1.2,0.27137279510498047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,4,power_law_1.2,0.085343998670578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,16,power_law_1.2,0.2492095947265625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,16,power_law_1.2,0.3178688049316406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,4,power_law_1.2,0.07256320118904114
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,16,power_law_1.2,0.38915839195251467
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,4,power_law_1.2,0.0832319974899292
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,16,power_law_1.2,0.5241407871246337
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,4,power_law_1.2,0.10837759971618652
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,16,power_law_1.2,0.6701759815216064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,4,power_law_1.2,0.1292736053466797
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,2,power_law_1.01,0.0883520007133484
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,16,power_law_1.2,0.9986368179321289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,1,power_law_1.01,0.07197440266609192
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,4,power_law_1.2,0.17560319900512694
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,2,power_law_1.01,0.12060799598693847
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,1,power_law_1.01,0.09701120257377624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,4,power_law_1.2,0.18458880186080934
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,16,power_law_1.2,1.3585472106933594
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,2,power_law_1.01,0.09548799991607666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,1,power_law_1.01,0.10632319450378418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,4,power_law_1.2,0.19645440578460693
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,2,power_law_1.01,0.16165759563446044
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,16,power_law_1.2,1.4973567962646483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,4,power_law_1.2,0.19435520172119142
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,1,power_law_1.01,0.19409919977188111
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,2,power_law_1.01,0.22509438991546632
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,4,power_law_1.2,0.20635519027709961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,1,power_law_1.01,0.29742720127105715
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,2,power_law_1.01,0.2725248098373413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,4,power_law_1.2,0.2134848117828369
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,16,power_law_1.2,2.9984832763671876
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,1,power_law_1.01,0.40078082084655764
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,2,power_law_1.01,0.39336960315704345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,4,power_law_1.2,0.2157696008682251
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,1,power_law_1.01,0.5691391944885253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,2,power_law_1.01,0.40609278678894045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,4,power_law_1.2,0.2153856039047241
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,1,power_law_1.01,0.6153664112091064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,2,power_law_1.01,0.4335936069488525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,4,power_law_1.2,0.22506239414215087
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,1,power_law_1.01,0.6482687950134277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,2,power_law_1.01,0.4240447998046875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,16,power_law_1.2,5.692780685424805
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,4,power_law_1.2,0.23031039237976075
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,1,power_law_1.01,0.6682176113128662
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,2,power_law_1.01,0.45505919456481936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,4,power_law_1.2,0.2387455940246582
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,1,power_law_1.01,0.6816832065582276
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,2,power_law_1.01,0.4717376232147217
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,4,power_law_1.2,0.25075840950012207
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,1,power_law_1.01,0.7204224109649658
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,2,power_law_1.01,0.46749439239501955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,4,power_law_1.2,0.27701120376586913
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,2,power_law_1.01,0.4901951789855957
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,1,power_law_1.01,0.7284736156463623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,4,power_law_1.2,0.3112384080886841
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,2,power_law_1.01,0.4991424083709717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,1,power_law_1.01,0.7484479904174804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,4,power_law_1.2,0.36446080207824705
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,2,power_law_1.01,0.5242559909820557
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,1,power_law_1.01,0.7876416206359863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,4,power_law_1.2,0.4079808235168457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,2,power_law_1.01,0.537004804611206
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,1,power_law_1.01,0.7781568050384522
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,4,power_law_1.2,0.5303743839263916
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,2,power_law_1.01,0.5713471889495849
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,1,power_law_1.01,0.8444671630859375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,4,power_law_1.2,0.6156415939331055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,2,power_law_1.01,0.6220543861389161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,1,power_law_1.01,0.9469696044921875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,2,power_law_1.01,0.694163179397583
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,4,power_law_1.2,0.8325759887695312
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,1,power_law_1.01,0.9537535667419433
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,2,power_law_1.01,0.7459904193878174
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,1,power_law_1.01,1.1015487670898438
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,4,power_law_1.2,1.0789759635925293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,2,power_law_1.01,0.9000639915466309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,1,power_law_1.01,1.1737536430358886
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,2,power_law_1.01,1.0238783836364747
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,1,power_law_1.01,1.3837632179260253
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,4,power_law_1.2,1.4756287574768066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,16,power_law_1.01,0.06791039705276489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,2,power_law_1.01,1.4131391525268555
