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
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,1,1,power_law_1.01,0.19991552352905273
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,1,1,power_law_1.01,0.29856319427490235
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,1,1,power_law_1.01,0.3491078567504883
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,1,1,power_law_1.01,0.6201196670532226
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,1,1,power_law_1.01,0.6236140823364258
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,1,1,power_law_1.01,0.4879961776733398
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,1,1,power_law_1.01,0.6262073516845703
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,1,1,power_law_1.01,0.615591697692871
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,1,1,power_law_1.01,0.6220921707153321
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,1,1,power_law_1.01,0.6335833740234376
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,1,1,power_law_1.01,0.655467529296875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,1,1,power_law_1.01,0.6697952270507812
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,1,1,power_law_1.01,0.7088646697998047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,1,1,power_law_1.01,0.7875878143310546
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,1,1,power_law_1.01,0.8540934753417968
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,1,1,power_law_1.01,0.9975052642822266
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,1,1,power_law_1.01,1.1594400024414062
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,1,1,power_law_1.01,1.5004499816894532
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,1,1,power_law_1.01,2.0795910644531252
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,1,1,power_law_1.01,2.7250021362304686
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,1,1,power_law_1.01,3.6013958740234373
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,1,1,power_law_1.01,5.307197265625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,1,1,power_law_1.01,7.634945678710937
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,1,1,power_law_1.01,11.473027343750001
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,1,1,power_law_1.01,15.2733642578125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,1,1,power_law_1.01,23.872900390625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,1,1,power_law_1.01,33.67424072265625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,1,1,power_law_1.01,0.12149503707885742
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,1,1,power_law_1.01,0.1665990447998047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,1,1,power_law_1.01,0.19361536026000975
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,1,1,power_law_1.01,0.26410240173339844
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,1,1,power_law_1.01,0.29353727340698244
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,1,1,power_law_1.01,0.3037324714660644
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,1,1,power_law_1.01,0.2979929542541504
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,1,1,power_law_1.01,0.2976633644104004
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,1,1,power_law_1.01,0.31517120361328127
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,1,1,power_law_1.01,0.31591487884521485
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,1,1,power_law_1.01,0.30413312911987306
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,1,1,power_law_1.01,0.30602432250976563
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,1,1,power_law_1.01,0.3184531211853027
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,1,1,power_law_1.01,0.3578451156616211
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,1,1,power_law_1.01,0.3766508865356445
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,1,1,power_law_1.01,0.38512577056884767
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,1,1,power_law_1.01,0.4087974548339844
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,1,1,power_law_1.01,0.4770438385009766
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,1,1,power_law_1.01,0.5805887985229492
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,1,1,power_law_1.01,0.7611641693115234
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,1,1,power_law_1.01,0.9725408172607422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,1,1,power_law_1.01,1.6008531188964845
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,1,1,power_law_1.01,42.630869140624995
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,1,1,power_law_1.01,1.8462290954589844
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,1,1,power_law_1.01,2.8874591064453123
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,1,1,power_law_1.01,3.973475341796875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,1,1,power_law_1.01,5.686737670898437
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,1,1,power_law_1.01,7.820128784179687
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,1,1,power_law_1.01,10.230301513671876
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,1,1,power_law_1.01,0.05467967987060547
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,1,1,power_law_1.01,0.1482323169708252
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,1,1,power_law_1.01,0.14818559646606447
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,1,1,power_law_1.01,0.14926976203918457
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,1,1,power_law_1.01,0.14772159576416016
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,1,1,power_law_1.01,0.14879103660583498
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,1,1,power_law_1.01,0.15112064361572267
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,1,1,power_law_1.01,0.15396672248840332
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,1,1,power_law_1.01,0.08041664123535155
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,1,1,power_law_1.01,0.10581184387207032
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,1,1,power_law_1.01,0.1455564785003662
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,1,1,power_law_1.01,0.15393856048583984
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,1,1,power_law_1.01,0.16187519073486328
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,1,1,power_law_1.01,0.1817638397216797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,1,1,power_law_1.01,0.1917407989501953
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,1,1,power_law_1.01,0.20204288482666016
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,1,1,power_law_1.01,0.22541248321533205
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,1,1,power_law_1.01,0.29352319717407227
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,1,1,power_law_1.01,0.35294143676757816
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,1,1,power_law_1.01,0.47047359466552735
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,1,1,power_law_1.01,0.5948204803466797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,1,1,power_law_1.01,0.8836627197265624
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,1,1,power_law_1.01,1.10487548828125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,1,1,power_law_1.01,1.6290559387207033
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,1,1,power_law_1.01,2.173315124511719
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,1,1,power_law_1.2,0.20315135955810545
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,1,1,power_law_1.2,0.29149951934814455
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,1,1,power_law_1.2,0.3151155281066894
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,1,1,power_law_1.2,0.4863840103149414
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,1,1,power_law_1.2,0.5877727890014649
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,1,1,power_law_1.2,0.6192236709594726
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,1,1,power_law_1.2,0.6222163009643554
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,1,1,power_law_1.2,0.6123276901245117
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,1,1,power_law_1.2,0.6275635147094727
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,1,1,power_law_1.2,0.631077766418457
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,1,1,power_law_1.2,0.665815658569336
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,1,1,power_law_1.2,0.6860300445556641
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,1,1,power_law_1.2,0.6954713439941406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,1,1,power_law_1.2,0.7765164947509765
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,1,1,power_law_1.2,0.8324614715576171
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,1,1,power_law_1.2,1.0341478729248048
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,1,1,power_law_1.2,1.1677881622314454
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,1,1,power_law_1.2,1.5214732360839842
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,1,1,power_law_1.2,1.9016978454589843
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,1,1,power_law_1.01,3.3039013671874997
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,1,1,power_law_1.2,2.643555297851562
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,1,1,power_law_1.2,3.478701477050781
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,1,1,power_law_1.2,5.263546752929687
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,1,1,power_law_1.01,4.379647216796875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,1,1,power_law_1.2,7.329217529296875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,1,1,power_law_1.2,11.424154052734375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,1,1,power_law_1.01,5.7150341796875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,1,1,power_law_1.2,15.472835693359375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,1,1,power_law_1.2,0.12129664421081543
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,1,1,power_law_1.2,0.1624825668334961
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,1,1,power_law_1.2,0.17788864135742188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,1,1,power_law_1.2,0.2650393676757813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,1,1,power_law_1.2,0.2922547149658203
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,1,1,power_law_1.2,0.2981292724609375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,1,1,power_law_1.2,0.30716543197631835
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,1,1,power_law_1.2,0.2981292724609375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,1,1,power_law_1.2,0.3096518325805664
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,1,1,power_law_1.2,0.3054252815246582
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,1,1,power_law_1.2,0.3149190330505371
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,1,1,power_law_1.2,0.3121536064147949
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,1,1,power_law_1.2,0.3192012786865234
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,1,1,power_law_1.2,0.35623233795166015
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,1,1,power_law_1.2,0.3879916763305664
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,1,1,power_law_1.2,0.39432128906249997
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,1,1,power_law_1.2,0.4117292785644532
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,1,1,power_law_1.2,0.4732921600341797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,1,1,power_law_1.2,0.5864128112792969
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,1,1,power_law_1.2,0.7424940490722657
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,1,1,power_law_1.2,0.9463820648193358
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,1,1,power_law_1.2,23.807124023437503
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,1,1,power_law_1.2,1.461015625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,1,1,power_law_1.2,1.9226272583007813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,1,1,power_law_1.2,2.941212158203125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,1,1,power_law_1.2,3.7642471313476564
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,1,1,power_law_1.2,33.8926318359375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,1,1,power_law_1.2,43.3646826171875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,1,1,power_law_1.2,0.05482495784759521
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,1,1,power_law_1.2,0.07778687953948975
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,1,1,power_law_1.2,0.0882643222808838
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,1,1,power_law_1.2,0.14526528358459473
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,1,1,power_law_1.2,0.142740478515625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,1,1,power_law_1.2,0.14815872192382812
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,1,1,power_law_1.2,0.14879615783691408
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,1,1,power_law_1.2,0.1473465633392334
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,1,1,power_law_1.2,0.15057151794433593
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,1,1,power_law_1.2,0.1509836769104004
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,1,1,power_law_1.2,0.15376704216003417
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,1,1,power_law_1.2,0.1564121627807617
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,1,1,power_law_1.2,0.1631372833251953
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,1,1,power_law_1.2,0.18168895721435546
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,1,1,power_law_1.2,0.19602176666259766
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,1,1,power_law_1.2,0.21108480453491213
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,1,1,power_law_1.2,5.679849243164062
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,1,1,power_law_1.2,7.890074462890626
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,1,1,power_law_1.2,0.23317184448242187
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,1,1,power_law_1.2,0.2982828712463379
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,1,1,power_law_1.2,0.47312320709228517
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,1,1,power_law_1.2,10.087433471679688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,1,1,power_law_1.2,0.3557503890991211
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,1,1,power_law_1.2,0.5973823928833009
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,1,1,power_law_1.2,0.8464620971679688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,1,1,power_law_1.2,1.1229138946533204
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,1,1,power_law_1.2,1.6493606567382812
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,1,1,power_law_1.2,2.2001068115234377
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,1,1,power_law_1.01,0.31760000228881835
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,1,1,power_law_1.01,0.458419189453125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,1,1,power_law_1.01,0.5341836929321289
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,1,1,power_law_1.01,0.809955825805664
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,1,1,power_law_1.01,0.9295200347900391
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,1,1,power_law_1.01,0.9352678680419922
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,1,1,power_law_1.01,0.9412198638916015
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,1,1,power_law_1.01,0.9382848358154297
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,1,1,power_law_1.01,0.9503359985351562
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,1,1,power_law_1.01,0.986084442138672
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,1,1,power_law_1.01,1.0681107330322266
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,1,1,power_law_1.01,1.1869222259521484
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,1,1,power_law_1.01,1.2952774047851563
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,1,1,power_law_1.01,1.4026976013183594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,1,1,power_law_1.2,3.5466860961914066
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,1,1,power_law_1.01,1.5140870666503905
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,1,1,power_law_1.01,1.6217388916015625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,1,1,power_law_1.01,1.9905133056640625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,1,1,power_law_1.01,2.804599609375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,1,1,power_law_1.01,3.366776428222656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,1,1,power_law_1.01,5.003383178710938
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,1,1,power_law_1.01,6.43763916015625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,1,1,power_law_1.01,10.40165771484375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,1,1,power_law_1.2,4.375010681152344
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,1,1,power_law_1.01,13.335787353515624
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,1,1,power_law_1.01,19.995412597656248
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,1,1,power_law_1.01,26.16923095703125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,1,1,power_law_1.2,5.466286010742188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,1,1,power_law_1.01,0.16148544311523438
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,1,1,power_law_1.01,0.24155712127685547
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,1,1,power_law_1.01,0.2851417541503906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,1,1,power_law_1.01,0.43350463867187494
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,1,1,power_law_1.01,0.47012607574462895
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,1,1,power_law_1.01,0.47318847656249996
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,1,1,power_law_1.01,0.47594047546386714
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,1,1,power_law_1.01,0.47015232086181646
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,1,1,power_law_1.01,0.4724140930175781
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,1,1,power_law_1.01,0.48301441192626954
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,1,1,power_law_1.01,0.4852518463134766
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,1,1,power_law_1.01,0.48045822143554684
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,1,1,power_law_1.01,0.4953664016723633
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,1,1,power_law_1.01,0.517331199645996
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,1,1,power_law_1.01,0.5245856094360352
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,1,1,power_law_1.01,0.548389778137207
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,1,1,power_law_1.01,0.5991372680664062
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,1,1,power_law_1.01,0.7598521423339843
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,1,1,power_law_1.01,0.9389977264404298
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,1,1,power_law_1.01,1.3508837890624998
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,1,1,power_law_1.01,41.488876953125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,1,1,power_law_1.01,1.7779335021972655
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,1,1,power_law_1.01,2.6979467773437498
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,1,1,power_law_1.01,3.3540826416015626
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,1,1,power_law_1.01,5.005539245605469
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,1,1,power_law_1.01,6.353030395507813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,1,1,power_law_1.01,59.41730468749999
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,1,1,power_law_1.01,10.254256591796874
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,1,1,power_law_1.01,13.345670166015626
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,1,1,power_law_1.01,77.55796875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,1,1,power_law_1.01,17.5898291015625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,1,1,power_law_1.01,0.07998208045959473
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,1,1,power_law_1.01,0.24213951110839843
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,1,1,power_law_1.01,0.24260543823242187
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,1,1,power_law_1.01,0.24318399429321289
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,1,1,power_law_1.01,0.23986240386962893
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,1,1,power_law_1.01,0.24020351409912108
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,1,1,power_law_1.01,0.24629823684692384
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,1,1,power_law_1.01,0.24907903671264647
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,1,1,power_law_1.01,0.13133248329162597
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,1,1,power_law_1.01,0.15806400299072265
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,1,1,power_law_1.01,0.24023168563842773
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,1,1,power_law_1.01,0.24776256561279295
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,1,1,power_law_1.01,0.2606489562988281
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,1,1,power_law_1.01,0.29481216430664064
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,1,1,power_law_1.01,0.30543872833251956
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,1,1,power_law_1.01,0.321673583984375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,1,1,power_law_1.01,0.3694508743286133
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,1,1,power_law_1.01,0.4830643081665039
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,1,1,power_law_1.01,0.5843718338012696
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,1,1,power_law_1.01,0.8067462158203125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,1,1,power_law_1.01,1.0174752044677735
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,1,1,power_law_1.01,1.5634144592285157
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,1,1,power_law_1.01,2.1153305053710936
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,1,1,power_law_1.01,2.984738464355469
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,1,1,power_law_1.01,3.9934738159179686
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,1,1,power_law_1.2,0.3123097610473633
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,1,1,power_law_1.2,0.4337881469726563
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,1,1,power_law_1.2,0.4838745498657227
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,1,1,power_law_1.2,0.8014080047607421
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,1,1,power_law_1.2,0.9038432312011718
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,1,1,power_law_1.2,0.9315462493896485
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,1,1,power_law_1.2,0.9405165100097657
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,1,1,power_law_1.2,0.9361280059814453
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,1,1,power_law_1.2,0.9605197143554687
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,1,1,power_law_1.2,0.9798770904541015
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,1,1,power_law_1.2,1.0902079772949218
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,1,1,power_law_1.2,1.2004518127441406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,1,1,power_law_1.2,1.2748889923095703
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,1,1,power_law_1.2,1.3840716552734373
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,1,1,power_law_1.2,1.4802873229980469
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,1,1,power_law_1.2,1.6468038940429686
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,1,1,power_law_1.2,1.9545164489746092
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,1,1,power_law_1.2,2.728077392578125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,1,1,power_law_1.01,5.708119506835938
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,1,1,power_law_1.2,3.3988671875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,1,1,power_law_1.2,4.878953552246093
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,1,1,power_law_1.2,6.354430541992188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,1,1,power_law_1.2,9.90172119140625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,1,1,power_law_1.01,8.325242919921875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,1,1,power_law_1.2,13.245247802734374
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,1,1,power_law_1.01,10.244779052734376
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,1,1,power_law_1.2,19.989608154296874
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,1,1,power_law_1.2,27.836440429687503
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,1,1,power_law_1.2,0.16426368713378908
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,1,1,power_law_1.2,0.23285951614379882
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,1,1,power_law_1.2,0.2610515213012695
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,1,1,power_law_1.2,0.43785663604736325
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,1,1,power_law_1.2,0.456445426940918
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,1,1,power_law_1.2,0.47366977691650386
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,1,1,power_law_1.2,0.47645824432373046
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,1,1,power_law_1.2,0.4710732650756836
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,1,1,power_law_1.2,0.47995967864990235
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,1,1,power_law_1.2,0.4839193725585938
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,1,1,power_law_1.2,0.4869862365722656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,1,1,power_law_1.2,0.4893273544311524
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,1,1,power_law_1.2,0.4921343994140625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,1,1,power_law_1.2,0.5170816040039062
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,1,1,power_law_1.2,0.5325120162963868
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,1,1,power_law_1.2,0.5596217727661132
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,1,1,power_law_1.2,0.6274969482421875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,1,1,power_law_1.2,0.7870553588867187
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,1,1,power_law_1.2,0.9735449218750001
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,1,1,power_law_1.2,1.2857594299316406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,1,1,power_law_1.2,1.9729145812988282
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,1,1,power_law_1.2,2.7015866088867186
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,1,1,power_law_1.2,42.0478564453125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,1,1,power_law_1.2,3.539440002441406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,1,1,power_law_1.2,4.97174072265625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,1,1,power_law_1.2,6.842897338867187
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,1,1,power_law_1.2,59.9519775390625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,1,1,power_law_1.2,10.262762451171875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,1,1,power_law_1.2,0.08140607833862304
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,1,1,power_law_1.2,0.12499199867248537
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,1,1,power_law_1.2,0.1423846435546875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,1,1,power_law_1.2,0.23578687667846682
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,1,1,power_law_1.2,0.23131071090698244
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,1,1,power_law_1.2,0.24098047256469726
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,1,1,power_law_1.2,0.24165376663208007
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,1,1,power_law_1.2,0.23834815979003907
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,1,1,power_law_1.2,0.24338304519653323
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,1,1,power_law_1.2,0.24465856552124024
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,1,1,power_law_1.2,0.24755008697509764
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,1,1,power_law_1.2,0.2513094329833984
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,1,1,power_law_1.2,0.2653734397888184
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,1,1,power_law_1.2,0.29733631134033206
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,1,1,power_law_1.2,79.2261669921875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,1,1,power_law_1.2,14.162664794921875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,1,1,power_law_1.2,0.3149452781677246
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,1,1,power_law_1.2,0.3378611373901367
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,1,1,power_law_1.2,16.836080322265623
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,1,1,power_law_1.2,0.7653798675537109
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,1,1,power_law_1.2,0.38456832885742187
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,1,1,power_law_1.2,0.48568576812744146
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,1,1,power_law_1.2,0.5995238494873047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,1,1,power_law_1.2,1.014672622680664
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,1,1,power_law_1.2,1.5115219116210938
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,1,1,power_law_1.2,2.0743615722656252
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,1,1,power_law_1.2,2.8944671630859373
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,1,1,power_law_1.2,3.8885247802734377
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,1,power_law_1.01,0.14400639533996581
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,1,power_law_1.01,0.2847756767272949
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,1,power_law_1.01,0.49764160156250004
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,1,power_law_1.01,0.9172096252441406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,1,power_law_1.01,1.3788262939453124
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,1,power_law_1.01,1.7778585815429686
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,1,1,power_law_1.2,6.1874890136718745
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,1,power_law_1.01,2.5226348876953124
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,1,power_law_1.01,2.7824057006835936
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,1,power_law_1.01,2.722534484863281
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,1,power_law_1.01,2.9589944458007813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,1,power_law_1.01,2.9950714111328125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,1,power_law_1.01,3.494376220703125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,1,power_law_1.01,3.3282983398437502
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,1,power_law_1.01,3.8596697998046876
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,1,power_law_1.01,3.466738586425781
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,1,power_law_1.01,3.801514282226562
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,1,power_law_1.01,3.956718139648437
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,1,power_law_1.01,5.072107543945313
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,1,power_law_1.01,4.8999853515625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,1,power_law_1.01,6.28456298828125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,1,power_law_1.01,6.9083056640625005
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,1,power_law_1.01,9.443757934570312
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,1,power_law_1.01,10.673446044921874
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,1,1,power_law_1.2,7.957256469726562
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,1,power_law_1.01,15.019752197265626
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,1,power_law_1.01,17.495643310546875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,1,power_law_1.01,23.569677734375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,1,1,power_law_1.2,10.15593505859375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,1,power_law_1.01,0.09236224174499512
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,1,power_law_1.01,33.27900390625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,1,power_law_1.01,0.15132479667663573
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,1,power_law_1.01,0.2658969688415527
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,1,power_law_1.01,0.9385209655761718
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,1,power_law_1.01,0.7090745544433593
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,1,power_law_1.01,1.2810791015625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,1,power_law_1.01,0.4771097564697266
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,1,1,power_law_1.01,38.198039550781246
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,1,power_law_1.01,1.39051513671875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,1,power_law_1.01,1.3505235290527344
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,1,power_law_1.01,1.5151303100585938
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,1,power_law_1.01,1.5515653991699219
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,1,power_law_1.01,1.43387451171875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,1,power_law_1.01,1.4745280456542968
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,1,power_law_1.01,1.5631564331054686
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,1,power_law_1.01,1.7350233459472655
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,1,power_law_1.01,1.8730335998535157
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,1,power_law_1.01,1.9305862426757812
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,1,power_law_1.01,2.0224896240234376
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,1,power_law_1.01,2.163962860107422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,1,power_law_1.01,2.4701971435546874
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,1,power_law_1.01,2.788215026855469
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,1,power_law_1.01,3.1568670654296875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,1,power_law_1.01,3.704879455566406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,1,power_law_1.01,5.227334594726562
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,1,power_law_1.01,5.94138916015625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,1,power_law_1.01,7.960726318359375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,1,power_law_1.01,10.1316357421875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,1,1,power_law_1.01,13.22491455078125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,1,1,power_law_1.01,0.0897439956665039
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,1,1,power_law_1.01,0.1232051181793213
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,1,1,power_law_1.01,0.4272505569458008
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,1,1,power_law_1.01,0.5430227279663086
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,1,1,power_law_1.01,0.7668000030517578
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,1,1,power_law_1.01,0.17568960189819335
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,1,1,power_law_1.01,0.8177670288085939
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,1,1,power_law_1.01,0.8341120147705079
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,1,1,power_law_1.01,0.8465535736083984
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,1,1,power_law_1.01,0.8888883209228515
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,1,1,power_law_1.01,0.9224089813232421
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,1,1,power_law_1.01,0.9500736236572266
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,1,1,power_law_1.01,0.9284172821044923
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,1,1,power_law_1.01,0.29192768096923827
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,1,1,power_law_1.01,1.004126739501953
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,1,1,power_law_1.01,0.9949753570556641
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,1,1,power_law_1.01,1.0749324798583983
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,1,1,power_law_1.01,1.2867756652832032
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,1,1,power_law_1.01,1.41076416015625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,1,1,power_law_1.01,1.6519007873535156
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,1,1,power_law_1.01,1.1423385620117188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,1,1,power_law_1.01,2.0111680603027344
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,1,1,power_law_1.01,2.3648716735839845
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,1,1,power_law_1.01,3.17355712890625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,1,1,power_law_1.01,4.041922607421875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,1,power_law_1.2,0.14181119918823243
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,1,power_law_1.2,0.28113088607788084
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,1,power_law_1.2,0.49484607696533206
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,1,power_law_1.2,0.8440294647216797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,1,power_law_1.2,1.2491295623779297
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,1,power_law_1.2,1.6852371215820312
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,1,power_law_1.2,2.5396389770507812
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,1,power_law_1.2,2.6916314697265626
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,1,1,power_law_1.01,5.815686645507812
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,1,power_law_1.2,2.826195983886719
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,1,1,power_law_1.01,7.503127441406249
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,1,power_law_1.2,3.1784478759765626
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,1,power_law_1.2,2.750335388183594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,7168,2048,8,256,1,1,power_law_1.01,9.335732421874999
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,1,power_law_1.2,3.7397811889648436
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,1,power_law_1.2,3.0831768798828127
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,1,power_law_1.2,3.7166302490234373
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,1,power_law_1.2,3.601502075195312
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,1,power_law_1.2,4.109814453125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,1,power_law_1.2,4.019483032226562
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,1,power_law_1.2,4.982107543945313
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,1,power_law_1.2,4.52637451171875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,1,power_law_1.2,6.554184570312501
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,1,power_law_1.2,6.913357543945312
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,1,power_law_1.2,9.04733642578125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,1,power_law_1.2,10.75312255859375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,1,power_law_1.2,15.054041748046876
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,1,power_law_1.2,17.047705078125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,1,power_law_1.2,0.09177599906921387
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,1,power_law_1.2,0.15015168190002443
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,1,power_law_1.2,0.2656217575073242
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,1,power_law_1.2,0.4413772964477539
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,1,power_law_1.2,0.6487430572509766
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,1,power_law_1.2,0.8857740783691407
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,1,power_law_1.2,1.2625727844238281
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,1,power_law_1.2,1.32173828125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,1,power_law_1.2,24.4133447265625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,1,power_law_1.2,1.3696109008789061
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,1,power_law_1.2,1.418406982421875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,1,power_law_1.2,1.4545574951171876
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,1,power_law_1.2,1.527904052734375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,1,power_law_1.2,1.531091766357422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,1,power_law_1.2,1.6262387084960939
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,1,power_law_1.2,30.61661865234375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,1,power_law_1.2,1.8089337158203125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,1,power_law_1.2,1.9376101684570313
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,1,power_law_1.2,1.9390380859375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,1,power_law_1.2,2.1706636047363284
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,1,power_law_1.2,2.3756422424316406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,1,power_law_1.2,2.5923193359375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,1,power_law_1.2,2.8477163696289063
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,1,power_law_1.2,3.4757791137695313
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,1,power_law_1.2,3.9638177490234376
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,1,power_law_1.2,4.838796691894531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,1,1,power_law_1.2,40.3263720703125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,1,power_law_1.2,6.125543212890625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,1,1,power_law_1.2,0.09013888359069824
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,1,power_law_1.2,8.257137451171875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,1,1,power_law_1.2,0.12340479850769044
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,1,1,power_law_1.2,0.17334848403930664
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,1,1,power_law_1.2,0.2716147232055664
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,1,power_law_1.2,10.107801513671875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,1,1,power_law_1.2,0.3885433578491211
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,1,1,power_law_1.2,0.5110739135742188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,1,1,power_law_1.2,0.757132797241211
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,1,1,power_law_1.2,0.7920403289794922
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,1,1,power_law_1.2,0.8237983703613281
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,1,1,power_law_1.2,0.8538566589355469
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,1,1,power_law_1.2,0.8765644836425782
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,1,1,power_law_1.2,0.9197484588623046
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,1,1,power_law_1.2,13.37791015625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,1,1,power_law_1.2,0.9427244567871094
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,1,1,power_law_1.2,0.9873907470703125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,1,1,power_law_1.2,1.0285504150390625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,1,1,power_law_1.2,1.050561294555664
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,1,1,power_law_1.2,1.1179341125488282
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,1,1,power_law_1.2,1.2713075256347657
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,1,1,power_law_1.2,1.7032984924316406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,1,1,power_law_1.2,1.3838847351074217
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,1,1,power_law_1.2,1.8615795898437502
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,1,1,power_law_1.2,2.4881100463867187
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,1,1,power_law_1.2,3.1440069580078123
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,1,1,power_law_1.2,4.420372619628906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,1,power_law_1.01,0.03288640022277832
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,1,power_law_1.01,0.07787263870239258
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,1,power_law_1.01,0.1113260841369629
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,1,power_law_1.01,0.1335110378265381
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,1,power_law_1.01,0.18801984786987305
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,1,power_law_1.01,0.19671167373657225
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,1,power_law_1.01,0.19911104202270508
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,1,power_law_1.01,0.211661434173584
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,1,power_law_1.01,0.217061767578125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,1,power_law_1.01,0.2340774345397949
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,1,power_law_1.01,0.23988672256469729
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,1,power_law_1.01,0.242553596496582
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,1,power_law_1.01,0.24964927673339843
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,1,power_law_1.01,0.264256649017334
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,1,power_law_1.01,0.27231359481811523
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,1,power_law_1.01,0.2866579246520996
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,1,power_law_1.01,0.3192044830322266
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,1,power_law_1.01,0.35485057830810546
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,1,power_law_1.01,0.4342073440551758
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,1,power_law_1.01,0.5179884719848633
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,1,power_law_1.01,0.6856781005859375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,1,power_law_1.01,0.8507974243164063
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,1,power_law_1.01,1.1960467529296874
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,1,power_law_1.01,1.5358956909179686
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,1,1,power_law_1.2,5.833592529296875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,1,power_law_1.01,2.2637120056152344
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,1,power_law_1.01,2.928686218261719
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,2048,768,8,128,1,1,power_law_1.01,3.639573669433594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,1,power_law_1.01,0.033714559078216556
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,1,power_law_1.01,0.040873599052429196
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,1,power_law_1.01,0.0792902421951294
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,1,power_law_1.01,0.09798208236694336
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,1,power_law_1.01,0.13424384117126464
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,1,power_law_1.01,0.13967616081237794
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,1,power_law_1.01,0.14111552238464356
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,1,power_law_1.01,0.14516223907470702
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,1,power_law_1.01,0.06383296012878417
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,1,power_law_1.01,0.14499327659606934
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,1,power_law_1.01,0.14880640029907227
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,1,power_law_1.01,0.1691059112548828
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,1,power_law_1.01,0.17060352325439454
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,1,power_law_1.01,0.17332544326782226
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,1,power_law_1.01,0.18112960815429688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,1,power_law_1.01,0.18389568328857422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,1,power_law_1.01,0.19501632690429688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,1,power_law_1.01,0.20265663146972654
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,1,power_law_1.01,0.2197849655151367
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,1,power_law_1.01,0.2547871971130371
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,1,power_law_1.01,0.2963520050048828
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,1,power_law_1.01,0.37220672607421873
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,1,power_law_1.01,0.4456972885131836
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,1,power_law_1.01,0.6037657546997071
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,1,power_law_1.01,0.7557241821289062
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,1,power_law_1.01,1.0739046478271486
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,1,power_law_1.01,1.3688281249999998
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,2048,768,8,128,1,1,power_law_1.01,1.6850988769531248
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,1,1,power_law_1.01,0.03680576086044311
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,1,1,power_law_1.01,0.04815616130828858
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,1,1,power_law_1.01,0.054236798286437994
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,1,1,power_law_1.01,0.04177792072296142
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,1,1,power_law_1.01,0.06113152027130127
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,1,1,power_law_1.01,0.0786079978942871
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,1,1,power_law_1.01,0.08207615852355957
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,1,1,power_law_1.01,0.084584321975708
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,1,1,power_law_1.01,0.08609536170959473
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,1,1,power_law_1.01,0.08750080108642579
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,1,1,power_law_1.01,0.09205120086669923
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,1,1,power_law_1.01,0.09522111892700195
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,1,1,power_law_1.01,0.09939007759094239
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,1,1,power_law_1.01,0.10545087814331056
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,1,1,power_law_1.01,0.08988927841186524
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,1,1,power_law_1.01,0.11801983833312987
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,1,1,power_law_1.01,0.10803008079528809
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,1,1,power_law_1.01,0.15707648277282715
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,1,1,power_law_1.01,0.1389094352722168
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,1,1,power_law_1.01,0.20324287414550782
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,1,1,power_law_1.01,0.2532876777648926
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,1,1,power_law_1.01,0.35304832458496094
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,1,1,power_law_1.01,0.4570272064208984
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,1,1,power_law_1.01,0.67146240234375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,1,1,power_law_1.01,0.8891468811035157
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,1,power_law_1.2,0.03783360004425049
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,1,1,power_law_1.01,1.3332762145996093
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,1,power_law_1.2,0.05255680084228516
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,1,power_law_1.2,0.10512127876281738
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,1,1,power_law_1.01,1.790987548828125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,1,power_law_1.2,0.12895551681518555
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,1,power_law_1.2,0.19167295455932618
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,1,power_law_1.2,0.19688127517700196
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,1,power_law_1.2,0.20837312698364258
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,1,power_law_1.2,0.21788351058959962
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,1,power_law_1.2,0.21950592041015624
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,1,power_law_1.2,0.23702720642089842
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,1,power_law_1.2,0.23965120315551758
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,1,power_law_1.2,0.25101696014404296
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,1,power_law_1.2,0.26788864135742185
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,1,power_law_1.2,0.244323844909668
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,1,power_law_1.2,0.27573951721191403
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,2048,768,8,128,1,1,power_law_1.01,2.2543673706054688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,1,power_law_1.2,0.2892780876159668
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,1,power_law_1.2,0.32488574981689455
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,1,power_law_1.2,0.3610707092285156
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,1,power_law_1.2,0.44221057891845705
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,1,power_law_1.2,0.5174086380004883
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,1,power_law_1.2,0.6915634918212891
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,1,power_law_1.2,0.07730432033538819
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,1,power_law_1.2,0.8616627502441407
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,1,power_law_1.2,1.1886962890625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,1,power_law_1.2,1.5280953979492187
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,1,power_law_1.2,0.03424511909484863
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,1,power_law_1.2,0.041535358428955074
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,1,power_law_1.2,0.06310143947601318
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,1,power_law_1.2,0.07512832164764405
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,1,power_law_1.2,0.09500800132751465
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,1,power_law_1.2,0.1335097599029541
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,1,power_law_1.2,0.1401683235168457
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,1,power_law_1.2,0.14412287712097166
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,1,power_law_1.2,0.1467865562438965
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,1,power_law_1.2,0.1470131206512451
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,1,power_law_1.2,0.14835519790649415
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,1,power_law_1.2,0.16986495971679688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,1,power_law_1.2,0.1709087944030762
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,1,power_law_1.2,0.17918272018432618
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,1,power_law_1.2,0.18386560440063476
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,1,power_law_1.2,0.1891756820678711
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,1,power_law_1.2,0.19689023971557618
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,1,power_law_1.2,0.20583040237426758
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,1,power_law_1.2,0.22604671478271485
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,1,power_law_1.2,2.3236927795410156
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,1,power_law_1.2,0.26532991409301754
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,1,power_law_1.2,0.37616321563720706
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,1,power_law_1.2,0.2991488075256348
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,1,power_law_1.2,0.45539905548095705
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,1,power_law_1.2,0.6044947052001953
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,1,power_law_1.2,0.7641932678222656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,1,power_law_1.2,2.945672912597656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,1,power_law_1.2,1.0723725128173829
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,1,1,power_law_1.2,0.036595199108123776
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,1,1,power_law_1.2,0.04148416042327881
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,1,1,power_law_1.2,0.04768064022064209
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,1,1,power_law_1.2,0.05230080127716065
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,1,1,power_law_1.2,0.058801279067993165
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,1,1,power_law_1.2,0.07917183876037598
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,1,1,power_law_1.2,0.08275584220886231
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,1,1,power_law_1.2,0.08499967575073242
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,1,1,power_law_1.2,0.08763392448425293
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,2048,768,8,128,1,1,power_law_1.2,3.58694580078125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,1,1,power_law_1.2,0.08830080032348633
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,1,1,power_law_1.2,0.0914367961883545
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,1,1,power_law_1.2,0.09523712158203125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,1,1,power_law_1.2,0.09771007537841797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,1,1,power_law_1.2,0.11073984146118163
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,1,1,power_law_1.2,0.11656384468078614
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,1,1,power_law_1.2,0.1045849609375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,1,1,power_law_1.2,0.15995967864990235
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,1,1,power_law_1.2,0.13000896453857422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,1,power_law_1.2,1.3775814819335939
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,1,1,power_law_1.2,0.19293760299682616
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,1,1,power_law_1.2,0.2663007926940918
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,1,1,power_law_1.2,0.34292480468750003
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,1,1,power_law_1.2,0.4974790573120117
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,1,1,power_law_1.2,0.6627654266357422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,2048,768,8,128,1,1,power_law_1.2,1.6875640869140625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,1,1,power_law_1.2,0.976890869140625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,1,1,power_law_1.2,1.2831039428710938
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,1,power_law_1.01,0.081277437210083
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,1,power_law_1.01,0.13121727943420408
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,1,power_law_1.01,0.2345683288574219
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,1,power_law_1.01,0.3352422332763672
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,1,power_law_1.01,0.42301376342773434
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,1,power_law_1.01,0.6445145416259765
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,1,power_law_1.01,0.6802003479003906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,1,power_law_1.01,0.6916172790527344
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,1,power_law_1.01,0.7278598022460938
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,1,power_law_1.01,0.7383699035644531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,1,power_law_1.01,0.8107437133789063
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,1,power_law_1.01,0.8456800079345703
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,1,1,power_law_1.2,1.9077299499511717
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,1,power_law_1.01,0.8593670654296875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,1,power_law_1.01,0.8779519653320312
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,1,power_law_1.01,0.9097325134277344
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,1,power_law_1.01,0.9251168060302735
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,1,power_law_1.01,0.9849734497070312
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,1,power_law_1.01,1.1005235290527344
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,1,power_law_1.01,1.2765740966796875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,1,power_law_1.01,1.6737510681152343
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,1,power_law_1.01,2.116330261230469
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,1,power_law_1.01,2.70172607421875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,1,power_law_1.01,3.453537902832031
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,1,power_law_1.01,4.717276306152344
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,1,power_law_1.01,6.2549072265625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,1,power_law_1.01,9.039263916015624
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,1,power_law_1.01,12.5107275390625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,1,power_law_1.01,0.057912321090698236
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,1,1,power_law_1.01,14.52599609375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,1,power_law_1.01,0.14086912155151368
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,1,power_law_1.01,0.19319168090820313
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,1,power_law_1.01,0.24765567779541015
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,1,power_law_1.01,0.08999423980712891
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,1,power_law_1.01,0.3571097564697266
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,1,power_law_1.01,0.39543041229248044
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,1,1,power_law_1.2,2.531014404296875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,1,power_law_1.01,0.4065151977539062
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,1,power_law_1.01,0.4047468948364258
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,1,power_law_1.01,0.4233958435058594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,1,power_law_1.01,0.3847321701049805
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,1,power_law_1.01,0.45700736999511715
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,1,power_law_1.01,0.45870849609375003
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,1,power_law_1.01,0.4698271942138672
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,1,power_law_1.01,0.4838835144042969
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,1,power_law_1.01,0.486321907043457
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,1,power_law_1.01,0.5097811126708984
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,1,power_law_1.01,0.5364326477050781
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,1,power_law_1.01,0.575272331237793
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,1,power_law_1.01,0.6665446472167968
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,1,power_law_1.01,0.7941677093505859
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,1,power_law_1.01,0.9619583892822267
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,1,power_law_1.01,1.1930847930908204
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,1,power_law_1.01,1.6240576171874999
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,1,power_law_1.01,2.1153439331054686
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,1,power_law_1.01,3.1845843505859373
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,1,power_law_1.01,3.9425344848632813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,1,1,power_law_1.01,4.823627014160157
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,1,1,power_law_1.01,0.04924287796020508
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,1,1,power_law_1.01,0.11646528244018554
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,1,1,power_law_1.01,0.0640825605392456
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,1,1,power_law_1.01,0.1427116775512695
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,1,1,power_law_1.01,0.21057600021362305
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,1,1,power_law_1.01,0.08783552169799805
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,1,1,power_law_1.01,0.22433536529541015
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,1,1,power_law_1.01,0.23167488098144534
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,1,1,power_law_1.01,0.23863807678222657
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,1,1,power_law_1.01,0.24001279830932618
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,1,1,power_law_1.01,0.25090560913085935
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,1,1,power_law_1.01,0.2552537536621094
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,1,1,power_law_1.01,0.27261119842529297
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,1,1,power_law_1.01,0.26032831192016603
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,1,1,power_law_1.01,0.2848671913146973
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,1,1,power_law_1.01,0.29361215591430667
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,1,1,power_law_1.01,0.3155353546142578
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,2048,768,8,128,1,1,power_law_1.2,3.163321533203125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,1,1,power_law_1.01,0.36493438720703125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,1,1,power_law_1.01,0.41950977325439454
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,1,1,power_law_1.01,0.5303116989135742
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,1,1,power_law_1.01,0.6581267547607422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,1,1,power_law_1.01,0.886948471069336
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,1,1,power_law_1.01,1.1343014526367188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,1,1,power_law_1.01,1.624919738769531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,1,1,power_law_1.01,2.1353932189941407
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,1,1,power_law_1.01,3.188054504394531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,1,power_law_1.2,0.08125056266784668
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,1,power_law_1.2,0.13248448371887206
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,1,1,power_law_1.01,4.253887939453125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,1536,8,128,1,1,power_law_1.01,5.362393188476562
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,1,power_law_1.2,0.3115033531188965
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,1,power_law_1.2,0.39303489685058596
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,1,power_law_1.2,0.6561830139160156
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,1,power_law_1.2,0.6884652709960937
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,1,power_law_1.2,0.7229510498046875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,1,power_law_1.2,0.7487814331054687
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,1,power_law_1.2,0.7498233795166016
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,1,power_law_1.2,0.8114393615722657
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,1,power_law_1.2,0.8611270141601561
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,1,power_law_1.2,0.8646451568603515
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,1,power_law_1.2,0.8900806427001953
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,1,power_law_1.2,0.23235967636108396
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,1,power_law_1.2,0.9317536163330079