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,1,power_law_1.01,1.652729606628418
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,16,power_law_1.01,0.09106559753417968
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,4,power_law_1.2,1.8974336624145507
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,2,power_law_1.01,1.6564863204956055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,1,power_law_1.01,2.110099220275879
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,16,power_law_1.01,0.053830397129058835
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,2,power_law_1.01,2.27258243560791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,1,power_law_1.01,2.4892095565795898
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,4,power_law_1.2,2.245254325866699
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,16,power_law_1.01,0.05873280167579651
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,2,power_law_1.01,2.867955207824707
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,1,power_law_1.01,3.369945526123047
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,4,power_law_1.2,3.6105857849121095
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,16,power_law_1.01,0.06151679754257202
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,2,power_law_1.01,3.498099136352539
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,1,power_law_1.01,4.3279296875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,16,power_law_1.01,0.07304319739341736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,16,power_law_1.01,0.08008319735527039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,1,power_law_1.01,5.21385612487793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,2,power_law_1.01,5.192262268066406
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,4,power_law_1.2,6.601248168945313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,16,power_law_1.01,0.07768959999084472
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,16,power_law_1.01,0.08183040022850037
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,1,power_law_1.01,7.620313262939453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,16,power_law_1.01,0.08069120049476623
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,2,power_law_1.01,9.431737518310547
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,16,power_law_1.01,0.08455680012702942
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,16,power_law_1.01,0.08727679848670959
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,1,power_law_1.01,14.42108154296875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,16,power_law_1.01,0.08921599984169007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,16,power_law_1.01,0.08982399702072144
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,16,power_law_1.01,0.09950079917907714
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,4,power_law_1.2,0.050310397148132326
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,16,power_law_1.01,0.10368640422821045
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,4,power_law_1.2,0.048403200507164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,4,power_law_1.2,0.04748159945011139
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,16,power_law_1.01,0.10395519733428955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,4,power_law_1.2,0.05910400152206421
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,4,power_law_1.2,0.06664320230484008
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,16,power_law_1.01,0.11931519508361817
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,128,power_law_1.01,0.04778240025043488
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,4,power_law_1.2,0.08803840279579163
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,4,power_law_1.2,0.0906112015247345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,16,power_law_1.01,0.1331455945968628
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,128,power_law_1.01,0.05504639744758606
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,4,power_law_1.2,0.09342079758644103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,4,power_law_1.2,0.09143679738044738
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,16,power_law_1.01,0.1628224015235901
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,128,power_law_1.01,0.04799999892711639
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,4,power_law_1.2,0.09557120203971863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,2,power_law_1.2,0.07080960273742676
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,4,power_law_1.2,0.09764479994773864
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,16,power_law_1.01,0.196396803855896
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,128,power_law_1.01,0.05073919892311096
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,2,power_law_1.2,0.11137280464172364
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,4,power_law_1.2,0.10173439979553223
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,4,power_law_1.2,0.1045375943183899
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,16,power_law_1.01,0.25508480072021483
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,2,power_law_1.2,0.0991104006767273
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,128,power_law_1.01,0.04947839975357056
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,4,power_law_1.2,0.10489599704742432
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,4,power_law_1.2,0.11228799819946289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,2,power_law_1.2,0.11352959871292115
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,128,power_law_1.01,0.05123839974403381
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,4,power_law_1.2,0.11861759424209595
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,16,power_law_1.01,0.28354558944702146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,2,power_law_1.2,0.15077760219573974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,4,power_law_1.2,0.12444800138473511
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,128,power_law_1.01,0.052211201190948485
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,4,power_law_1.2,0.14789119958877564
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,2,power_law_1.2,0.1933632016181946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,16,power_law_1.01,0.385260796546936
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,4,power_law_1.2,0.16062719821929933
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,128,power_law_1.01,0.052697598934173584
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,2,power_law_1.2,0.22634239196777345
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,4,power_law_1.2,0.19212160110473633
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,4,power_law_1.2,0.22522239685058593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,2,power_law_1.2,0.2868607997894287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,16,power_law_1.01,0.4450943946838379
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,128,power_law_1.01,0.052025598287582395