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,1,power_law_1.2,0.9542668914794922
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,1,power_law_1.2,0.9990585327148438
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,1,power_law_1.2,1.140901107788086
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,1,power_law_1.2,1.372380828857422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,1,power_law_1.2,1.6512197875976562
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,1,power_law_1.2,2.0371775817871094
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,1,power_law_1.2,2.6290194702148435
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,1,power_law_1.2,3.392611083984375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,1,power_law_1.2,4.647791442871094
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,1,power_law_1.2,5.989774780273438
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,1,power_law_1.2,0.05743167877197266
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,1,power_law_1.2,0.09221759796142578
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,1,power_law_1.2,0.13713472366333007
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,1,power_law_1.2,0.18140224456787107
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,1,power_law_1.2,0.23464448928833007
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,1,power_law_1.2,0.37180736541748044
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,1,power_law_1.2,0.387828483581543
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,1,power_law_1.2,0.40034111022949215
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,1,power_law_1.2,0.4101510238647461
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,1,power_law_1.2,0.4081868743896484
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,1,power_law_1.2,0.41684223175048823
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,1,power_law_1.2,0.4594400024414062
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,1,power_law_1.2,0.46423679351806635
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,1,power_law_1.2,0.4809324645996094
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,1,power_law_1.2,0.4935283279418945
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,1,power_law_1.2,0.49588672637939457
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,1,power_law_1.2,0.5110739135742188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,1,power_law_1.2,0.5447910308837891
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,1,power_law_1.2,0.5885548782348633
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,1,power_law_1.2,0.6810342407226562
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,1,power_law_1.2,0.7789273834228515
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,1,power_law_1.2,0.953812484741211
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,1,power_law_1.2,1.2009420776367188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,1,power_law_1.2,8.882369995117188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,1,power_law_1.2,1.676862030029297
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,1,power_law_1.2,2.045509796142578
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,1,power_law_1.2,12.369483642578125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,1,power_law_1.2,2.980970153808594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,1,1,power_law_1.2,0.04921984195709229
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,1,1,power_law_1.2,0.06369728088378906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,1,1,power_law_1.2,0.08560704231262208
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,1,1,power_law_1.2,0.10892864227294922
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,1,1,power_law_1.2,0.13422847747802735
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,1,1,power_law_1.2,0.21421632766723633
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,1,1,power_law_1.2,0.22787071228027345
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,1,1,power_law_1.2,0.23667327880859376
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,1,1,power_law_1.2,0.2442246437072754
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,1,1,power_law_1.2,0.24544895172119138
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,1,1,power_law_1.2,0.2530617523193359
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,1,1,power_law_1.2,0.26406976699829104
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,1,1,power_law_1.2,0.26926464080810547
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,1,1,power_law_1.2,0.28547264099121095
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,1,1,power_law_1.2,0.3019756889343262
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,1,1,power_law_1.2,0.31450624465942384
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,1,1,power_law_1.2,14.370098876953126
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,1,1,power_law_1.2,0.35006271362304686
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,1,1,power_law_1.2,0.4188812637329102
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,1,1,power_law_1.2,0.49535678863525395
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,1,1,power_law_1.2,0.6631712341308594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,1,power_law_1.2,3.8301464843749997
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,1,1,power_law_1.2,0.849491195678711
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,1,1,power_law_1.2,4.9299923706054685
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,1,1,power_law_1.2,1.173034896850586
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,1,1,power_law_1.2,1.538019256591797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,1,1,power_law_1.2,2.2375027465820314
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,1,1,power_law_1.2,2.931721496582031
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,1,power_law_1.01,0.1517849636077881
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,1,power_law_1.01,0.3095027160644531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,1,power_law_1.01,0.516568946838379
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,1,power_law_1.01,0.8208211517333984
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,1,power_law_1.01,1.1259053039550782
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,1,power_law_1.01,1.8414585876464844
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,1,power_law_1.01,1.9535430908203124
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,1,power_law_1.01,2.045250549316406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,1,power_law_1.01,2.1155206298828126
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,1,power_law_1.01,2.2383424377441408
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,1,power_law_1.01,2.4660263061523438
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,1,power_law_1.01,2.66087158203125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,1,1,power_law_1.2,4.3161248779296875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,1,power_law_1.01,2.4960633850097658
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,1,power_law_1.01,2.5878393554687498
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,1,power_law_1.01,2.7521875000000002
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,1,power_law_1.01,2.6825234985351565
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,1,power_law_1.01,3.2216146850585936
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,1,power_law_1.01,3.5452166748046876
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,1,power_law_1.01,4.295881652832032
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,1,power_law_1.01,4.952566528320313
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,1,power_law_1.01,6.139830322265626
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,1,power_law_1.01,7.853317871093751
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,1,power_law_1.01,10.39437744140625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,1,power_law_1.01,13.427049560546873
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,1,power_law_1.01,17.831793212890624
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,1,power_law_1.01,24.03445556640625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,1,1,power_law_1.2,5.715113525390625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,1,power_law_1.01,0.1143507194519043
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,1,power_law_1.01,32.94968505859375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,1,power_law_1.01,0.17136192321777344
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,1,power_law_1.01,0.28648511886596684
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,1,power_law_1.01,0.434837760925293
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,1,power_law_1.01,0.5945049667358399
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,1,power_law_1.01,0.9245760345458984
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,1,power_law_1.01,0.9846208190917969
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,1,power_law_1.01,1.0143449401855469
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,1,power_law_1.01,1.062392349243164
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,1,power_law_1.01,1.056895980834961
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,1,power_law_1.01,1.1064134216308594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,1,power_law_1.01,1.1177645111083985
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,1,power_law_1.01,1.2383026885986328
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,1,power_law_1.01,1.3034048461914063
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,1,power_law_1.01,1.3180831909179687
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,1,power_law_1.01,1.3193434143066407
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,1,power_law_1.01,1.3888294982910157
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,1,power_law_1.01,1.4686105346679688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,1,power_law_1.01,1.595927734375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,1,power_law_1.01,1.7702464294433593
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,1,1,power_law_1.01,38.373952636718755
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,1,power_law_1.01,2.0943763732910154
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,1,power_law_1.01,2.6939596557617187
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,1,power_law_1.01,3.345888671875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,1,power_law_1.01,4.583400268554688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,1,power_law_1.01,5.4076904296875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,1,power_law_1.01,7.78259521484375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,1536,8,128,1,1,power_law_1.2,7.123001708984376
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,1,power_law_1.01,10.7146142578125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,1,1,power_law_1.01,12.137406005859376
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,1,1,power_law_1.01,0.08105216026306153
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,1,1,power_law_1.01,0.11380096435546876
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,1,1,power_law_1.01,0.340590705871582
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,1,1,power_law_1.01,0.5488966369628907
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,1,1,power_law_1.01,0.5848697662353516
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,1,1,power_law_1.01,0.1692723274230957
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,1,1,power_law_1.01,0.6072012710571288
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,1,1,power_law_1.01,0.6254380798339844
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,1,1,power_law_1.01,0.6367583847045898
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,1,1,power_law_1.01,0.6562246704101563
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,1,1,power_law_1.01,0.6708914947509765
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,1,1,power_law_1.01,0.6776979064941406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,1,1,power_law_1.01,0.7205689239501953
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,1,1,power_law_1.01,0.7323212432861328
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,1,1,power_law_1.01,0.7439218902587891
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,1,1,power_law_1.01,0.25499135971069337
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,1,1,power_law_1.01,0.7840857696533203
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,1,1,power_law_1.01,0.8727852630615235
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,1,1,power_law_1.01,0.9600089263916015
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,1,1,power_law_1.01,1.1598572540283203
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,1,1,power_law_1.01,1.3711750793457032
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,1,1,power_law_1.01,1.7924537658691406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,1,1,power_law_1.01,2.1718118286132815
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,1,1,power_law_1.01,3.0497665405273438
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,1,1,power_law_1.01,3.901628723144531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,1,power_law_1.2,0.15369728088378906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,1,power_law_1.2,0.31289920806884763
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,1,1,power_law_1.01,5.7273193359375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,1,power_law_1.2,0.5022003173828125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,1,power_law_1.2,0.7481350708007812
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,1,power_law_1.2,1.0299378967285155
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,1,power_law_1.2,1.818476104736328
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,1,1,power_law_1.01,7.614472045898438
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,1,power_law_1.2,1.9046719360351563
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,1,power_law_1.2,2.0963591003417967
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,1,power_law_1.2,2.1602496337890624
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,1,power_law_1.2,2.2587718200683593
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,1,power_law_1.2,2.5999609374999997
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,1,power_law_1.2,2.5411814880371093
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,1,power_law_1.2,2.5743563842773436
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,1,power_law_1.2,2.8619256591796876
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,1,power_law_1.2,2.810418701171875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,1,power_law_1.2,2.8525729370117188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,1,power_law_1.2,3.4729669189453127
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,1,power_law_1.2,3.7299398803710937
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,1,power_law_1.2,4.227909851074219
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,1,power_law_1.2,5.483876953125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,1,power_law_1.2,5.931306762695312
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,1,power_law_1.2,7.923805541992188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,1,power_law_1.2,10.6376806640625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,1,power_law_1.2,13.526337890625001
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,2560,8,160,1,1,power_law_1.01,9.445137939453124
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,1,power_law_1.2,16.704727783203126
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,1,power_law_1.2,0.11534976005554198
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,1,power_law_1.2,0.170067195892334
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,1,power_law_1.2,0.27284095764160154
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,1,power_law_1.2,0.39145599365234374
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,1,power_law_1.2,0.5454367828369141
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,1,power_law_1.2,0.9128300476074219
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,1,power_law_1.2,0.9661644744873048
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,1,power_law_1.2,1.0295180511474609
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,1,power_law_1.2,1.0593023681640625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,1,power_law_1.2,1.0959097290039064
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,1,power_law_1.2,1.1222483062744142
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,1,power_law_1.2,1.1391117095947265
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,1,power_law_1.2,1.3106912231445311
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,1,power_law_1.2,1.3188435363769533
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,1,power_law_1.2,1.3627456665039062
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,1,power_law_1.2,25.716184082031248
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,1,power_law_1.2,1.3785536193847656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,1,power_law_1.2,1.5129991149902344
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,1,power_law_1.2,1.5807781982421876
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,1,power_law_1.2,1.8252359008789063
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,1,power_law_1.2,1.9071513366699218
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,1,power_law_1.2,2.433887939453125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,1,power_law_1.2,2.7721170043945316
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,1,power_law_1.2,3.5731442260742186
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,1,power_law_1.2,4.358802490234375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,1,power_law_1.2,5.92614013671875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,1,power_law_1.2,31.631425781249998
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,1,power_law_1.2,7.726112060546875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,1,1,power_law_1.2,0.08147647857666016
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,1,1,power_law_1.2,0.10835904121398925
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,1,1,power_law_1.2,0.16501312255859374
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,1,1,power_law_1.2,0.22909887313842772
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,1,1,power_law_1.2,0.31571775436401367
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,1,1,power_law_1.2,0.5398912048339843
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,1,1,power_law_1.2,0.5765388870239259
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,1,1,power_law_1.2,0.6136172866821289
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,1,1,power_law_1.2,0.6391987228393555
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,1,1,power_law_1.2,0.6535276794433594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,1,1,power_law_1.2,0.6854399871826172
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,1,power_law_1.2,10.52623046875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,1,1,power_law_1.2,0.6926342773437499
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,1,1,power_law_1.2,0.7238931274414062
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,1,1,power_law_1.2,0.7373004913330078
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,1,1,power_law_1.2,0.7949260711669922
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,1,1,power_law_1.2,38.316123046875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,1,1,power_law_1.2,12.20707763671875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,1,1,power_law_1.2,0.7705849456787109
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,1,1,power_law_1.2,0.8691327667236328
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,1,1,power_law_1.2,0.9700550079345703
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,1,1,power_law_1.2,1.0941587066650391
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,1,1,power_law_1.2,1.3601522827148438
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,1,1,power_law_1.2,1.6553567504882813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,1,1,power_law_1.2,2.966539001464844
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,1,1,power_law_1.2,2.2885247802734376
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,1,1,power_law_1.2,4.329902038574219
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,1,1,power_law_1.2,5.737130737304687
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,1,power_law_1.01,0.1492953586578369
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,1,power_law_1.01,0.28182464599609375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,1,power_law_1.01,0.4932799911499023
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,1,power_law_1.01,0.9108806610107422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,1,power_law_1.01,1.6376850891113282
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,1,power_law_1.01,2.292469787597656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,1,power_law_1.01,2.637792663574219
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,1,power_law_1.01,3.4187393188476562
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,1,power_law_1.01,3.608302001953125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,1,1,power_law_1.2,8.653938598632813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,1,power_law_1.01,3.801443786621094
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,1,power_law_1.01,3.965968017578125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,1,power_law_1.01,4.000879211425781
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,1,power_law_1.01,4.322017822265625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,1,power_law_1.01,4.670138244628906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,1,power_law_1.01,5.028761596679688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,1,power_law_1.01,5.044659729003906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,1,power_law_1.01,5.6508374023437495
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,1,power_law_1.01,6.262918701171875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,1,power_law_1.01,7.614517822265626
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,1,power_law_1.01,8.312518920898437
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,1,1,power_law_1.2,11.480096435546875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,1,power_law_1.01,10.637979736328125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,1,power_law_1.01,12.1495068359375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,1,power_law_1.01,15.4692138671875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,1,power_law_1.01,19.367828369140625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,1,power_law_1.01,25.7237646484375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,1,power_law_1.01,0.09778495788574218
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,1,power_law_1.01,0.15294272422790528
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,1,power_law_1.01,0.267509765625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,1,power_law_1.01,0.4760723114013672
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,1,power_law_1.01,0.8333958435058595
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,1,power_law_1.01,1.1832870483398437
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,1,power_law_1.01,32.426298828125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,1,power_law_1.01,1.3986604309082031
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,1,power_law_1.01,1.7944064331054688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,1,power_law_1.01,1.8655903625488282
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,1,power_law_1.01,1.9855218505859376
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,1,power_law_1.01,2.0065100097656248
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,1,power_law_1.01,2.093669128417969
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,1,power_law_1.01,2.088534393310547
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,1,power_law_1.01,2.2019282531738282
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,1,power_law_1.01,2.2352703857421874
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,1,1,power_law_1.01,40.05912841796875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,1,power_law_1.01,2.7478533935546876
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,1,power_law_1.01,2.72932861328125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,1,power_law_1.01,3.06178955078125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,1,power_law_1.01,3.295587158203125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,1,power_law_1.01,3.527037353515625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,1,power_law_1.01,2.3321446228027343
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,1,power_law_1.01,4.231892395019531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,1,power_law_1.01,4.683019409179687
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,1,power_law_1.01,5.78439697265625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,1,power_law_1.01,7.213467407226562
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,1,power_law_1.01,9.256978149414063
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,1,power_law_1.01,11.101563720703124
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,1,1,power_law_1.01,13.298846435546874
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,1,1,power_law_1.01,0.10528127670288086
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,1,1,power_law_1.01,0.10301247596740723
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,1,1,power_law_1.01,0.11782719612121581
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,1,1,power_law_1.01,0.13157440185546876
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,1,1,power_law_1.01,0.1473484802246094
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,1,1,power_law_1.01,0.16014400482177735
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,1,1,power_law_1.01,0.17591808319091798
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,1,1,power_law_1.01,0.1890732765197754
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,1,1,power_law_1.01,0.10997056007385253
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,1,1,power_law_1.01,0.21884799957275392
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,1,1,power_law_1.01,0.10739199638366699
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,1,1,power_law_1.01,0.2828096008300781
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,1,1,power_law_1.01,0.31306047439575196
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,1,1,power_law_1.01,0.3774310302734375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,1,1,power_law_1.01,0.46119678497314454
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,1,1,power_law_1.01,0.5227961730957031
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,1,1,power_law_1.01,0.666917724609375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,1,1,power_law_1.01,0.9620185852050781
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,1,1,power_law_1.01,1.2504422760009766
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,1,1,power_law_1.01,1.8312364196777344
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,1,1,power_law_1.01,2.4135090637207033
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,1,1,power_law_1.01,3.5856326293945315
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,1,1,power_law_1.01,4.739326782226563
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,1,1,power_law_1.01,7.0483898925781245
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,1,power_law_1.2,0.14764863967895508
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,1,power_law_1.2,0.5131103897094726
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,1,power_law_1.2,0.9252102661132813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,1,power_law_1.2,1.4968984985351563
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,1,power_law_1.2,0.28987199783325196
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,1,power_law_1.2,2.0881869506835935
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,1,power_law_1.2,2.482974090576172
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,1,power_law_1.2,3.3650457763671873
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,1,power_law_1.2,3.6600302124023436
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,1,power_law_1.2,3.816837768554687
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,1,power_law_1.2,3.8155712890624995
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,1,power_law_1.2,4.287705688476562
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,1,power_law_1.2,4.310536193847656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,1,power_law_1.2,4.853855285644531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,1,power_law_1.2,5.092156066894531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,1,power_law_1.2,5.5895361328124995
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,1,power_law_1.2,5.320332641601563
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,1,power_law_1.2,6.119712524414062
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,1,power_law_1.2,6.629097900390624
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,1,power_law_1.2,8.194735107421874
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,1,power_law_1.2,8.94394287109375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,1,power_law_1.2,11.553521728515625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,1,power_law_1.2,12.3579638671875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,1,power_law_1.2,15.8909423828125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,1,power_law_1.2,19.32480224609375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,1,power_law_1.2,0.09546751976013183
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,1,power_law_1.2,0.15326592445373535
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,1,power_law_1.2,0.268221435546875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,1,power_law_1.2,0.47760383605957035
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,1,power_law_1.2,0.7632012939453124
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,1,power_law_1.2,1.0919551849365234
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,1,power_law_1.2,1.2987091064453125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,1,power_law_1.2,1.7363923645019532
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,1,power_law_1.2,27.870751953125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,1,power_law_1.2,1.828362274169922
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,1,power_law_1.2,1.953387451171875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,1,power_law_1.2,1.9475736999511717
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,1,power_law_1.2,2.118732757568359
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,1,power_law_1.2,2.093251190185547
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,1,power_law_1.2,2.204552917480469
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,1,power_law_1.2,32.46180419921875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,1,power_law_1.2,2.2155116271972655
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,1,power_law_1.2,2.4634515380859376
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,1,power_law_1.2,2.7182290649414065
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,1,power_law_1.2,2.9870379638671873
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,1,power_law_1.2,3.292109375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,1,power_law_1.2,3.413386840820313
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,1,power_law_1.2,3.7401809692382812
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,1,1,power_law_1.2,40.08693115234375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,1,power_law_1.2,4.685550842285156
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,1,power_law_1.2,4.878164367675781
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,1,power_law_1.2,6.047412719726562
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,1,power_law_1.2,7.34008056640625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,1,power_law_1.2,10.052066650390625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,1,1,power_law_1.2,0.10487232208251954
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,1,power_law_1.2,11.43758544921875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,1,1,power_law_1.2,0.10671232223510743
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,1,1,power_law_1.2,13.512548828125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,1,1,power_law_1.2,0.11734911918640137
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,1,1,power_law_1.2,0.1458035182952881
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,1,1,power_law_1.2,0.13194047927856445
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,1,1,power_law_1.2,0.16104320526123045
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,1,1,power_law_1.2,0.17569215774536134
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,1,1,power_law_1.2,0.18987455368041992
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,1,1,power_law_1.2,0.21914304733276366
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,1,1,power_law_1.2,0.283001594543457
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,1,1,power_law_1.2,0.3779916763305664
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,1,1,power_law_1.2,0.46037567138671875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,1,1,power_law_1.2,0.5251136016845703
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,1,1,power_law_1.2,0.10570816040039062
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,1,1,power_law_1.2,0.11002559661865234
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,1,1,power_law_1.2,0.3157913589477539
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,1,1,power_law_1.2,0.9629183959960936
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,1,1,power_law_1.2,0.6679206085205078
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,1,1,power_law_1.2,1.2498892974853515
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,1,1,power_law_1.2,2.418254089355469
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,1,1,power_law_1.2,3.573619689941406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,1,1,power_law_1.2,1.8334924316406251
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,1,1,power_law_1.2,4.7320275878906255
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,1,1,power_law_1.2,7.065162963867188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,1,power_law_1.01,0.10413184165954589
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,1,power_law_1.01,0.1645996856689453
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,1,power_law_1.01,0.23587776184082032
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,1,power_law_1.01,0.31185983657836913
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,1,power_law_1.01,0.47839809417724605
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,1,power_law_1.01,0.5048479843139648
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,1,power_law_1.01,0.5293977737426758
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,1,power_law_1.01,0.5481139373779297
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,1,power_law_1.01,0.5626028823852539
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,1,power_law_1.01,0.5859539031982421
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,1,power_law_1.01,0.6488358306884765
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,1,power_law_1.01,0.6654720306396484
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,1,power_law_1.01,0.6853939056396484
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,1,power_law_1.01,0.6940576171874999
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,1,power_law_1.01,0.7030496215820312
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,1,power_law_1.01,0.7495532989501953
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,1,power_law_1.01,0.8278662109375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,1,power_law_1.01,0.9010527801513671
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,1,power_law_1.01,1.0656396484375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,1,power_law_1.01,1.330645751953125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,1,power_law_1.01,1.740113983154297
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,1,power_law_1.01,2.157117462158203
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,1,power_law_1.01,2.9383526611328126
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,1,power_law_1.01,3.6499053955078127
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,1,power_law_1.01,7.2338452148437495
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,1,power_law_1.01,5.555100708007812
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,2688,1856,6,128,1,1,power_law_1.01,8.802903442382812
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,1,1,power_law_1.01,0.042685441970825196
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,1,1,power_law_1.01,0.11720640182495116
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,1,1,power_law_1.01,0.16700159072875978
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,1,1,power_law_1.01,0.17357055664062498
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,1,1,power_law_1.01,0.05306879997253418
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,1,1,power_law_1.01,0.18301952362060547
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,1,1,power_law_1.01,0.18638080596923828
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,1,1,power_law_1.01,0.1937984085083008
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,1,1,power_law_1.01,0.19687423706054688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,1,1,power_law_1.01,0.2073740768432617
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,1,1,power_law_1.01,0.20326528549194336
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,1,1,power_law_1.01,0.21450624465942383
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,1,1,power_law_1.01,0.22291135787963867
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,1,1,power_law_1.01,0.22698432922363282
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,1,1,power_law_1.01,0.09256511688232422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,1,1,power_law_1.01,0.06881087779998779
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,1,1,power_law_1.01,0.2779795265197754
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,1,1,power_law_1.01,0.24605567932128905
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,1,1,power_law_1.01,0.31619264602661135
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,1,1,power_law_1.01,0.3923238372802734
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,1,1,power_law_1.01,0.47041599273681645
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,1,1,power_law_1.01,0.7918943786621094
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,1,1,power_law_1.01,0.6282182312011718
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,1,1,power_law_1.01,1.1318495941162108
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,1,1,power_law_1.01,1.4780307006835938
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,1,1,power_law_1.01,2.172613067626953
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,1,1,power_law_1.01,2.929066162109375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,2688,1856,6,128,1,1,power_law_1.01,3.6447198486328127
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,1,power_law_1.2,0.16256767272949219
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,1,power_law_1.2,0.2969599914550781
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,1,power_law_1.2,0.06381375789642334
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,1,power_law_1.2,0.4771116638183594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,1,power_law_1.2,0.5388351821899414
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,1,power_law_1.2,0.08652288436889648
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,1,power_law_1.2,0.5060441589355469
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,1,power_law_1.2,0.5554272079467774
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,1,power_law_1.2,0.5994777679443359
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,1,power_law_1.2,0.667891845703125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,1,power_law_1.2,0.6842880249023438
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,1,power_law_1.2,0.7013619232177735
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,1,power_law_1.2,0.5703468704223633
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,1,power_law_1.2,0.7107923126220703
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,1,power_law_1.2,0.7395475006103516
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,1,power_law_1.2,0.7665203094482422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,1,power_law_1.2,0.8507942199707031
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,1,power_law_1.2,0.9227494049072267
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,1,power_law_1.2,1.118042221069336
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,1,power_law_1.2,0.22423999786376952
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,1,power_law_1.2,1.3013594055175781
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,1,power_law_1.2,1.7893907165527345
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,1,power_law_1.2,2.114245758056641
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,1,power_law_1.2,2.8791595458984376
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,1,power_law_1.2,3.8366827392578124
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,1,1,power_law_1.2,0.04262847900390625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,1,1,power_law_1.2,0.05354112148284912
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,1,1,power_law_1.2,0.06940864086151123
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,1,1,power_law_1.2,0.08717951774597169
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,1,power_law_1.2,5.25745361328125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,1,1,power_law_1.2,0.11013055801391602
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,1,1,power_law_1.2,0.16649152755737304
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,1,1,power_law_1.2,0.17595264434814453
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,1,1,power_law_1.2,0.1859833526611328
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,1,1,power_law_1.2,0.19036031723022462
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,1,1,power_law_1.2,0.19581504821777346
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,1,1,power_law_1.2,0.19996608734130858
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,1,1,power_law_1.2,0.2102956771850586
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,1,1,power_law_1.2,0.21423999786376954
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,1,1,power_law_1.2,0.22581760406494142
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,1,power_law_1.2,7.043001708984375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,1,1,power_law_1.2,0.2479884719848633
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,2688,1856,6,128,1,1,power_law_1.2,8.859385986328125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,1,1,power_law_1.2,0.23502464294433595
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,1,1,power_law_1.2,0.2691059112548828
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,1,1,power_law_1.2,0.3211648178100586
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,1,1,power_law_1.2,0.37216255187988284
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,1,1,power_law_1.2,0.48794624328613284
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,1,1,power_law_1.2,0.6065260696411132
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,1,1,power_law_1.2,0.8704972839355468
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,1,1,power_law_1.2,1.1414527893066406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,1,1,power_law_1.2,1.684405059814453
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,1,1,power_law_1.2,2.2585728454589846
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,1,1,power_law_1.2,3.5044390869140627
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,1,power_law_1.01,0.2976959991455078
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,1,power_law_1.01,0.5022137451171875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,1,power_law_1.01,0.9322547149658202
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,1,power_law_1.01,1.6054733276367188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,1,power_law_1.01,2.2621472167968752
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,1,1,power_law_1.2,4.668251647949218
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,1,power_law_1.01,3.8369741821289063
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,1,power_law_1.01,4.054913330078125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,1,power_law_1.01,4.512153015136719
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,1,power_law_1.01,4.282508850097656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,1,power_law_1.01,4.520773010253906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,1,power_law_1.01,4.995530090332031
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,1,power_law_1.01,5.253955078125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,1,power_law_1.01,5.564541625976562
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,1,power_law_1.01,6.5465588378906245
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,1,power_law_1.01,7.101640625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,1,power_law_1.01,7.795042724609376
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,2688,1856,6,128,1,1,power_law_1.2,5.83435791015625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,1,power_law_1.01,9.238273315429687
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,1,power_law_1.01,7.940668334960938
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,1,power_law_1.01,8.85472412109375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,1,power_law_1.01,11.638302001953125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,1,power_law_1.01,12.932620849609375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,1,power_law_1.01,16.186923828125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,1,power_law_1.01,19.489783935546875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,1,power_law_1.01,28.1654345703125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,1,power_law_1.01,33.67060546875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,1,power_law_1.01,0.17468544006347658
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,1,power_law_1.01,0.2841574478149414
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,1,power_law_1.01,0.5144044876098632
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,1,power_law_1.01,46.8006982421875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,1,power_law_1.01,0.8616217803955077
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,1,power_law_1.01,2.033072052001953
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,1,power_law_1.01,1.2463488006591796
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,1,power_law_1.01,2.1789170837402345
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,1,power_law_1.01,2.335491790771484
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,1,power_law_1.01,2.2650131225585937
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,1,power_law_1.01,60.019316406250006
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,1,power_law_1.01,2.413868865966797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,1,power_law_1.01,2.4699136352539064
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,1,power_law_1.01,2.581742858886719
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,1,power_law_1.01,2.645992431640625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,1,power_law_1.01,2.7144818115234375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,1,power_law_1.01,2.477763824462891
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,1,power_law_1.01,2.8440338134765626
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,1,power_law_1.01,3.0093350219726562
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,1,power_law_1.01,3.777780456542969
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,1,power_law_1.01,3.458248291015625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,1,power_law_1.01,4.464396057128907
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,1,power_law_1.01,4.974783935546875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,1,power_law_1.01,6.023079223632813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,1,power_law_1.01,7.357546997070313
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,1,power_law_1.01,9.874379272460938
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,1,power_law_1.01,12.057805175781251
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,1,power_law_1.01,16.72789794921875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,1,power_law_1.01,20.025419921875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,1,1,power_law_1.01,0.16150527954101562
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,1,1,power_law_1.01,0.16723583221435548
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,1,1,power_law_1.01,0.17697792053222655
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,1,1,power_law_1.01,0.16391359329223634
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,1,1,power_law_1.01,0.195295352935791
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,1,1,power_law_1.01,0.21258495330810545
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,1,1,power_law_1.01,0.2307948875427246
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,1,1,power_law_1.01,0.24829824447631835
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,1,1,power_law_1.01,0.1591584014892578
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,1,1,power_law_1.01,0.3004057693481445
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,1,1,power_law_1.01,0.2656703948974609
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,1,1,power_law_1.01,0.40583232879638675
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,1,1,power_law_1.01,0.3698534393310547
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,1,1,power_law_1.01,0.475109748840332
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,1,1,power_law_1.01,0.5832716751098632
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,1,1,power_law_1.01,0.6515020751953126
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,1,1,power_law_1.01,0.8389151763916015
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,1,1,power_law_1.01,1.1949414062500001
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,1,1,power_law_1.01,1.5645114135742186
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,1,1,power_law_1.01,3.0091058349609376
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,1,1,power_law_1.01,2.2785760498046876
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,1,1,power_law_1.01,4.4337048339843745
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,1,1,power_law_1.01,5.856219482421875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,1,1,power_law_1.01,8.753610229492187
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,1,1,power_law_1.01,11.5669970703125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,1,power_law_1.2,0.3236537551879883
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,1,power_law_1.2,0.9341081237792969
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,1,power_law_1.2,1.4749600219726564
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,1,power_law_1.2,2.0463629150390625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,1,power_law_1.2,3.8607162475585937
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,1,power_law_1.2,4.302653503417969
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,1,power_law_1.2,4.489007263183593
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,1,power_law_1.2,4.407058410644531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,1,power_law_1.2,4.991329956054687
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,1,power_law_1.2,5.005488586425781
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,1,power_law_1.2,5.790478515625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,1,power_law_1.2,5.794221191406249
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,1,power_law_1.2,6.946697387695312
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,1,power_law_1.2,7.388035278320312
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,1,power_law_1.2,8.49343994140625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,1,power_law_1.2,9.19917236328125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,1,power_law_1.2,8.894628295898437
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,1,power_law_1.2,9.240001220703125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,1,power_law_1.2,11.065191650390625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,1,power_law_1.2,13.661866455078126
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,1,power_law_1.2,16.03547607421875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,1,power_law_1.2,19.747751464843752
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,1,power_law_1.2,26.3258251953125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,1,power_law_1.2,34.62358154296875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,1,power_law_1.2,0.1759129524230957
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,1,power_law_1.2,0.2857939147949219
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,1,power_law_1.2,0.5155020904541016
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,1,power_law_1.2,0.7952365112304688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,1,power_law_1.2,1.132360305786133
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,1,power_law_1.2,2.0437831115722656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,1,power_law_1.2,46.971669921875005
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,1,power_law_1.2,2.254288635253906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,1,power_law_1.2,2.389559631347656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,1,power_law_1.2,2.2352665710449218
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,1,power_law_1.2,2.5601278686523434
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,1,power_law_1.2,2.5185005187988283
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,1,power_law_1.2,2.62889404296875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,1,power_law_1.2,2.6120184326171874
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,1,power_law_1.2,2.8504122924804687
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,1,power_law_1.2,2.850172119140625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,1,power_law_1.2,3.1245022583007813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,1,power_law_1.2,3.228071594238281
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,1,power_law_1.2,3.8750668334960934
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,1,power_law_1.2,4.099479064941407
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,1,power_law_1.2,4.902010803222656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,1,power_law_1.2,5.138112182617188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,1,power_law_1.2,6.560086669921875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,1,power_law_1.2,62.935273437499994
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,1,power_law_1.2,7.462277221679687
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,1,power_law_1.2,10.788813476562499
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,1,power_law_1.2,12.3869384765625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,1,power_law_1.2,16.479036865234374
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,1,1,power_law_1.2,0.16164159774780273
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,1,1,power_law_1.2,0.1646784019470215
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,1,power_law_1.2,21.56259765625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,1,1,power_law_1.2,0.16758592605590822
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,1,1,power_law_1.2,0.1762656021118164
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,1,1,power_law_1.2,0.19555904388427733
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,1,1,power_law_1.2,0.21240575790405272
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,1,1,power_law_1.2,0.2308140754699707
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,1,1,power_law_1.2,0.24857856750488283
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,1,1,power_law_1.2,0.26568576812744144
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,1,1,power_law_1.2,0.3003027153015137
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,1,1,power_law_1.2,0.4068115234375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,1,1,power_law_1.2,0.4752166366577148
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,1,1,power_law_1.2,0.15844544410705566
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,1,1,power_law_1.2,0.5848012924194336
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,1,1,power_law_1.2,0.3714694213867188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,1,1,power_law_1.2,0.6513401794433593
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,1,1,power_law_1.2,0.835555191040039
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,1,1,power_law_1.2,1.198252182006836
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,1,1,power_law_1.2,2.283891906738281
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,1,1,power_law_1.2,1.5579206848144531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,1,1,power_law_1.2,3.0115264892578124
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,1,1,power_law_1.2,4.438521728515625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,1,1,power_law_1.2,5.85712890625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,1,1,power_law_1.2,8.74972900390625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,1,1,power_law_1.2,11.49189697265625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,2,1,power_law_1.01,0.1521356773376465
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,2,1,power_law_1.01,0.16053504943847657
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,2,1,power_law_1.01,0.20211135864257815
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,2,1,power_law_1.01,0.26580928802490233
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,2,1,power_law_1.01,0.32277889251708985
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,2,1,power_law_1.01,0.3249164962768555
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,2,1,power_law_1.01,0.327086067199707
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,2,1,power_law_1.01,0.32619647979736327
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,2,1,power_law_1.01,0.3285951995849609
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,2,1,power_law_1.01,0.3451571273803711
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,2,1,power_law_1.01,0.3810073471069336
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,2,1,power_law_1.01,0.3788224029541015
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,2,1,power_law_1.01,0.39324222564697264
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,2,1,power_law_1.01,0.42245952606201176
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,2,1,power_law_1.01,0.46820865631103514
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,2,1,power_law_1.01,0.5414220809936523
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,2,1,power_law_1.01,0.6170521545410156
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,2,1,power_law_1.01,0.7655987548828125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,2,1,power_law_1.01,0.9305766296386718
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,2,1,power_law_1.01,1.3171026611328125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,2,1,power_law_1.01,1.6270016479492189
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,2,1,power_law_1.01,2.5364959716796873
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,2,1,power_law_1.01,3.404535827636719
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,2,1,power_law_1.01,4.99876220703125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,2,1,power_law_1.01,6.802272338867188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,2,1,power_law_1.01,10.982142333984374
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,2,1,power_law_1.01,0.09789183616638183
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,2,1,power_law_1.01,0.11039744377136232
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,2,1,power_law_1.01,0.14838656425476074
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,2,1,power_law_1.01,0.16089536666870116
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,2,1,power_law_1.01,0.16176063537597657
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,2,1,power_law_1.01,0.07102335929870605
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,2,1,power_law_1.01,0.1643391990661621
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,2,1,power_law_1.01,0.16287935256958008
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,2,1,power_law_1.01,0.15597760200500488
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,2,1,power_law_1.01,0.1671673583984375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,2,1,power_law_1.01,0.16382015228271485
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,2,1,power_law_1.01,0.1672902488708496
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,2,1,power_law_1.01,0.17447423934936523
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,2,1,power_law_1.01,0.1992748832702637
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,2,1,power_law_1.01,0.20756864547729492
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,2,1,power_law_1.01,0.21771135330200195
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,2,1,power_law_1.01,0.22604608535766602
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,2,1,power_law_1.01,0.2677222442626953
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,2,1,power_law_1.01,0.32354942321777347
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,2,1,power_law_1.01,0.41664768218994136
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,2,1,power_law_1.01,0.49009536743164067
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,2,1,power_law_1.01,0.6872492980957031
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,2,1,power_law_1.01,0.8750195312500001
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,2,1,power_law_1.01,1.281610870361328
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,2,1,power_law_1.01,18.447160644531248
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,2,1,power_law_1.01,1.7352876281738283
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,2,1,power_law_1.01,2.8015451049804687
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,2,1,power_law_1.01,3.710516357421875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,2,1,power_law_1.01,0.03578176021575928
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,2,1,power_law_1.01,0.08177663803100585
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,2,1,power_law_1.01,0.0828831958770752
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,2,1,power_law_1.01,0.08487104415893555
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,2,1,power_law_1.01,0.08503552436828613
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,2,1,power_law_1.01,0.08463616371154785
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,2,1,power_law_1.01,0.050079998970031736
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,2,1,power_law_1.01,0.08481663703918457
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,2,1,power_law_1.01,0.08514176368713379
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,2,1,power_law_1.01,0.08691200256347656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,2,1,power_law_1.01,4.770308532714844
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,2,1,power_law_1.01,0.06194623947143555
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,2,1,power_law_1.01,0.08712639808654785
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,2,1,power_law_1.01,0.09049792289733886
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,2,1,power_law_1.01,0.10000576019287108
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,2,1,power_law_1.01,0.10461567878723144
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,2,1,power_law_1.01,0.10801664352416993
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,2,1,power_law_1.01,0.12126463890075682
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,2,1,power_law_1.01,0.15142080307006836
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,2,1,power_law_1.01,0.18435264587402345
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,2,1,power_law_1.01,0.24168512344360354
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,2,1,power_law_1.01,0.30451904296875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,2,1,power_law_1.01,0.4227468872070313
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,2,1,power_law_1.01,0.547591667175293
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,2,1,power_law_1.01,0.78989501953125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,2,1,power_law_1.01,1.0792179107666016
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,2,1,power_law_1.2,0.11774847984313966
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,2,1,power_law_1.2,0.15813183784484863
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,2,1,power_law_1.2,0.1736096000671387
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,2,1,power_law_1.2,0.26411327362060544
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,2,1,power_law_1.2,0.32501758575439454
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,2,1,power_law_1.2,0.3336403274536133
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,2,1,power_law_1.2,0.3360806274414062
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,2,1,power_law_1.2,0.3368473434448242
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,2,1,power_law_1.2,0.3413587188720703
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,2,1,power_law_1.2,0.35551742553710936
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,2,1,power_law_1.2,0.38541183471679685
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,2,1,power_law_1.2,0.3890591812133789
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,2,1,power_law_1.2,0.39476734161376953
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,2,1,power_law_1.2,0.4264275360107422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,2,1,power_law_1.2,0.4610400009155273
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,2,1,power_law_1.2,0.559233283996582
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,2,1,power_law_1.2,0.6181235122680664
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,2,1,power_law_1.2,0.7744083404541016
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,2,1,power_law_1.2,0.938837127685547
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,2,1,power_law_1.01,1.603724822998047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,2,1,power_law_1.2,1.2813229370117187
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,2,1,power_law_1.01,2.1389779663085937
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,2,1,power_law_1.2,1.6646092224121094
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,2,1,power_law_1.2,2.5467321777343748
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,2,1,power_law_1.2,3.2405044555664064
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,2,1,power_law_1.2,5.111796569824219
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,2,1,power_law_1.01,2.646198425292969
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,2,1,power_law_1.2,6.838441772460937
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,2,1,power_law_1.2,0.07502528190612792
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,2,1,power_law_1.2,0.09989439964294433