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,4,power_law_1.2,0.27524480819702146
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,2,power_law_1.2,0.29871358871459963
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,4,power_law_1.2,0.33228800296783445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,128,power_law_1.01,0.05306879878044128
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,4,power_law_1.2,0.45488638877868653
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,2,power_law_1.2,0.31503360271453856
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,16,power_law_1.01,0.6268159866333007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,128,power_law_1.01,0.05545600056648255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,4,power_law_1.2,0.6468224048614502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,2,power_law_1.2,0.3303168058395386
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,128,power_law_1.01,0.05813120007514953
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,4,power_law_1.2,0.8792192459106445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,2,power_law_1.2,0.3505215883255005
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,16,power_law_1.01,0.7861311912536622
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,4,power_law_1.2,1.0597951889038086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,128,power_law_1.01,0.059334397315979004
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,2,power_law_1.2,0.34979839324951173
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,2,power_law_1.2,0.3733695983886719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,128,power_law_1.01,0.06060799956321716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,4,power_law_1.2,1.3932671546936035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,2,power_law_1.2,0.393126392364502
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,4,16,power_law_1.01,0.9650176048278809
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,128,power_law_1.01,0.06751360297203064
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,4,power_law_1.2,2.13253116607666
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,2,power_law_1.2,0.4148928165435791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,128,power_law_1.01,0.06873599886894226
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,2,power_law_1.2,0.43518719673156736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,128,power_law_1.01,0.07537919878959656
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,2,power_law_1.2,0.47870721817016604
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,4,power_law_1.2,4.936819076538086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,4,16,power_law_1.01,1.396883201599121
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,2,power_law_1.2,0.523795223236084
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,128,power_law_1.01,0.08865919709205627
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,2,power_law_1.2,0.6126143932342529
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,128,power_law_1.01,0.0945792019367218
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,2,power_law_1.2,0.6732863903045654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,128,power_law_1.01,0.11964160203933716
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,2,power_law_1.2,0.8348992347717286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,128,power_law_1.01,0.14017280340194702
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,2,power_law_1.2,0.9693568229675293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,4,16,power_law_1.01,3.010188865661621
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,128,power_law_1.01,0.1868607997894287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,2,power_law_1.2,1.2629631996154784
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,128,power_law_1.01,0.19864319562911986
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,2,power_law_1.2,1.4932095527648925
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,128,power_law_1.01,0.26910719871520994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,2,power_law_1.2,2.046713638305664
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,1,balanced,0.05649599929650625
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,1,balanced,0.06337066491444905
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,1,balanced,0.0825386643409729
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,1,balanced,0.11602666974067688
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,1,balanced,0.1723840037981669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,1,balanced,0.2850079933802287
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,1,balanced,0.2897813320159912
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,1,balanced,0.28654932975769043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,1,balanced,0.2889066735903422
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,1,balanced,0.29205334186553955
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,1,balanced,0.29527467489242554
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,1,balanced,0.30062933762868244
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,1,balanced,0.30646934111913043
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,1,balanced,0.3137600024541219
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,1,balanced,0.32469866673151654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,1,balanced,0.33727999528249103
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,2,power_law_1.2,2.643289566040039
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,128,power_law_1.01,0.323744010925293
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,1,balanced,0.36773868401845294
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,1,balanced,0.41097601254781085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,1,balanced,0.43936534722646076
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,1,balanced,0.514245351155599
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,1,balanced,0.5864479939142863
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,1,balanced,0.8009866873423258
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,1,balanced,0.9134026368459066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,1,balanced,1.3067039648691814
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,1,balanced,1.5957493782043457
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,1,balanced,2.3140320777893066
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,2,power_law_1.2,3.2403839111328123
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,128,power_law_1.01,0.43168001174926757
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,1,balanced,2.9491306940714517
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,1,balanced,3.746389389038086
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,1,balanced,5.632272084554036
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,8,balanced,0.03770133356253306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,8,balanced,0.03802666564782461
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,8,balanced,0.03770133356253306