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,2,1,power_law_1.2,0.1055513572692871
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,2,1,power_law_1.2,0.14947903633117676
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,2,1,power_law_1.2,0.15572671890258788
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,2,1,power_law_1.2,0.16180736541748048
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,2,1,power_law_1.2,0.16256959915161134
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,2,1,power_law_1.2,0.161911678314209
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,2,1,power_law_1.2,0.16484352111816408
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,2,1,power_law_1.2,0.16627904891967774
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,2,1,power_law_1.2,0.16747135162353516
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,2,1,power_law_1.2,0.16990272521972655
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,2,1,power_law_1.2,0.1760915184020996
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,2,1,power_law_1.2,0.19786176681518555
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,2,1,power_law_1.2,0.2154003143310547
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,2,1,power_law_1.2,0.22506879806518554
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,2,1,power_law_1.2,0.22686208724975584
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,2,1,power_law_1.2,0.2661087989807129
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,2,1,power_law_1.2,0.32468544006347655
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,2,1,power_law_1.2,0.4026694488525391
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,2,1,power_law_1.2,0.49169857025146485
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,2,1,power_law_1.2,10.4236376953125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,2,1,power_law_1.2,0.6981203460693359
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,2,1,power_law_1.2,0.8989100646972655
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,2,1,power_law_1.2,1.3480531311035155
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,2,1,power_law_1.2,1.804973449707031
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,2,1,power_law_1.2,14.50815185546875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,2,1,power_law_1.2,18.229963378906252
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,2,1,power_law_1.2,0.038065280914306644
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,2,1,power_law_1.2,0.04839871883392334
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,2,1,power_law_1.2,0.05369984149932862
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,2,1,power_law_1.2,0.0826144027709961
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,2,1,power_law_1.2,0.0823526382446289
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,2,1,power_law_1.2,0.08415488243103028
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,2,1,power_law_1.2,0.08445311546325683
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,2,1,power_law_1.2,0.08394495964050293
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,2,1,power_law_1.2,0.08535231590270996
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,2,1,power_law_1.2,0.08533568382263183
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,2,1,power_law_1.2,0.08634943962097168
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,2,1,power_law_1.2,0.08767616271972656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,2,1,power_law_1.2,0.09042688369750976
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,2,1,power_law_1.2,0.1006054401397705
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,2,1,power_law_1.2,0.1070470428466797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,2,1,power_law_1.2,0.11252032279968263
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,2,1,power_law_1.2,2.8883584594726566
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,2,1,power_law_1.2,0.12310720443725587
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,2,1,power_law_1.2,3.749718322753906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,2,1,power_law_1.2,0.15593791961669923
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,2,1,power_law_1.2,0.3100320053100586
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,2,1,power_law_1.2,0.18511871337890623
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,2,1,power_law_1.2,4.785245971679688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,2,1,power_law_1.2,0.2466579246520996
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,2,1,power_law_1.2,0.445588493347168
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,2,1,power_law_1.2,0.5580217742919922
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,2,1,power_law_1.2,0.8506483459472657
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,2,1,power_law_1.2,1.0874092864990235
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,2,1,power_law_1.01,0.15658944129943847
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,2,1,power_law_1.01,0.2562259292602539
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,2,1,power_law_1.01,0.3124038314819336
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,2,1,power_law_1.01,0.4438431930541992
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,2,1,power_law_1.01,0.501635856628418
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,2,1,power_law_1.01,0.5235641479492188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,2,1,power_law_1.01,0.5499270248413086
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,2,1,power_law_1.01,0.5415097427368164
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,2,1,power_law_1.01,0.547056007385254
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,2,1,power_law_1.01,0.5609094238281249
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,2,1,power_law_1.01,0.5917606353759766
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,2,1,power_law_1.01,0.633314552307129
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,2,1,power_law_1.01,0.67704833984375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,2,1,power_law_1.01,0.7378630065917969
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,2,1,power_law_1.01,0.7747545623779297
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,2,1,power_law_1.01,0.8187641906738282
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,2,1,power_law_1.01,0.9792371368408203
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,2,1,power_law_1.01,1.2814259338378906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,2,1,power_law_1.01,1.6476185607910157
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,2,1,power_law_1.01,2.2825241088867188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,2,1,power_law_1.01,3.025087890625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,2,1,power_law_1.01,4.394645080566407
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,2,1,power_law_1.2,1.6967379760742187
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,2,1,power_law_1.01,5.7866143798828125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,2,1,power_law_1.01,9.300914306640625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,2,1,power_law_1.01,12.2946044921875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,2,1,power_law_1.2,2.1343865966796876
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,2,1,power_law_1.01,19.0846337890625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,2,1,power_law_1.01,0.09117888450622559
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,2,1,power_law_1.01,0.13818367958068847
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,2,1,power_law_1.01,0.16715967178344726
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,2,1,power_law_1.01,0.24295936584472658
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,2,1,power_law_1.01,0.2628057670593262
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,2,1,power_law_1.01,0.26482751846313474
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,2,1,power_law_1.01,0.2662892723083496
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,2,1,power_law_1.01,0.26173440933227543
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,2,1,power_law_1.01,0.2636128044128418
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,2,1,power_law_1.01,0.26830463409423827
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,2,1,power_law_1.01,0.27034879684448243
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,2,1,power_law_1.01,0.26827903747558596
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,2,1,power_law_1.01,0.27486143112182615
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,2,1,power_law_1.01,0.2868979263305664
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,2,1,power_law_1.01,0.2917094421386719
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,2,1,power_law_1.01,0.3048723220825195
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,2,1,power_law_1.01,0.3351161575317383
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,2,1,power_law_1.01,0.40416000366210936
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,2,1,power_law_1.01,0.47540737152099605
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,2,1,power_law_1.01,25.3556640625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,2,1,power_law_1.01,0.61770751953125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,2,1,power_law_1.2,2.7397753906249998
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,2,1,power_law_1.01,0.7580671691894532
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,2,1,power_law_1.01,1.0965599822998047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,2,1,power_law_1.01,1.607898864746094
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,2,1,power_law_1.01,2.3988659667968752
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,2,1,power_law_1.01,3.2075347900390625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,2,1,power_law_1.01,4.770588073730469
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,2,1,power_law_1.01,33.1760693359375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,2,1,power_law_1.01,6.125569458007813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,2,1,power_law_1.01,8.137053833007812
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,2,1,power_law_1.01,0.048414721488952636
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,2,1,power_law_1.01,0.13406720161437988
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,2,1,power_law_1.01,0.13443072319030763
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,2,1,power_law_1.01,0.13440511703491212
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,2,1,power_law_1.01,0.13224384307861328
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,2,1,power_law_1.01,0.13271552085876465
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,2,1,power_law_1.01,0.13612863540649414
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,2,1,power_law_1.01,0.13756671905517578
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,2,1,power_law_1.01,0.09042431831359862
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,2,1,power_law_1.01,0.07811200141906738
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,2,1,power_law_1.01,0.13017408370971678
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,2,1,power_law_1.01,0.13679679870605468
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,2,1,power_law_1.01,0.14251903533935545
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,2,1,power_law_1.01,0.15706944465637207
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,2,1,power_law_1.01,0.1590828800201416
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,2,1,power_law_1.01,0.1669011116027832
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,2,1,power_law_1.01,0.18819520950317384
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,2,1,power_law_1.01,0.2369651222229004
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,2,1,power_law_1.01,0.28296895980834963
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,2,1,power_law_1.01,0.3742572784423828
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,2,1,power_law_1.01,0.4727347183227539
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,2,1,power_law_1.01,0.8645951843261719
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,2,1,power_law_1.01,0.6604627227783203
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,2,1,power_law_1.01,1.3615782165527344
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,2,1,power_law_1.01,1.8237171936035157
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,2,1,power_law_1.2,0.16656511306762695
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,2,1,power_law_1.2,0.24213695526123047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,2,1,power_law_1.2,0.27817792892456056
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,2,1,power_law_1.2,0.4435635375976562
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,2,1,power_law_1.2,0.4754355239868164
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,2,1,power_law_1.2,0.5123443222045898
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,2,1,power_law_1.2,0.5434527969360351
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,2,1,power_law_1.2,0.5377433776855469
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,2,1,power_law_1.2,0.5538035202026367
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,2,1,power_law_1.2,0.5608582305908203
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,2,1,power_law_1.2,0.5874009704589843
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,2,1,power_law_1.2,0.6425202941894531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,2,1,power_law_1.2,0.6663545227050781
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,2,1,power_law_1.2,0.7329926300048828
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,2,1,power_law_1.2,0.7770079803466797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,2,1,power_law_1.2,0.829059829711914
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,2,1,power_law_1.2,0.9676774597167969
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,2,1,power_law_1.2,1.2854470825195312
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,2,1,power_law_1.2,1.5756672668457032
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,2,1,power_law_1.2,2.2733363342285156
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,2,1,power_law_1.2,2.9992410278320314
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,2,1,power_law_1.01,2.622839050292969
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,2,1,power_law_1.2,4.385471496582031
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,2,1,power_law_1.2,5.7395605468749995
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,2,1,power_law_1.01,3.563246765136719
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,2,1,power_law_1.01,4.599143676757812
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,2,1,power_law_1.2,9.096232299804687
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,2,1,power_law_1.2,12.3911572265625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,2,1,power_law_1.2,0.09935744285583496
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,2,1,power_law_1.2,0.13433792114257812
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,2,1,power_law_1.2,0.14791744232177734
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,2,1,power_law_1.2,0.24299711227416992
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,2,1,power_law_1.2,0.2525062370300293
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,2,1,power_law_1.2,0.2645561599731445
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,2,1,power_law_1.2,0.2665868759155273
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,2,1,power_law_1.2,0.2618176078796387
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,2,1,power_law_1.2,0.269299201965332
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,2,1,power_law_1.2,0.26881280899047855
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,2,1,power_law_1.2,0.27031167984008786
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,2,1,power_law_1.2,0.2731532859802246
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,2,1,power_law_1.2,0.27115392684936523
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,2,1,power_law_1.2,0.28777599334716797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,2,1,power_law_1.2,0.29723072052001953
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,2,1,power_law_1.2,0.3125056076049805
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,2,1,power_law_1.2,0.3481356811523438
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,2,1,power_law_1.2,0.40005249023437506
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,2,1,power_law_1.2,0.4758841705322266
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,2,1,power_law_1.2,0.6018041610717774
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,2,1,power_law_1.2,0.7384339141845703
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,2,1,power_law_1.2,1.200517120361328
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,2,1,power_law_1.2,18.85573974609375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,2,1,power_law_1.2,1.588792266845703
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,2,1,power_law_1.2,26.505458984375004
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,2,1,power_law_1.2,2.4064845275878906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,2,1,power_law_1.2,3.2757144165039067
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,2,1,power_law_1.2,34.21934814453125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,2,1,power_law_1.2,0.04805312156677246
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,2,1,power_law_1.2,0.07358208179473877
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,2,1,power_law_1.2,0.08376447677612305
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,2,1,power_law_1.2,0.12993408203125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,2,1,power_law_1.2,0.12836288452148437
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,2,1,power_law_1.2,0.13316672325134277
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,2,1,power_law_1.2,0.13328703880310058
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,2,1,power_law_1.2,0.13151616096496582
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,2,1,power_law_1.2,0.1339577579498291
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,2,1,power_law_1.2,0.13491456031799315
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,2,1,power_law_1.2,0.13607616424560548
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,2,1,power_law_1.2,0.13834112167358398
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,2,1,power_law_1.2,0.14285759925842284
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,2,1,power_law_1.2,0.15894975662231445
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,2,1,power_law_1.2,4.800296325683594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,2,1,power_law_1.2,0.17303232192993162
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,2,1,power_law_1.2,0.16822399139404295
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,2,1,power_law_1.2,6.1938146972656245
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,2,1,power_law_1.2,0.1915519905090332
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,2,1,power_law_1.2,0.37644161224365236
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,2,1,power_law_1.2,8.139537353515625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,2,1,power_law_1.2,0.24317695617675783
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,2,1,power_law_1.2,0.28408639907836913
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,2,1,power_law_1.2,0.4784678268432617
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,2,1,power_law_1.2,0.6812793731689453
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,2,1,power_law_1.2,0.8606988525390624
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,2,1,power_law_1.2,1.3140748596191405
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,2,1,power_law_1.2,2.7674105834960936
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,1,power_law_1.01,0.10010432243347167
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,2,1,power_law_1.2,1.7484300231933596
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,1,power_law_1.01,0.14227519989013673
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,1,power_law_1.01,0.4973664093017578
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,1,power_law_1.01,0.719656982421875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,1,power_law_1.01,0.9251872253417968
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,1,power_law_1.01,1.280166473388672
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,1,power_law_1.01,0.28201215744018554
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,1,power_law_1.01,1.3436947631835938
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,1,power_law_1.01,1.4026284790039063
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,1,power_law_1.01,1.3967800903320313
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,1,power_law_1.01,1.4819328308105468
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,1,power_law_1.01,1.6184217834472654
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,1,power_law_1.01,1.6435328674316405
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,1,power_law_1.01,1.8066784667968752
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,1,power_law_1.01,1.7853753662109377
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,1,power_law_1.01,1.8240345764160157
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,1,power_law_1.01,1.901848907470703
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,1,power_law_1.01,2.147199401855469
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,1,power_law_1.01,2.4271685791015623
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,1,power_law_1.01,3.0618560791015623
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,1,power_law_1.01,3.2754284667968747
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,1,power_law_1.01,4.341477661132812
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,1,power_law_1.01,4.918499145507813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,1,power_law_1.01,6.97197998046875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,1,power_law_1.01,8.648386840820312
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,1,power_law_1.01,12.697191162109375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,1,power_law_1.01,15.233509521484375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,1,power_law_1.01,0.06831168174743653
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,1,power_law_1.01,0.09606783866882325
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,1,power_law_1.01,0.15784832000732424
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,1,power_law_1.01,0.28293567657470703
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,2,1,power_law_1.2,3.625393981933594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,1,power_law_1.01,0.41248512268066406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,1,power_law_1.01,0.7178643035888672
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,1,power_law_1.01,0.75142333984375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,1,power_law_1.01,0.7691654205322266
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,1,power_law_1.01,0.7783238220214843
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,1,power_law_1.01,0.8124070739746093
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,1,power_law_1.01,0.8359744262695312
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,2,1,power_law_1.01,19.522491455078125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,1,power_law_1.01,0.5362771224975587
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,1,power_law_1.01,0.8497222137451171
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,1,power_law_1.01,0.8756006622314454
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,1,power_law_1.01,0.9452934265136719
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,1,power_law_1.01,0.9686041259765625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,1,power_law_1.01,1.0233990478515627
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,1,power_law_1.01,1.0342425537109374
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,1,power_law_1.01,1.128067855834961
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,1,power_law_1.01,1.2900518798828124
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,1,power_law_1.01,1.3213215637207032
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,1,power_law_1.01,1.6711424255371092
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,1,power_law_1.01,2.0373651123046876
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,1,power_law_1.01,2.4205946350097656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,1,power_law_1.01,3.0668145751953126
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,1,power_law_1.01,4.3650949096679685
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,1,power_law_1.01,5.0704653930664065
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,2,1,power_law_1.01,6.145966796874999
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,2,1,power_law_1.01,0.07144383907318115
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,2,1,power_law_1.01,0.1738431930541992
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,2,1,power_law_1.01,0.09147263526916503
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,2,1,power_law_1.01,0.24473024368286134
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,2,1,power_law_1.01,0.30884096145629886
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,2,1,power_law_1.01,0.1198361587524414
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,2,1,power_law_1.01,0.448603515625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,2,1,power_law_1.01,0.42582145690917966
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,2,1,power_law_1.01,0.45919166564941405
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,2,1,power_law_1.01,0.4670348739624023
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,2,1,power_law_1.01,0.4907865524291992
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,2,1,power_law_1.01,0.5073740768432617
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,2,1,power_law_1.01,0.5165983963012695
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,2,1,power_law_1.01,0.528985595703125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,2,1,power_law_1.01,0.5539123153686523
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,2,1,power_law_1.01,0.5607340621948242
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,2,1,power_law_1.01,0.5958380889892578
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,2,1,power_law_1.2,4.584223327636719
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,2,1,power_law_1.01,0.6537100982666015
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,2,1,power_law_1.01,0.7196339416503906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,2,1,power_law_1.01,0.8401427459716796
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,2,1,power_law_1.01,0.9779974365234375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,2,1,power_law_1.01,1.5471372985839844
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,2,1,power_law_1.01,1.2565715026855468
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,2,1,power_law_1.01,2.189181365966797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,2,1,power_law_1.01,2.805877685546875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,1,power_law_1.2,0.10115391731262206
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,1,power_law_1.2,0.14232576370239258
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,2,1,power_law_1.01,4.120841064453125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,1,power_law_1.2,0.28148479461669923
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,2,1,power_law_1.01,5.252872314453125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,1,power_law_1.2,0.6513651275634766
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,1,power_law_1.2,0.4499961471557617
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,1,power_law_1.2,0.8804486083984375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,1,power_law_1.2,1.2501651000976561
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,1,power_law_1.2,1.314385986328125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,1,power_law_1.2,1.353765106201172
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,7168,2048,8,256,2,1,power_law_1.01,6.565528564453125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,1,power_law_1.2,1.3989605712890625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,1,power_law_1.2,1.4800518798828126
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,1,power_law_1.2,1.606559295654297
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,1,power_law_1.2,1.7193760681152344
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,1,power_law_1.2,1.7930732727050782
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,1,power_law_1.2,1.814254150390625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,1,power_law_1.2,1.8821157836914062
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,1,power_law_1.2,1.985694122314453
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,1,power_law_1.2,2.142453155517578
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,1,power_law_1.2,2.416350708007813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,1,power_law_1.2,2.978949890136719
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,1,power_law_1.2,3.520712890625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,1,power_law_1.2,4.145618591308594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,1,power_law_1.2,5.023403015136719
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,1,power_law_1.2,6.82110107421875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,1,power_law_1.2,8.984042358398437
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,1,power_law_1.2,0.07020095825195313
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,1,power_law_1.2,0.09559935569763184
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,1,power_law_1.2,0.15780927658081054
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,1,power_law_1.2,11.8260302734375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,1,power_law_1.2,0.26044416427612305
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,1,power_law_1.2,0.3709990310668945
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,1,power_law_1.2,0.5154547119140626
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,1,power_law_1.2,0.7059629058837891
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,1,power_law_1.2,0.7357907104492187
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,1,power_law_1.2,0.7561740875244141
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,1,power_law_1.2,14.932999267578126
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,1,power_law_1.2,0.7816313934326172
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,1,power_law_1.2,0.8015187072753907
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,1,power_law_1.2,0.8268927764892577
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,1,power_law_1.2,0.8460492706298828
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,1,power_law_1.2,0.8747154998779297
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,1,power_law_1.2,0.9581120300292969
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,1,power_law_1.2,0.989908447265625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,1,power_law_1.2,1.0082527923583986
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,1,power_law_1.2,1.0860499572753906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,1,power_law_1.2,1.1849158477783202
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,1,power_law_1.2,1.2591423797607422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,1,power_law_1.2,1.4458169555664062
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,1,power_law_1.2,1.8101574707031252
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,1,power_law_1.2,1.9856448364257813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,2,1,power_law_1.2,18.486928710937498
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,1,power_law_1.2,2.5850439453125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,1,power_law_1.2,3.3394168090820315
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,2,1,power_law_1.2,0.07146431922912597
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,1,power_law_1.2,4.085545654296875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,2,1,power_law_1.2,0.09080960273742675
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,1,power_law_1.2,5.38654296875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,2,1,power_law_1.2,0.16448255538940432
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,2,1,power_law_1.2,0.22687744140625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,2,1,power_law_1.2,6.361193237304688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,2,1,power_law_1.2,0.2911584091186524
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,2,1,power_law_1.2,0.41805057525634765
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,2,1,power_law_1.2,0.4380928039550781
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,2,1,power_law_1.2,0.4537798309326172
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,2,1,power_law_1.2,0.4707968139648438
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,2,1,power_law_1.2,0.48968959808349605
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,2,1,power_law_1.2,0.5065811157226563
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,2,1,power_law_1.2,0.5220243072509765
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,2,1,power_law_1.2,0.5490291213989258
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,2,1,power_law_1.2,0.5707500839233399
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,2,1,power_law_1.2,0.5922841644287109
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,2,1,power_law_1.2,0.6363270568847657
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,2,1,power_law_1.2,0.11667263984680176
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,2,1,power_law_1.2,0.7297190093994141
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,2,1,power_law_1.2,0.8123583984374999
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,2,1,power_law_1.2,1.017611541748047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,2,1,power_law_1.2,1.6763487243652342
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,2,1,power_law_1.2,2.1575776672363283
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,2,1,power_law_1.2,3.1031732177734375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,2,1,power_law_1.2,1.2190290832519532
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,2,1,power_law_1.2,4.159757995605469
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,1,power_law_1.01,0.0808467197418213
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,1,power_law_1.01,0.08311488151550292
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,1,power_law_1.01,0.11796095848083496
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,1,power_law_1.01,0.12140928268432619
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,1,power_law_1.01,0.12390591621398925
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,1,power_law_1.01,0.12683199882507323
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,1,power_law_1.01,0.12742655754089355
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,1,power_law_1.01,0.13705792427062988
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,1,power_law_1.01,0.14716416358947754
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,1,power_law_1.01,0.14904000282287597
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,1,power_law_1.01,0.15279552459716797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,1,power_law_1.01,0.15700415611267088
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,1,power_law_1.01,0.1595372772216797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,1,power_law_1.01,0.1676192092895508
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,1,power_law_1.01,0.1941049575805664
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,1,power_law_1.01,0.21641279220581056
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,1,power_law_1.01,0.26349056243896485
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,1,power_law_1.01,0.3111071968078613
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,1,power_law_1.01,0.41072319030761717
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,1,power_law_1.01,0.5095001602172852
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,1,power_law_1.01,0.7097785949707032
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,1,power_law_1.01,0.905257568359375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,1,power_law_1.01,1.2951187133789062
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,1,power_law_1.01,1.6978604125976564
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,2048,768,8,128,2,1,power_law_1.01,2.0736647033691407
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,1,power_law_1.01,0.032115840911865236
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,1,power_law_1.01,0.036154239177703856
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,1,power_law_1.01,0.04645055770874024
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,1,power_law_1.01,0.059067521095275875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,1,power_law_1.01,0.0751475191116333
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,1,power_law_1.01,0.09623040199279785
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,1,power_law_1.01,0.09937919616699219
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,1,power_law_1.01,0.1005254364013672
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,1,power_law_1.01,0.1026028823852539
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,1,power_law_1.01,0.10398079872131347
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,1,power_law_1.01,0.10511103630065918
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,1,power_law_1.01,0.11916288375854492
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,1,power_law_1.01,0.12064319610595704
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,1,power_law_1.01,0.1225875186920166
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,1,power_law_1.01,0.12620863914489747
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,1,power_law_1.01,0.12795583724975584
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,1,power_law_1.01,0.13436415672302246
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,1,power_law_1.01,0.140448637008667
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,1,power_law_1.01,0.1566867160797119
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,1,power_law_1.01,0.17876031875610351
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,1,power_law_1.01,0.21236928939819336
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,1,power_law_1.01,0.2646681594848633
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,1,power_law_1.01,0.31661951065063476
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,1,power_law_1.01,0.43196479797363285
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,1,power_law_1.01,0.5418137741088868
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,1,power_law_1.01,0.7694822692871094
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,1,power_law_1.01,0.9899533081054688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,2048,768,8,128,2,1,power_law_1.01,1.210537567138672
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,2,1,power_law_1.01,0.03699712038040161
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,2,1,power_law_1.01,0.05214399814605712
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,2,1,power_law_1.01,0.0635750389099121
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,2,1,power_law_1.01,0.06604991912841797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,2,1,power_law_1.01,0.06679296016693115
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,2,1,power_law_1.01,0.037931520938873295
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,2,1,power_law_1.01,0.06834559917449952
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,2,1,power_law_1.01,0.06784639835357667
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,2,1,power_law_1.01,0.04272704124450684
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,2,1,power_law_1.01,0.071112961769104
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,2,1,power_law_1.01,0.07157695770263672
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,2,1,power_law_1.01,0.07244287967681885
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,2,1,power_law_1.01,0.07786176204681397
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,2,1,power_law_1.01,0.04614848136901856
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,2,1,power_law_1.01,0.08246784210205078
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,2,1,power_law_1.01,0.08289728164672852
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,2,1,power_law_1.01,0.09202176094055176
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,2,1,power_law_1.01,0.10611200332641602
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,2,1,power_law_1.01,0.12438655853271485
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,2,1,power_law_1.01,0.15620415687561034
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,2,1,power_law_1.01,0.19276735305786133
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,2,1,power_law_1.01,0.2662816047668457
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,2,1,power_law_1.01,0.3432953643798828
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,2,1,power_law_1.01,0.5050387191772461
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,2,1,power_law_1.01,0.6805369567871093
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,1,power_law_1.2,0.028817920684814452
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,1,power_law_1.2,0.04145535945892334
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,1,power_law_1.2,0.053884158134460444
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,1,power_law_1.2,0.06543168067932129
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,1,power_law_1.2,0.07817728042602538
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,1,power_law_1.2,0.12180224418640137
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,1,power_law_1.2,0.12374591827392578
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,1,power_law_1.2,0.12650367736816406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,1,power_law_1.2,0.12819840431213378
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,1,power_law_1.2,0.13009023666381836
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,1,power_law_1.2,0.13790847778320312
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,1,power_law_1.2,0.14785023689270022
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,1,power_law_1.2,0.14900416374206543
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,1,power_law_1.2,0.15356608390808107
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,1,power_law_1.2,0.15764287948608396
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,1,power_law_1.2,0.160262393951416
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,1,power_law_1.2,0.1687513542175293
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,1,power_law_1.2,0.19978368759155274
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,1,power_law_1.2,0.217960319519043
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,1,power_law_1.2,0.26884031295776367
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,1,power_law_1.2,0.3122073554992676
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,1,power_law_1.2,0.41265727996826174
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,2,1,power_law_1.01,1.374554901123047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,1,power_law_1.2,0.5142464065551758
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,2,1,power_law_1.01,1.0218380737304689
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,2048,768,8,128,2,1,power_law_1.01,1.7297293090820312
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,1,power_law_1.2,0.7167731475830078
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,1,power_law_1.2,0.9104402923583985
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,1,power_law_1.2,0.03248512029647827
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,1,power_law_1.2,0.037246079444885255
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,1,power_law_1.2,0.04638976097106934
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,1,power_law_1.2,0.054145917892456055
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,1,power_law_1.2,0.07303552150726318
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,1,power_law_1.2,0.09665408134460449
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,1,power_law_1.2,0.09944319725036621
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,1,power_law_1.2,0.1011347198486328
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,1,power_law_1.2,0.10367808341979981
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,1,power_law_1.2,0.10414143562316895
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,1,power_law_1.2,0.10481087684631347
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,1,power_law_1.2,0.11952896118164062
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,1,power_law_1.2,0.12103615760803224
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,1,power_law_1.2,0.12392831802368165
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,1,power_law_1.2,0.12766016006469727
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,1,power_law_1.2,0.1299903964996338
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,1,power_law_1.2,0.1351046371459961
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,1,power_law_1.2,0.14334464073181152
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,1,power_law_1.2,0.16005376815795896
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,1,power_law_1.2,0.18624000549316405
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,1,power_law_1.2,0.21350591659545898
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,1,power_law_1.2,0.2653651237487793
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,1,power_law_1.2,0.3219014358520508
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,1,power_law_1.2,0.4338937759399414
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,1,power_law_1.2,1.3080665588378906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,1,power_law_1.2,0.5416729736328125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,1,power_law_1.2,1.6917645263671877
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,1,power_law_1.2,0.7721958160400391
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,2,1,power_law_1.2,0.03701632022857666
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,2,1,power_law_1.2,0.03991231918334961
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,2,1,power_law_1.2,0.0439628791809082
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,2048,768,8,128,2,1,power_law_1.2,2.083374786376953
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,2,1,power_law_1.2,0.04610047817230224
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,2,1,power_law_1.2,0.050749440193176266
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,2,1,power_law_1.2,0.0646988821029663
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,2,1,power_law_1.2,0.06648255825042725
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,2,1,power_law_1.2,0.06825920104980468
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,2,1,power_law_1.2,0.06960383892059327
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,2,1,power_law_1.2,0.07012608051300048
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,2,1,power_law_1.2,0.07209152221679688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,2,1,power_law_1.2,0.07695104122161865
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,2,1,power_law_1.2,0.07502336025238036
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,2,1,power_law_1.2,0.08188032150268555
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,2,1,power_law_1.2,0.09135744094848633
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,1,power_law_1.2,0.9904467010498047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,2,1,power_law_1.2,0.10155584335327147
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,2,1,power_law_1.2,0.08711423873901367
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,2,1,power_law_1.2,0.12439935684204102
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,2,1,power_law_1.2,0.1507692813873291
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,2,1,power_law_1.2,0.2054207992553711
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,2048,768,8,128,2,1,power_law_1.2,1.2131027221679687
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,2,1,power_law_1.2,0.2598924827575684
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,2,1,power_law_1.2,0.380682258605957
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,2,1,power_law_1.2,0.503628158569336
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,2,1,power_law_1.2,0.7478502655029298
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,2,1,power_law_1.2,0.9856998443603515
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,1,power_law_1.01,0.062449917793273926
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,1,power_law_1.01,0.08281344413757324
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,1,power_law_1.01,0.1275609588623047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,1,power_law_1.01,0.19753087997436522
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,1,power_law_1.01,0.23716159820556643
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,1,power_law_1.01,0.3479462432861328
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,1,power_law_1.01,0.36293121337890627
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,1,power_law_1.01,0.36858177185058594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,1,power_law_1.01,0.3936044692993164
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,1,power_law_1.01,0.39982334136962894
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,1,power_law_1.01,0.4294380950927734
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,1,power_law_1.01,0.4416345596313477
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,1,power_law_1.01,0.44813758850097657
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,1,power_law_1.01,0.46927486419677733
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,1,power_law_1.01,0.5073939132690429
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,1,power_law_1.01,0.5125030517578125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,1,power_law_1.01,0.529375991821289
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,1,power_law_1.01,0.5962566375732422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,1,power_law_1.01,0.6550828552246093
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,1,power_law_1.01,0.798779525756836
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,1,power_law_1.01,0.9748863983154298
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,1,power_law_1.01,1.3599244689941405
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,1,power_law_1.01,1.7187199401855469
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,1,power_law_1.01,2.4070477294921875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,1,power_law_1.01,3.14723388671875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,2,1,power_law_1.2,1.4712812805175781
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,1,power_law_1.01,4.590073547363281
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,1,power_law_1.01,5.740386352539063
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,2,1,power_law_1.01,7.699713745117188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,1,power_law_1.01,0.06239359855651856
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,1,power_law_1.01,0.09647104263305664
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,1,power_law_1.01,0.12591936111450194
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,1,power_law_1.01,0.03906048059463501
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,1,power_law_1.01,0.16162111282348632
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,1,power_law_1.01,0.2309734344482422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,1,power_law_1.01,0.23504575729370117
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,1,power_law_1.01,0.24159936904907225
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,1,power_law_1.01,0.24144384384155276
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,1,power_law_1.01,0.24853696823120117
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,1,power_law_1.01,0.2772524833679199
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,1,power_law_1.01,0.28088960647583006
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,1,power_law_1.01,0.2857567977905274
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,1,power_law_1.01,0.22109760284423827
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,1,power_law_1.01,0.29678527832031254
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,1,power_law_1.01,0.3005900764465332
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,1,power_law_1.01,0.3158316802978516
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,1,power_law_1.01,0.3325862503051758
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,1,power_law_1.01,0.3607020950317383
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,1,power_law_1.01,0.41719425201416016
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,1,power_law_1.01,0.4853139114379883
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,1,power_law_1.01,0.6027366256713866
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,1,power_law_1.01,0.7219891357421875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,1,power_law_1.01,0.9733216094970704
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,1,power_law_1.01,1.2189017486572264
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,1,power_law_1.01,1.7187008666992187
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,1,power_law_1.01,2.2430995178222655
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,2,1,power_law_1.01,2.691400451660156
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,2,1,power_law_1.01,0.03965183973312378
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,2,1,power_law_1.01,0.062489600181579584
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,2,1,power_law_1.01,0.05047359943389893
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,2,1,power_law_1.01,0.07699840068817139
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,2,1,power_law_1.01,0.09206080436706543
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,2,1,power_law_1.01,0.12697152137756346
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,2,1,power_law_1.01,0.1360102367401123
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,2,1,power_law_1.01,0.13834752082824706
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,2,1,power_law_1.01,0.14359935760498047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,2,1,power_law_1.01,0.1441427230834961
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,2,1,power_law_1.01,0.15166784286499024
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,2,1,power_law_1.01,0.15480128288269043
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,2,1,power_law_1.01,0.15827263832092284
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,2,1,power_law_1.01,0.16559360504150392
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,2,1,power_law_1.2,1.9605126953125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,2048,768,8,128,2,1,power_law_1.2,2.4538841247558594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,2,1,power_law_1.01,0.17719680786132813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,2,1,power_law_1.01,0.18219839096069337
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,2,1,power_law_1.01,0.2002348709106445
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,2,1,power_law_1.01,0.2347545623779297
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,2,1,power_law_1.01,0.2729984092712402
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,2,1,power_law_1.01,0.35236671447753903
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,2,1,power_law_1.01,0.6100019073486328
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,2,1,power_law_1.01,0.4440639877319336
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,2,1,power_law_1.01,0.7851219177246094
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,2,1,power_law_1.01,1.1503097534179687
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,2,1,power_law_1.01,1.5241900634765624
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,2,1,power_law_1.01,3.029558410644531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,1,power_law_1.2,0.08359935760498047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,2,1,power_law_1.01,2.2907621765136716
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,1,power_law_1.2,0.12332736015319826
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,1,power_law_1.2,0.23107839584350587
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,1,power_law_1.2,0.06457280158996583
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,1,power_law_1.2,0.35243839263916016
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,1,power_law_1.2,0.37009601593017577
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,1,power_law_1.2,0.38895038604736326
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,1,power_law_1.2,0.40364990234375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,1,power_law_1.2,0.4068511962890625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,1,power_law_1.2,0.44557121276855466
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,1,power_law_1.2,0.4333689498901367
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,1,power_law_1.2,0.4521350479125976
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,1,power_law_1.2,0.487831039428711
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,1,power_law_1.2,0.5110073471069336
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,1,power_law_1.2,0.5132511901855469
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,1,power_law_1.2,0.5318540954589844
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,1,power_law_1.2,0.6109958267211913
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,1,power_law_1.2,0.6695891571044922
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,1,power_law_1.2,0.8163980865478516
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,1536,8,128,2,1,power_law_1.01,3.8626867675781247
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,1,power_law_1.2,0.18202560424804687
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,1,power_law_1.2,0.970804443359375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,1,power_law_1.2,1.2822047424316405
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,1,power_law_1.2,1.7081056213378907
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,1,power_law_1.2,2.358203582763672
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,1,power_law_1.2,2.9546322631835937
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,1,power_law_1.2,0.045251197814941406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,1,power_law_1.2,0.06174784183502198
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,1,power_law_1.2,0.09488703727722167
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,1,power_law_1.2,0.1150393581390381
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,1,power_law_1.2,0.15281855583190918
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,1,power_law_1.2,0.22280384063720704
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,1,power_law_1.2,0.23271615982055663
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,1,power_law_1.2,0.23828672409057616
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,1,power_law_1.2,0.2437785530090332
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,1,power_law_1.2,0.2428883171081543
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,1,power_law_1.2,0.24681472778320312
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,1,power_law_1.2,0.28203008651733397
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,1,power_law_1.2,0.28505664825439453
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,1,power_law_1.2,0.2977491188049316
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,1,power_law_1.2,0.305219841003418
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,1,power_law_1.2,0.3082137680053711
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,1,power_law_1.2,0.3195590400695801
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,1,power_law_1.2,0.33863296508789065
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,1,power_law_1.2,0.36949249267578127
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,1,power_law_1.2,0.42980350494384767
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,1,power_law_1.2,0.4917440032958984
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,1,power_law_1.2,0.6097651290893554
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,1,power_law_1.2,0.733842544555664
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,1,power_law_1.2,4.307219238281251
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,1,power_law_1.2,0.9808057403564453
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,1,power_law_1.2,1.223623046875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,1,power_law_1.2,5.931038818359375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,1,power_law_1.2,1.7214784240722658
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,2,1,power_law_1.2,0.040285439491271974
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,2,1,power_law_1.2,0.05023615837097169
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,2,1,power_law_1.2,0.06222464084625244
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,2,1,power_law_1.2,0.07356544017791748
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,2,1,power_law_1.2,0.08706048011779785
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,2,1,power_law_1.2,0.1294054412841797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,2,1,power_law_1.2,0.13655296325683594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,2,1,power_law_1.2,0.1422860813140869
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,2,1,power_law_1.2,0.1475648021697998
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,2,1,power_law_1.2,0.14781311988830564
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,2,1,power_law_1.2,0.15355839729309081
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,2,1,power_law_1.2,0.16015743255615233
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,2,1,power_law_1.2,0.1639641571044922
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,2,1,power_law_1.2,0.17663871765136718
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,2,1,power_law_1.2,0.1886406326293945
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,2,1,power_law_1.2,0.19750463485717773
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,1,power_law_1.2,2.1952294921875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,2,1,power_law_1.2,0.22257728576660157
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,2,1,power_law_1.2,0.2720767974853516
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,2,1,power_law_1.2,0.32919166564941404
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,2,1,power_law_1.2,7.302539672851563
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,2,1,power_law_1.2,0.4494271850585937
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,2,1,power_law_1.2,0.5691020965576172
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,2,1,power_law_1.2,0.8317958068847655
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,2,1,power_law_1.2,2.716259765625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,2,1,power_law_1.2,1.093930206298828
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,2,1,power_law_1.2,1.6088677978515626
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,2,1,power_law_1.2,2.113093719482422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,1,power_law_1.01,0.10244159698486328
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,1,power_law_1.01,0.15754560470581053
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,1,power_law_1.01,0.3067257690429687
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,1,power_law_1.01,0.4546470260620117
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,1,power_law_1.01,0.6048582458496095
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,1,power_law_1.01,0.9507250976562499
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,1,power_law_1.01,0.987533416748047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,1,power_law_1.01,1.0243705749511718
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,1,power_law_1.01,1.0606848144531251
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,1,power_law_1.01,1.0851910400390625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,1,power_law_1.01,1.1644818878173828
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,1,power_law_1.01,1.2489612579345704
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,1,power_law_1.01,1.279839324951172
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,2,1,power_law_1.2,3.143634033203125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,1,power_law_1.01,1.3298822021484376
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,1,power_law_1.01,1.358046112060547
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,1,power_law_1.01,1.372270050048828
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,1,power_law_1.01,1.4987295532226563
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,1,power_law_1.01,1.7035647583007811
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,1,power_law_1.01,2.06669189453125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,1,power_law_1.01,2.3281210327148436
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,1,power_law_1.01,3.0345755004882813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,1,power_law_1.01,3.7198080444335937
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,1,power_law_1.01,4.752876892089843
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,1,power_law_1.01,6.404655151367187
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,1,power_law_1.01,8.894205322265625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,1,power_law_1.01,12.018374023437499
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,2,1,power_law_1.2,4.167337646484375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,1,power_law_1.01,16.1845703125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,1,power_law_1.01,0.11348480224609374
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,1,power_law_1.01,0.07513343811035156
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,1,power_law_1.01,0.17606208801269532
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,1,power_law_1.01,0.34218048095703124
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,1,power_law_1.01,0.24921215057373045
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,1,power_law_1.01,0.5250495910644531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,1,power_law_1.01,0.5426240158081055
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,1,power_law_1.01,0.561602554321289
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,1,power_law_1.01,0.5777011108398438
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,1,power_law_1.01,0.5866444778442383
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,1,power_law_1.01,0.5998976135253906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,1,power_law_1.01,0.6161126327514649
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,1,power_law_1.01,0.6722860717773438
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,1,power_law_1.01,0.7062079620361328
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,1,power_law_1.01,0.6997894287109375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,1,power_law_1.01,0.7067161560058594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,1,power_law_1.01,0.729345932006836
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,2,1,power_law_1.01,19.137528076171876