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,8,balanced,0.03989866624275843
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,8,balanced,0.04804266492525736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,2,power_law_1.2,4.451308822631836
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,8,balanced,0.05011733373006185
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,8,balanced,0.05000533163547516
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,8,balanced,0.05002133548259735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,128,power_law_1.01,0.5436223983764649
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,8,balanced,0.050144001841545105
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,8,balanced,0.05197333296140035
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,8,balanced,0.051738664507865906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,8,balanced,0.052229334910710655
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,8,balanced,0.054133335749308266
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,8,balanced,0.05384533107280731
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,8,balanced,0.05859733124574026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,8,balanced,0.059290667374928795
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,8,balanced,0.06443200012048085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,8,balanced,0.0691840002934138
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,8,balanced,0.0831520011027654
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,8,balanced,0.09243733684221904
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,8,balanced,0.0985599954922994
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,8,balanced,0.12239467104276021
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,1,balanced,10.85098648071289
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,8,balanced,0.14146133263905844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,8,balanced,0.18311466773351034
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,8,balanced,0.2200266718864441
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,8,balanced,0.30523733297983807
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,8,balanced,0.38142398993174237
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,8,balanced,0.4676106770833333
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,8,balanced,0.7042506535847982
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,8,balanced,1.3337492942810059
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,20480,4096,2688,22,512,2,128,power_law_1.01,0.6675519943237305
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,2,power_law_1.2,9.10672607421875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,32768,4096,2688,22,512,2,128,power_law_1.01,1.0088959693908692
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,6144,16384,2,8,8,4,power_law_1.01,0.0353983998298645
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,8,power_law_1.2,0.07110400199890136
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,6144,16384,2,8,8,4,power_law_1.01,0.037651199102401736
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,6144,16384,2,8,8,4,power_law_1.01,0.03496319949626923
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,6144,16384,2,8,8,4,power_law_1.01,0.03498879969120026
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,6144,16384,2,8,8,4,power_law_1.01,0.033888000249862674
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,6144,16384,2,8,8,4,power_law_1.01,0.03845759928226471
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,6144,16384,2,8,8,4,power_law_1.01,0.041203200817108154
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,fp8,65536,4096,2688,22,512,2,128,power_law_1.01,2.0898303985595703
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,6144,16384,2,8,8,4,power_law_1.01,0.039577600359916684
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,6144,16384,2,8,8,4,power_law_1.01,0.042822399735450746
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,6144,16384,2,8,8,4,power_law_1.01,0.044556799530982974
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,6144,16384,2,8,8,4,power_law_1.01,0.04529919922351837
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,8,power_law_1.2,0.09824640154838563
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,6144,16384,2,8,8,4,power_law_1.01,0.04902400076389313
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,6144,16384,2,8,8,4,power_law_1.01,0.05200639963150024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,6144,16384,2,8,8,4,power_law_1.01,0.054841601848602296
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,6144,16384,2,8,8,4,power_law_1.01,0.06663039922714234
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,6144,16384,2,8,8,4,power_law_1.01,0.0686464011669159
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,6144,16384,2,8,8,4,power_law_1.01,0.07291520237922669
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,6144,16384,2,8,8,4,power_law_1.01,0.08480640053749085
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1,6144,16384,2,8,1,2,power_law_1.01,0.21001598834991456
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,8,4,power_law_1.01,0.0945472002029419
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,8,4,power_law_1.01,0.12209919691085816
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2,6144,16384,2,8,1,2,power_law_1.01,0.26990718841552735
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4,6144,16384,2,8,1,2,power_law_1.01,0.24164481163024903
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,8,4,power_law_1.01,0.1401471972465515
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8,6144,16384,2,8,1,2,power_law_1.01,0.3764800071716309
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,8,4,power_law_1.01,0.21703040599822998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16,6144,16384,2,8,1,2,power_law_1.01,0.37117440700531007
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,8,4,power_law_1.01,0.24449920654296875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32,6144,16384,2,8,1,2,power_law_1.01,0.3944639921188354
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,8,4,power_law_1.01,0.305676794052124
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,48,6144,16384,2,8,1,2,power_law_1.01,0.39463679790496825
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,8,4,power_law_1.01,0.47628159523010255
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,8,power_law_1.2,0.0887615978717804
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,64,6144,16384,2,8,1,2,power_law_1.01,0.4034111976623535
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,8,4,power_law_1.01,0.7078015804290771
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,80,6144,16384,2,8,1,2,power_law_1.01,0.4369344234466553
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,96,6144,16384,2,8,1,2,power_law_1.01,0.41603841781616213
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,8,4,power_law_1.01,0.9007103919982911
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,128,6144,16384,2,8,1,2,power_law_1.01,0.42165122032165525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,160,6144,16384,2,8,1,2,power_law_1.01,0.432147216796875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,8,4,power_law_1.01,1.0582847595214844