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,1,power_law_1.01,0.784936294555664
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,1,power_law_1.01,0.8237510681152344
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,1,power_law_1.01,0.9761926269531249
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,1,power_law_1.01,1.1719789123535156
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,1,power_law_1.01,1.3934739685058593
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,1,power_law_1.01,1.684800567626953
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,1,power_law_1.01,2.3543052673339844
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,1,power_law_1.01,2.748929138183594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,1,power_law_1.01,3.974217529296875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,1,power_law_1.01,5.270722045898437
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,2,1,power_law_1.01,6.029752807617188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,1536,8,128,2,1,power_law_1.2,5.1997119140625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,2,1,power_law_1.01,0.0795091199874878
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,2,1,power_law_1.01,0.19449472427368164
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,2,1,power_law_1.01,0.10680831909179686
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,2,1,power_law_1.01,0.30313343048095703
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,2,1,power_law_1.01,0.3194278335571289
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,2,1,power_law_1.01,0.3328467178344726
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,2,1,power_law_1.01,0.34417152404785156
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,2,1,power_law_1.01,0.3487955093383789
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,2,1,power_law_1.01,0.06393983840942383
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,2,1,power_law_1.01,0.36093055725097656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,2,1,power_law_1.01,0.37176128387451174
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,2,1,power_law_1.01,0.3971891021728516
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,2,1,power_law_1.01,0.3760313415527344
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,2,1,power_law_1.01,0.4063315200805664
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,2,1,power_law_1.01,0.4109945678710938
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,2,1,power_law_1.01,0.15067392349243164
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,2,1,power_law_1.01,0.5064249420166016
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,2,1,power_law_1.01,0.5585356903076172
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,2,1,power_law_1.01,0.6841305541992188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,2,1,power_law_1.01,0.4446380615234375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,2,1,power_law_1.01,0.8306028747558594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,2,1,power_law_1.01,1.0995948791503907
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,2,1,power_law_1.01,1.3856294250488281
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,2,1,power_law_1.01,1.9918170166015625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,2,1,power_law_1.01,2.586134948730469
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,1,power_law_1.2,0.10043904304504395
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,1,power_law_1.2,0.15604607582092284
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,1,power_law_1.2,0.2912326431274414
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,1,power_law_1.2,0.4041759872436524
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,2,1,power_law_1.01,3.8137588500976562
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,1,power_law_1.2,0.5440812683105468
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,1,power_law_1.2,0.9192582702636718
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,1,power_law_1.2,0.9698790740966796
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,1,power_law_1.2,1.0268326568603516
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,1,power_law_1.2,1.0844403076171876
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,1,power_law_1.2,1.1229254150390626
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,1,power_law_1.2,1.2241990661621094
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,1,power_law_1.2,1.2718739318847656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,1,power_law_1.2,1.3296946716308593
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,1,power_law_1.2,1.3373881530761718
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,1,power_law_1.2,1.4208428955078125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,2,1,power_law_1.01,5.1171942138671875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,1,power_law_1.2,1.4478054809570313
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,1,power_law_1.2,1.529303741455078
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,1,power_law_1.2,1.6891162109375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,1,power_law_1.2,1.9213491821289062
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,1,power_law_1.2,2.5701370239257812
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,1,power_law_1.2,2.874034423828125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,1,power_law_1.2,3.663919372558594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,1,power_law_1.2,4.785503234863281
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,1,power_law_1.2,6.152767333984375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,1,power_law_1.2,8.025111083984374
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,2560,8,160,2,1,power_law_1.01,6.414475097656251
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,1,power_law_1.2,0.07181248188018799
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,1,power_law_1.2,0.1145952033996582
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,1,power_law_1.2,0.16821760177612305
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,1,power_law_1.2,0.234451847076416
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,1,power_law_1.2,0.3231500625610352
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,1,power_law_1.2,0.5180467224121094
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,1,power_law_1.2,0.5389683151245117
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,1,power_law_1.2,12.388859863281251
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,1,power_law_1.2,0.5656505584716797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,1,power_law_1.2,0.5855596923828126
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,1,power_law_1.2,0.5987308883666993
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,1,power_law_1.2,0.6137228775024414
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,1,power_law_1.2,0.621085433959961
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,1,power_law_1.2,0.7011475372314453
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,1,power_law_1.2,0.7077868652343751
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,1,power_law_1.2,0.7241011047363282
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,1,power_law_1.2,0.733846435546875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,1,power_law_1.2,0.7613625335693359
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,1,power_law_1.2,0.8053984069824219
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,1,power_law_1.2,0.8473664093017579
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,1,power_law_1.2,0.991327362060547
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,1,power_law_1.2,1.1298643493652345
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,1,power_law_1.2,1.4039244079589843
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,1,power_law_1.2,1.7323341369628906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,1,power_law_1.2,2.266367950439453
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,1,power_law_1.2,2.8568798828125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,1,power_law_1.2,15.037630615234374
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,1,power_law_1.2,3.9585101318359377
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,2,1,power_law_1.2,0.06367424011230469
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,2,1,power_law_1.2,0.07692607879638672
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,2,1,power_law_1.2,0.10478848457336425
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,2,1,power_law_1.2,0.13795519828796388
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,2,1,power_law_1.2,0.18189247131347658
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,2,1,power_law_1.2,18.400821533203125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,2,1,power_law_1.2,0.2982758331298828
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,2,1,power_law_1.2,0.3172934341430664
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,2,1,power_law_1.2,0.350118408203125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,2,1,power_law_1.2,0.33811584472656253
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,1,power_law_1.2,5.246708984375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,2,1,power_law_1.2,0.35963520050048825
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,2,1,power_law_1.2,0.37518463134765623
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,2,1,power_law_1.2,0.38117889404296873
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,2,1,power_law_1.2,0.3966732788085937
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,2,1,power_law_1.2,0.4095814514160156
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,2,1,power_law_1.2,5.891022949218749
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,2,1,power_law_1.2,0.5655807876586915
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,2,1,power_law_1.2,0.4322150421142578
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,2,1,power_law_1.2,0.4498368072509765
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,2,1,power_law_1.2,0.4902438354492188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,2,1,power_law_1.2,0.6541868591308594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,2,1,power_law_1.2,0.8501062774658203
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,2,1,power_law_1.2,1.0605158233642578
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,2,1,power_law_1.2,1.975810546875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,2,1,power_law_1.2,1.5138957214355468
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,2,1,power_law_1.2,2.9733377075195313
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,2,1,power_law_1.2,3.9944647216796874
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,1,power_law_1.01,0.10316415786743165
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,1,power_law_1.01,0.14434880256652832
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,1,power_law_1.01,0.28784128189086916
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,1,power_law_1.01,0.5040460968017578
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,1,power_law_1.01,0.8555827331542968
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,1,power_law_1.01,1.1727852630615234
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,1,power_law_1.01,1.352058868408203
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,1,power_law_1.01,1.7440045166015625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,1,power_law_1.01,1.8362028503417966
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,1,power_law_1.01,1.911637725830078
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,1,power_law_1.01,1.990784606933594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,1,power_law_1.01,2.0134495544433593
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,2,1,power_law_1.2,6.0296563720703125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,1,power_law_1.01,2.1516627502441406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,1,power_law_1.01,2.24483642578125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,1,power_law_1.01,2.4839628601074217
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,1,power_law_1.01,2.6080838012695313
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,1,power_law_1.01,2.539134674072266
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,1,power_law_1.01,2.72902587890625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,1,power_law_1.01,3.0373959350585937
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,1,power_law_1.01,3.5030810546875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,1,power_law_1.01,4.9388146972656255
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,1,power_law_1.01,5.6483544921875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,1,power_law_1.01,7.796299438476562
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,1,power_law_1.01,9.34534423828125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,2,1,power_law_1.2,8.115245971679688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,1,power_law_1.01,12.917625732421874
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,1,power_law_1.01,0.07353024005889892
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,1,power_law_1.01,0.0983993625640869
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,1,power_law_1.01,0.16034303665161134
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,1,power_law_1.01,0.28106496810913084
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,1,power_law_1.01,0.47164352416992184
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,1,power_law_1.01,0.6637785339355469
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,1,power_law_1.01,0.7880115509033203
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,1,power_law_1.01,15.925679931640625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,1,power_law_1.01,0.9750329589843749
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,1,power_law_1.01,1.0215897369384765
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,1,power_law_1.01,1.0595417785644532
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,2,1,power_law_1.01,19.5945556640625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,1,power_law_1.01,1.0933625793457031
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,1,power_law_1.01,1.1095136260986327
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,1,power_law_1.01,1.1516006469726563
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,1,power_law_1.01,1.1762675476074218
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,1,power_law_1.01,1.1894719696044922
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,1,power_law_1.01,1.2581632232666016
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,1,power_law_1.01,1.410576629638672
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,1,power_law_1.01,1.4104115295410158
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,1,power_law_1.01,1.5322335815429686
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,1,power_law_1.01,1.6345362854003906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,1,power_law_1.01,1.7605375671386718
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,1,power_law_1.01,2.108293762207031
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,1,power_law_1.01,2.3918829345703125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,1,power_law_1.01,2.941931457519531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,1,power_law_1.01,3.506780090332031
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,1,power_law_1.01,4.621402282714844
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,1,power_law_1.01,5.5463671875000005
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,2,1,power_law_1.01,6.83065185546875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,2,1,power_law_1.01,0.08752320289611817
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,2,1,power_law_1.01,0.09912320137023925
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,2,1,power_law_1.01,0.0871558380126953
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,2,1,power_law_1.01,0.11131135940551758
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,2,1,power_law_1.01,0.1238700771331787
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,2,1,power_law_1.01,0.1365017604827881
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,2,1,power_law_1.01,0.149237117767334
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,2,1,power_law_1.01,0.1612761688232422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,2,1,power_law_1.01,0.18686080932617188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,2,1,power_law_1.01,0.08820608139038086
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,2,1,power_law_1.01,0.09054143905639649
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,2,1,power_law_1.01,0.2242259216308594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,2,1,power_law_1.01,0.248089599609375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,2,1,power_law_1.01,0.29940864562988284
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,2,1,power_law_1.01,0.3604902267456055
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,2,1,power_law_1.01,0.5383039855957031
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,2,1,power_law_1.01,0.4143731307983399
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,2,1,power_law_1.01,1.0205023956298827
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,2,1,power_law_1.01,0.7879283142089843
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,2,1,power_law_1.01,1.4917330932617188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,2,1,power_law_1.01,1.9607026672363284
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,2,1,power_law_1.01,2.8994784545898438
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,2,1,power_law_1.01,5.715680541992187
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,2,1,power_law_1.01,3.8447921752929686
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,1,power_law_1.2,0.10452735900878907
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,1,power_law_1.2,0.14578047752380371
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,1,power_law_1.2,0.2872345542907715
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,1,power_law_1.2,0.49954814910888673
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,1,power_law_1.2,0.7691219329833985
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,1,power_law_1.2,1.2506265258789062
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,1,power_law_1.2,1.706636199951172
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,1,power_law_1.2,1.7795347595214843
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,1,power_law_1.2,1.84151611328125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,1,power_law_1.2,1.9262899780273437
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,1,power_law_1.2,2.0291539001464844
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,1,power_law_1.2,2.15185791015625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,1,power_law_1.2,2.3347628784179686
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,1,power_law_1.2,2.52927490234375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,1,power_law_1.2,2.7260064697265625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,1,power_law_1.2,2.594304504394531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,1,power_law_1.2,2.87782470703125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,1,power_law_1.2,3.1449484252929687
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,1,power_law_1.2,3.6419604492187503
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,1,power_law_1.2,4.231164245605468
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,1,power_law_1.2,4.921906433105469
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,1,power_law_1.2,5.946248168945313
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,1,power_law_1.2,7.907443847656251
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,1,power_law_1.2,9.77381103515625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,1,power_law_1.2,12.773297119140626
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,1,power_law_1.2,0.07321216106414795
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,1,power_law_1.2,0.16191360473632813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,1,power_law_1.2,16.889360351562498
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,1,power_law_1.2,0.28401472091674806
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,2,1,power_law_1.2,19.970369873046874
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,1,power_law_1.2,0.6176595306396484
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,1,power_law_1.2,0.7368313598632812
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,1,power_law_1.2,0.9551238250732421
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,1,power_law_1.2,0.9982803344726563
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,1,power_law_1.2,1.02391357421875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,1,power_law_1.2,1.0667552185058593
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,1,power_law_1.2,0.42903488159179687
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,1,power_law_1.2,1.09989501953125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,1,power_law_1.2,1.1386630249023437
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,1,power_law_1.2,0.09402239799499512
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,1,power_law_1.2,1.167952651977539
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,1,power_law_1.2,1.1876863861083984
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,1,power_law_1.2,1.223957748413086
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,1,power_law_1.2,1.343206329345703
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,1,power_law_1.2,1.483068084716797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,1,power_law_1.2,1.5304287719726564
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,1,power_law_1.2,1.722034606933594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,1,power_law_1.2,1.8593043518066406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,1,power_law_1.2,2.30238525390625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,1,power_law_1.2,2.4769856262207033
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,1,power_law_1.2,3.117760009765625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,1,power_law_1.2,3.5958489990234375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,1,power_law_1.2,4.898695678710938
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,2,1,power_law_1.2,0.08787327766418457
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,1,power_law_1.2,5.7404632568359375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,2,1,power_law_1.2,6.822793579101562
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,2,1,power_law_1.2,0.08974528312683105
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,2,1,power_law_1.2,0.11144703865051268
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,2,1,power_law_1.2,0.12436991691589357
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,2,1,power_law_1.2,0.13744192123413085
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,2,1,power_law_1.2,0.1492243194580078
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,2,1,power_law_1.2,0.16200960159301758
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,2,1,power_law_1.2,0.18705215454101562
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,2,1,power_law_1.2,0.22410240173339843
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,2,1,power_law_1.2,0.24900735855102538
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,2,1,power_law_1.2,0.08606719970703125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,2,1,power_law_1.2,0.09221183776855468
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,2,1,power_law_1.2,0.29957759857177735
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,2,1,power_law_1.2,0.09802816390991212
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,2,1,power_law_1.2,0.3620038223266602
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,2,1,power_law_1.2,0.41625408172607425
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,2,1,power_law_1.2,0.78487548828125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,2,1,power_law_1.2,0.5441619110107422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,2,1,power_law_1.2,1.0200141143798827
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,2,1,power_law_1.2,1.4941529846191406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,2,1,power_law_1.2,1.9595571899414062
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,2,1,power_law_1.2,2.9014291381835937
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,2,1,power_law_1.2,3.8369491577148436
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,2,1,power_law_1.2,5.7192315673828125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,1,power_law_1.01,0.04459072113037109
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,1,power_law_1.01,0.08849535942077637
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,1,power_law_1.01,0.1496121597290039
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,1,power_law_1.01,0.17403455734252932
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,1,power_law_1.01,0.26098175048828126
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,1,power_law_1.01,0.28026752471923827
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,1,power_law_1.01,0.29656703948974605
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,1,power_law_1.01,0.30433088302612304
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,1,power_law_1.01,0.3082073593139648
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,1,power_law_1.01,0.3238566589355469
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,1,power_law_1.01,0.35620670318603515
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,1,power_law_1.01,0.36990463256835937
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,1,power_law_1.01,0.37987838745117186
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,1,power_law_1.01,0.37988159179687503
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,1,power_law_1.01,0.3829811096191406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,1,power_law_1.01,0.3983059310913086
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,1,power_law_1.01,0.44484607696533207
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,1,power_law_1.01,0.48161151885986325
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,1,power_law_1.01,0.5611916732788086
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,1,power_law_1.01,0.653196792602539
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,1,power_law_1.01,0.8381938934326172
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,1,power_law_1.01,1.0442246246337892
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,1,power_law_1.01,1.4441011047363281
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,1,power_law_1.01,1.846961212158203
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,1,power_law_1.01,2.7063006591796874
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,1,power_law_1.01,3.513960266113281
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,2688,1856,6,128,2,1,power_law_1.01,4.286408386230469
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,2,1,power_law_1.01,0.03944384098052979
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,2,1,power_law_1.01,0.05538432121276855
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,2,1,power_law_1.01,0.0688704013824463
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,2,1,power_law_1.01,0.08176192283630371
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,2,1,power_law_1.01,0.10884544372558594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,2,1,power_law_1.01,0.11319616317749023
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,2,1,power_law_1.01,0.1183948802947998
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,2,1,power_law_1.01,0.043526401519775396
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,2,1,power_law_1.01,0.1217305564880371
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,2,1,power_law_1.01,0.1206886386871338
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,2,1,power_law_1.01,0.12734784126281737
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,2,1,power_law_1.01,0.12851519584655763
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,2,1,power_law_1.01,0.13382975578308104
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,2,1,power_law_1.01,0.13705280303955078
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,2,1,power_law_1.01,0.17856639862060547
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,2,1,power_law_1.01,0.158024320602417
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,2,1,power_law_1.01,0.14282560348510742
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,2,1,power_law_1.01,0.13992832183837892
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,2,1,power_law_1.01,0.2014041519165039
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,2,1,power_law_1.01,0.25383039474487307
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,2,1,power_law_1.01,0.3051705551147461
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,2,1,power_law_1.01,0.41192638397216796
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,2,1,power_law_1.01,0.5265113449096679
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,2,1,power_law_1.01,0.7622624206542968
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,2,1,power_law_1.01,0.9916960144042969
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,2,1,power_law_1.01,1.4488377380371094
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,1,power_law_1.2,0.04450240135192871
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,1,power_law_1.2,0.05943424224853515
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,1,power_law_1.2,0.08635392189025878
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,1,power_law_1.2,0.13244352340698243
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,1,power_law_1.2,0.16646015167236328
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,1,power_law_1.2,0.25987455368041995
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,1,power_law_1.2,0.298288631439209
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,1,power_law_1.2,0.3043379211425781
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,1,power_law_1.2,0.2805484771728516
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,1,power_law_1.2,0.31164159774780276
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,1,power_law_1.2,0.3311507034301758
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,2,1,power_law_1.01,1.9669247436523438
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,1,power_law_1.2,0.3721612930297852
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,1,power_law_1.2,0.38347583770751953
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,1,power_law_1.2,0.3845145416259766
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,1,power_law_1.2,0.37711040496826176
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,1,power_law_1.2,0.3917337417602539
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,1,power_law_1.2,0.41051326751708983
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,1,power_law_1.2,0.4517663955688477
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,1,power_law_1.2,0.4830201721191406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,1,power_law_1.2,0.5752710342407227
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,1,power_law_1.2,0.6651014709472657
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,1,power_law_1.2,0.8435692596435548
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,2688,1856,6,128,2,1,power_law_1.01,2.4747943115234374
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,1,power_law_1.2,1.0566246032714843
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,1,power_law_1.2,1.4530931091308594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,1,power_law_1.2,1.8295359802246094
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,2,1,power_law_1.2,0.039736959934234616
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,2,1,power_law_1.2,0.045935997962951655
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,2,1,power_law_1.2,0.054826240539550784
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,2,1,power_law_1.2,0.06429183959960938
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,1,power_law_1.2,2.6770944213867187
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,2,1,power_law_1.2,0.07789887905120849
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,2,1,power_law_1.2,0.10893759727478028
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,2,1,power_law_1.2,0.1132800006866455
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,2,1,power_law_1.2,0.12003775596618653
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,2,1,power_law_1.2,0.12583104133605957
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,2,1,power_law_1.2,0.1217632007598877
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,2,1,power_law_1.2,0.12904191970825196
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,2,1,power_law_1.2,0.13277248382568357
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,2,1,power_law_1.2,0.13852864265441894
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,2,1,power_law_1.2,0.14266624450683593
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,2,1,power_law_1.2,0.15151935577392578
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,1,power_law_1.2,3.5887597656250003
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,2688,1856,6,128,2,1,power_law_1.2,4.315734252929688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,2,1,power_law_1.2,0.15607935905456544
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,2,1,power_law_1.2,0.2040947151184082
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,2,1,power_law_1.2,0.1745631980895996
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,2,1,power_law_1.2,0.23898239135742189
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,2,1,power_law_1.2,0.31735103607177734
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,2,1,power_law_1.2,0.3959590530395508
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,2,1,power_law_1.2,0.5754662322998046