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,192,6144,16384,2,8,1,2,power_law_1.01,0.44387202262878417
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,256,6144,16384,2,8,1,2,power_law_1.01,0.4634751796722412
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,320,6144,16384,2,8,1,2,power_law_1.01,0.4793727874755859
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,8,4,power_law_1.01,1.8809280395507812
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,384,6144,16384,2,8,1,2,power_law_1.01,0.5057472229003906
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,8,power_law_1.2,0.07283200025558471
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,512,6144,16384,2,8,1,2,power_law_1.01,0.5471231937408447
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,768,6144,16384,2,8,1,2,power_law_1.01,0.6838592052459717
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1024,6144,16384,2,8,1,2,power_law_1.01,0.7612607955932618
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,8,4,power_law_1.01,3.545606231689453
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,1536,6144,16384,2,8,1,2,power_law_1.01,0.9027135848999024
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,2048,6144,16384,2,8,1,2,power_law_1.01,1.1139519691467286
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,3072,6144,16384,2,8,1,2,power_law_1.01,1.596895980834961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,8,power_law_1.2,0.08744320273399353
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,4096,6144,16384,2,8,1,2,power_law_1.01,1.9296064376831055
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,6144,6144,16384,2,8,1,2,power_law_1.01,2.904083251953125
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,8192,6144,16384,2,8,1,2,power_law_1.01,3.944736099243164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,8,power_law_1.2,0.11445120573043824
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,2,balanced,0.08861333131790161
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,2,balanced,0.13429866234461466
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,2,balanced,0.19393066565195718
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,2,balanced,0.3340746561686198
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,12288,6144,16384,2,8,1,2,power_law_1.01,6.820441436767578
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,2,balanced,0.5920159816741943
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,2,balanced,0.8407839934031168
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,8,power_law_1.2,0.12302080392837525
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,2,balanced,0.839850664138794
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,2,balanced,0.8489279747009277
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,2,balanced,0.8537546793619791
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,2,balanced,0.8578826586405436
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,2,balanced,0.8573173681894938
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,2,balanced,0.863264004389445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,2,balanced,0.8687946796417236
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,2,balanced,0.8803466955820719
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,16384,6144,16384,2,8,1,2,power_law_1.01,7.6495811462402346
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,2,balanced,0.894266684850057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,2,balanced,0.9039626916249593
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,2,balanced,0.92031462987264
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,8,power_law_1.2,0.1502400040626526
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,2,balanced,1.0040427049001057
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,2,balanced,0.9881227016448975
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,2,balanced,1.2877333164215088
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,2,balanced,1.1269013086954753
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,2,balanced,1.8605440457661946
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,2,balanced,1.3999253908793132
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,2,balanced,2.1568479537963867
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,8,power_law_1.2,0.15358079671859742
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,20480,6144,16384,2,8,1,2,power_law_1.01,11.031155395507813
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,2,balanced,2.104848066965739
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,2,balanced,3.1977173487345376
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,2,balanced,3.467616081237793
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,8,power_law_1.2,0.15971839427947998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,20480,4096,2688,22,512,2,2,balanced,4.498202641805013
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,32768,4096,2688,22,512,2,2,balanced,6.386117299397786
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,8,power_law_1.2,0.16268800497055053
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,32768,6144,16384,2,8,1,2,power_law_1.01,16.182342529296875
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,8,power_law_1.2,0.16380159854888915
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow_nongated,float16,65536,4096,2688,22,512,2,2,balanced,12.702229817708334
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,8,power_law_1.2,0.17568639516830445
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,8,power_law_1.2,0.1792448043823242
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,8,power_law_1.2,0.19221119880676268
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,float16,65536,6144,16384,2,8,1,2,power_law_1.01,33.2282958984375
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,8,power_law_1.2,0.1927296042442322
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,8,power_law_1.2,0.2082751989364624
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,8,power_law_1.2,0.23613440990447998
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,8,power_law_1.2,0.2648832082748413
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,8,power_law_1.2,0.3105792045593262
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,8,power_law_1.2,0.339897608757019
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,8,power_law_1.2,0.4532735824584961
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,8,power_law_1.2,0.4909183979034424
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,8,power_law_1.2,0.6360127925872803
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,8,power_law_1.2,0.7877312183380127
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,8,power_law_1.2,1.1202048301696776
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,8,power_law_1.2,1.3591168403625489
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,8,power_law_1.2,1.625503921508789
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,8,power_law_1.2,2.502841567993164
TRTLLM,1.2.0rc6.post3,NVIDIA GB300,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,8,power_law_1.2,5.187539291381836