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,2,1,power_law_1.2,2.329351043701172
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,2,1,power_law_1.2,0.7581766510009766
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,2,1,power_law_1.2,1.128463363647461
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,2,1,power_law_1.2,1.5180140686035155
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,1,power_law_1.01,0.17791807174682617
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,1,power_law_1.01,0.28221887588500977
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,1,power_law_1.01,0.8343769836425782
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,1,power_law_1.01,0.499699821472168
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,1,power_law_1.01,1.157871322631836
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,1,power_law_1.01,1.8931468200683592
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,1,power_law_1.01,2.1059341430664062
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,1,power_law_1.01,2.1360301208496093
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,1,power_law_1.01,2.250469055175781
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,1,power_law_1.01,2.246163787841797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,1,power_law_1.01,2.6294879150390624
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,2,1,power_law_1.2,3.1499118041992187
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,1,power_law_1.01,2.623882141113281
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,1,power_law_1.01,2.9449313354492186
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,1,power_law_1.01,3.116596984863281
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,1,power_law_1.01,3.372630310058594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,1,power_law_1.01,3.8938745117187503
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,1,power_law_1.01,4.311305541992188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,1,power_law_1.01,3.697721557617187
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,2688,1856,6,128,2,1,power_law_1.2,3.927029113769531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,1,power_law_1.01,4.3571405029296875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,1,power_law_1.01,4.964512634277344
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,1,power_law_1.01,5.833016967773437
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,1,power_law_1.01,7.900702514648438
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,1,power_law_1.01,10.443582763671875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,1,power_law_1.01,13.210650634765624
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,1,power_law_1.01,16.3920361328125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,1,power_law_1.01,0.11194623947143553
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,1,power_law_1.01,0.17719680786132813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,1,power_law_1.01,0.296882553100586
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,1,power_law_1.01,23.3323486328125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,1,power_law_1.01,0.48270721435546876
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,1,power_law_1.01,0.6771193695068359
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,1,power_law_1.01,1.0895667266845703
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,1,power_law_1.01,31.56772705078125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,1,power_law_1.01,1.206917724609375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,1,power_law_1.01,1.156825637817383
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,1,power_law_1.01,1.265525131225586
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,1,power_law_1.01,1.2992684936523438
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,1,power_law_1.01,1.3331718444824219
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,1,power_law_1.01,1.265198059082031
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,1,power_law_1.01,1.3613926696777345
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,1,power_law_1.01,1.4108082580566408
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,1,power_law_1.01,1.4352056884765625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,1,power_law_1.01,1.4879353332519532
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,1,power_law_1.01,1.5433439636230468
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,2688,22,512,2,1,power_law_1.01,36.80606201171875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,1,power_law_1.01,1.8072416687011716
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,1,power_law_1.01,1.978021697998047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,1,power_law_1.01,2.2025715637207033
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,1,power_law_1.01,2.5736398315429687
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,1,power_law_1.01,3.224839172363281
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,1,power_law_1.01,3.867059326171875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,1,power_law_1.01,4.891677551269531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,1,power_law_1.01,6.168484497070312
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,1,power_law_1.01,9.585037231445312
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,1,power_law_1.01,11.56676025390625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,2688,22,512,2,1,power_law_1.01,13.685716552734373
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,2,1,power_law_1.01,0.12977984428405762
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,2,1,power_law_1.01,0.1289478397369385
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,2,1,power_law_1.01,0.13881279945373534
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,2,1,power_law_1.01,0.1499244785308838
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,2,1,power_law_1.01,0.12996928215026854
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,2,1,power_law_1.01,0.17294208526611327
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,2,1,power_law_1.01,0.16321088790893554
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,2,1,power_law_1.01,0.18466432571411134
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,2,1,power_law_1.01,0.19573823928833006
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,2,1,power_law_1.01,0.2202412796020508
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,2,1,power_law_1.01,0.13128640174865722
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,2,1,power_law_1.01,0.26570432662963867
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,2,1,power_law_1.01,0.2886553573608398
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,2,1,power_law_1.01,0.4040915298461914
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,2,1,power_law_1.01,0.33766334533691406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,2,1,power_law_1.01,0.4531833648681641
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,2,1,power_law_1.01,0.5688985443115234
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,2,1,power_law_1.01,0.8030528259277343
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,2,1,power_law_1.01,1.040062713623047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,2,1,power_law_1.01,1.5049906921386718
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,2,1,power_law_1.01,1.9674584960937502
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,2,1,power_law_1.01,2.8983502197265625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,2,1,power_law_1.01,3.83900146484375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,2,1,power_law_1.01,5.704102783203124
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,1,power_law_1.2,0.17891328811645507
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,1,power_law_1.2,0.2815628814697265
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,1,power_law_1.2,0.4975455856323242
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,2,1,power_law_1.01,7.531187133789063
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,1,power_law_1.2,0.7669951629638672
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,1,power_law_1.2,1.0406694030761718
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,1,power_law_1.2,1.8671725463867186
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,1,power_law_1.2,2.0794406127929688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,1,power_law_1.2,2.1289208984375003
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,1,power_law_1.2,2.391643524169922
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,1,power_law_1.2,2.4492396545410156
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,1,power_law_1.2,2.738335266113281
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,1,power_law_1.2,2.7751455688476563
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,1,power_law_1.2,3.1133938598632813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,1,power_law_1.2,3.31861572265625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,1,power_law_1.2,3.824801940917969
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,1,power_law_1.2,3.9068493652343745
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,1,power_law_1.2,4.6583929443359375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,1,power_law_1.2,3.9808230590820317
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,1,power_law_1.2,4.651967468261718
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,1,power_law_1.2,5.221141967773438
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,1,power_law_1.2,6.358984375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,1,power_law_1.2,8.08897216796875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,1,power_law_1.2,10.61138916015625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,1,power_law_1.2,13.490159912109373
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,1,power_law_1.2,17.104532470703127
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,1,power_law_1.2,22.824030761718753
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,1,power_law_1.2,0.10805824279785156
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,1,power_law_1.2,0.17210304260253906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,1,power_law_1.2,0.44309249877929685
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,1,power_law_1.2,0.2917676734924316
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,1,power_law_1.2,0.627116813659668
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,1,power_law_1.2,1.1532364654541016
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,1,power_law_1.2,1.067855987548828
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,1,power_law_1.2,1.1897357177734373
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,1,power_law_1.2,1.2418355560302734
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,1,power_law_1.2,1.2762892913818358
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,1,power_law_1.2,1.3215231323242187
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,1,power_law_1.2,1.3567674255371094
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,1,power_law_1.2,1.3921363830566407
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,1,power_law_1.2,1.4612608337402344
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,1,power_law_1.2,30.359418945312502
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,1,power_law_1.2,1.4919506835937502
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,1,power_law_1.2,1.5271046447753907
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,1,power_law_1.2,1.6737432861328123
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,1,power_law_1.2,1.8968800354003907
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,1,power_law_1.2,1.9536723327636718
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,1,power_law_1.2,2.4615577697753905
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,1,power_law_1.2,2.618016662597656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,1,power_law_1.2,3.2510650634765623
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,1,power_law_1.2,3.906102294921875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,1,power_law_1.2,4.938252258300781
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,1,power_law_1.2,6.06549072265625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,1,power_law_1.2,9.032651367187501
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,2,1,power_law_1.2,0.13033920288085937
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,2,1,power_law_1.2,0.12911744117736818
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,2,1,power_law_1.2,0.13013119697570802
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,2,1,power_law_1.2,0.13179903984069824
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,2688,22,512,2,1,power_law_1.2,38.32854736328125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,2,1,power_law_1.2,0.13872768402099608
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,2,1,power_law_1.2,0.14980287551879884
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,2,1,power_law_1.2,0.1621548843383789
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,1,power_law_1.2,11.36712646484375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,2,1,power_law_1.2,0.1729011154174805
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,2,1,power_law_1.2,0.186060791015625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,2,1,power_law_1.2,0.2195136070251465
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,2,1,power_law_1.2,0.2662528038024902
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,2,1,power_law_1.2,0.289542407989502
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,2,1,power_law_1.2,0.3356076812744141
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,2,1,power_law_1.2,0.45275520324707036
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,2,1,power_law_1.2,0.19693248748779296
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,2,1,power_law_1.2,0.40394752502441406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,2,1,power_law_1.2,0.5705484771728516
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,2,1,power_law_1.2,1.5066310119628905
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,2688,22,512,2,1,power_law_1.2,13.72572021484375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,2,1,power_law_1.2,0.8039282989501952
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,2,1,power_law_1.2,1.0385734558105468
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,2,1,power_law_1.2,1.9755308532714841
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,2,1,power_law_1.2,2.908328857421875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,2,1,power_law_1.2,3.8408538818359377
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,2,1,power_law_1.2,5.691183471679688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,2,1,power_law_1.2,7.560986328125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,4,1,power_law_1.01,0.10869312286376953
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,4,1,power_law_1.01,0.10678655624389648
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,4,1,power_law_1.01,0.1443129634857178
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,4,1,power_law_1.01,0.175228157043457
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,4,1,power_law_1.01,0.17680320739746094
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,4,1,power_law_1.01,0.17848575592041016
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,4,1,power_law_1.01,0.17736127853393555
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,4,1,power_law_1.01,0.17845184326171876
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,4,1,power_law_1.01,0.18910783767700196
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,4,1,power_law_1.01,0.22559104919433595
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,4,1,power_law_1.01,0.22637567520141602
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,4,1,power_law_1.01,0.2372768020629883
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,4,1,power_law_1.01,0.24934335708618166
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,4,1,power_law_1.01,0.25775167465209964
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,4,1,power_law_1.01,0.2964153671264648
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,4,1,power_law_1.01,0.345838737487793
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,4,1,power_law_1.01,0.42351936340332036
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,4,1,power_law_1.01,0.5027532958984375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,4,1,power_law_1.01,0.6728256225585938
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,4,1,power_law_1.01,0.8445101165771485
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,4,1,power_law_1.01,1.1993587493896485
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,4,1,power_law_1.01,1.5476069641113281
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,4,1,power_law_1.01,2.3264729309082033
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,4,1,power_law_1.01,3.1367111206054688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,4,1,power_law_1.01,4.604457702636719
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,4,1,power_law_1.01,0.048309760093688967
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,4,1,power_law_1.01,0.060015358924865723
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,4,1,power_law_1.01,0.06860928058624267
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,4,1,power_law_1.01,0.09244159698486329
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,4,1,power_law_1.01,0.09690112113952637
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,4,1,power_law_1.01,0.09841343879699707
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,4,1,power_law_1.01,0.09881152153015137
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,4,1,power_law_1.01,0.10128000259399414
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,4,1,power_law_1.01,0.10144384384155274
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,4,1,power_law_1.01,0.10357503890991211
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,4,1,power_law_1.01,0.10439935684204102
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,4,1,power_law_1.01,0.10512448310852052
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,4,1,power_law_1.01,0.10691583633422852
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,4,1,power_law_1.01,0.11882431983947754
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,4,1,power_law_1.01,0.12544896125793456
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,4,1,power_law_1.01,0.13513216018676757
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,4,1,power_law_1.01,0.13708800315856934
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,4,1,power_law_1.01,0.16138879776000975
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,4,1,power_law_1.01,0.1896486473083496
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,4,1,power_law_1.01,0.2435193634033203
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,4,1,power_law_1.01,6.233034057617187
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,4,1,power_law_1.01,0.2814988708496094
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,4,1,power_law_1.01,0.3919116973876953
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,4,1,power_law_1.01,0.4950143814086914
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,4,1,power_law_1.01,0.7044268798828125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,4,1,power_law_1.01,0.9178374481201171
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,4,1,power_law_1.01,8.020130004882812
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,4,1,power_law_1.01,1.405749053955078
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,4,1,power_law_1.01,1.8280070495605467
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,4,1,power_law_1.01,0.024714241027832033
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,4,1,power_law_1.01,0.04940288066864014
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,4,1,power_law_1.01,0.050078082084655764
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,4,1,power_law_1.01,0.05188223838806152
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,4,1,power_law_1.01,0.05188799858093261
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,4,1,power_law_1.01,0.05177599906921386
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,4,1,power_law_1.01,0.05159999847412109
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,4,1,power_law_1.01,2.397646026611328
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,4,1,power_law_1.01,0.05224575996398926
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,4,1,power_law_1.01,0.035439999103546144
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,4,1,power_law_1.01,0.03968447923660278
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,4,1,power_law_1.01,0.05342400074005127
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,4,1,power_law_1.01,0.0513702392578125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,4,1,power_law_1.01,0.055602560043334956
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,4,1,power_law_1.01,0.06043007850646973
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,4,1,power_law_1.01,0.062707200050354
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,4,1,power_law_1.01,0.06758336067199708
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,4,1,power_law_1.01,0.07298111915588379
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,4,1,power_law_1.01,0.08862655639648438
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,4,1,power_law_1.01,0.10547391891479492
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,4,1,power_law_1.01,0.1381638431549072
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,4,1,power_law_1.01,0.17138048171997072
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,4,1,power_law_1.01,0.23853567123413083
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,4,1,power_law_1.01,0.30745983123779297
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,4,1,power_law_1.01,0.4482124710083008
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,4,1,power_law_1.01,0.5881049728393555
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,4,1,power_law_1.2,0.07977407932281494
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,4,1,power_law_1.2,0.09522111892700195
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,4,1,power_law_1.2,0.10203328132629394
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,4,1,power_law_1.2,0.14552639961242675
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,4,1,power_law_1.2,0.1726675224304199
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,4,1,power_law_1.2,0.17708223342895507
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,4,1,power_law_1.2,0.17912704467773438
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,4,1,power_law_1.2,0.1774700736999512
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,4,1,power_law_1.2,0.18060287475585937
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,4,1,power_law_1.01,0.8626316833496095
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,4,1,power_law_1.2,0.1869215965270996
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,4,1,power_law_1.2,0.22442367553710935
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,4,1,power_law_1.2,0.22953151702880858
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,4,1,power_law_1.2,0.2335161590576172
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,4,1,power_law_1.2,0.24926271438598632
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,4,1,power_law_1.2,0.25834623336791995
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,4,1,power_law_1.2,0.3093312072753906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,4,1,power_law_1.2,0.34459457397460935
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,4,1,power_law_1.2,0.42170177459716796
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,4,1,power_law_1.2,0.5017081451416016
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,4,1,power_law_1.2,0.6764691162109375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,4,1,power_law_1.2,0.8418029022216796
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,4,1,power_law_1.2,1.1883757019042969
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,4,1,power_law_1.2,1.5266873168945314
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,4,1,power_law_1.01,1.149993591308594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,4,1,power_law_1.01,1.4313612365722657
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,4,1,power_law_1.2,2.3565132141113283
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,4,1,power_law_1.2,3.0981350708007813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,4,1,power_law_1.2,0.05559103965759278
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,4,1,power_law_1.2,0.06392127990722656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,4,1,power_law_1.2,0.0664524793624878
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,4,1,power_law_1.2,0.0923583984375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,4,1,power_law_1.2,0.0958016014099121
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,4,1,power_law_1.2,0.09902912139892578
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,4,1,power_law_1.2,0.09973440170288086
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,4,1,power_law_1.2,0.10142784118652344
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,4,1,power_law_1.2,0.10105088233947754
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,4,1,power_law_1.2,0.10318464279174804
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,4,1,power_law_1.2,0.10315711975097656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,4,1,power_law_1.2,0.10595264434814453
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,4,1,power_law_1.2,0.10565247535705566
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,4,1,power_law_1.2,0.11887040138244628
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,4,1,power_law_1.2,0.12896832466125488
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,4,1,power_law_1.2,0.14059328079223632
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,4,1,power_law_1.2,0.13730112075805664
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,4,1,power_law_1.2,0.15915583610534667
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,4,1,power_law_1.2,0.18971263885498046
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,4,1,power_law_1.2,0.23423999786376953
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,4,1,power_law_1.2,0.28239616394042966
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,4,1,power_law_1.2,0.39148609161376957
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,4,1,power_law_1.2,4.798866577148438
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,4,1,power_law_1.2,0.49437057495117187
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,4,1,power_law_1.2,6.229376220703125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,4,1,power_law_1.2,0.7015251159667969
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,4,1,power_law_1.2,0.8972422027587891
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,4,1,power_law_1.2,8.159140625000001
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,4,1,power_law_1.2,0.025543038845062253
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,4,1,power_law_1.2,1.3944921875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,4,1,power_law_1.2,0.034067840576171876
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,4,1,power_law_1.2,0.036507520675659176
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,4,1,power_law_1.2,0.051330561637878414
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,4,1,power_law_1.2,0.05104320049285889
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,4,1,power_law_1.2,0.05220287799835205
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,4,1,power_law_1.2,0.052598400115966795
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,4,1,power_law_1.2,0.051925759315490726
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,4,1,power_law_1.2,0.05241919994354248
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,4,1,power_law_1.2,0.05343808174133301
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,4,1,power_law_1.2,0.05234816074371338
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,4,1,power_law_1.2,0.05593728065490723
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,4,1,power_law_1.2,0.06078144073486328
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,4,1,power_law_1.2,0.05395328044891358
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,4,1,power_law_1.2,0.06984960079193116
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,4,1,power_law_1.2,0.06480447769165039
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,4,1,power_law_1.2,1.915879669189453
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,4,1,power_law_1.2,0.10749631881713868
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,4,1,power_law_1.2,0.07375743865966797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,4,1,power_law_1.2,2.269137268066406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,4,1,power_law_1.2,0.09201536178588868
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,4,1,power_law_1.2,0.2478713607788086
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,4,1,power_law_1.2,0.13977408409118652
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,4,1,power_law_1.2,0.17681535720825195
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,4,1,power_law_1.2,0.31971263885498047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,4,1,power_law_1.2,0.46175361633300777
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,4,1,power_law_1.2,0.6023360061645507
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,4,1,power_law_1.01,0.1042137622833252
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,4,1,power_law_1.01,0.13091391563415528
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,4,1,power_law_1.01,0.15532352447509765
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,4,1,power_law_1.01,0.23930559158325196
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,4,1,power_law_1.01,0.2470911979675293
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,4,1,power_law_1.01,0.2504537582397461
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,4,1,power_law_1.01,0.25253120422363284
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,4,1,power_law_1.01,0.24949695587158205
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,4,1,power_law_1.01,0.2549267196655273
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,4,1,power_law_1.01,0.27758272171020504
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,4,1,power_law_1.01,0.34299774169921876
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,4,1,power_law_1.01,0.36429630279541014
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,4,1,power_law_1.01,0.3820851135253906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,4,1,power_law_1.01,0.3977139282226562
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,4,1,power_law_1.01,0.41626430511474605
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,4,1,power_law_1.01,0.4564614486694336
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,4,1,power_law_1.01,0.5271072006225586
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,4,1,power_law_1.01,0.6702502441406251
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,4,1,power_law_1.01,0.8112159729003906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,4,1,power_law_1.01,1.1193593597412108
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,4,1,power_law_1.01,1.4870291137695313
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,4,1,power_law_1.01,2.1928839111328124
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,4,1,power_law_1.2,0.8933971405029297
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,4,1,power_law_1.01,2.91502197265625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,4,1,power_law_1.01,4.145109863281251
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,4,1,power_law_1.01,5.703246459960938
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,4,1,power_law_1.01,8.746237182617188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,4,1,power_law_1.2,1.1862278747558594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,4,1,power_law_1.01,0.06484416007995605
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,4,1,power_law_1.01,0.08095616340637207
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,4,1,power_law_1.01,0.1022969627380371
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,4,1,power_law_1.01,0.137457275390625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,4,1,power_law_1.01,0.1470201587677002
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,4,1,power_law_1.01,0.14768128395080565
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,4,1,power_law_1.01,0.14925888061523437
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,4,1,power_law_1.01,0.14855551719665527
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,4,1,power_law_1.01,0.15004672050476073
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,4,1,power_law_1.01,0.1518899154663086
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,4,1,power_law_1.01,0.154071683883667
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,4,1,power_law_1.01,0.15403327941894532
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,4,1,power_law_1.01,0.15844672203063964
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,4,1,power_law_1.01,0.17179391860961915
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,4,1,power_law_1.01,0.17784704208374022
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,4,1,power_law_1.01,0.1801420783996582
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,4,1,power_law_1.01,0.19389888763427734
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,4,1,power_law_1.01,0.23102144241333006
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,4,1,power_law_1.01,0.27627519607543943
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,4,1,power_law_1.01,0.3555327987670899
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,4,1,power_law_1.01,0.4201087951660156
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,4,1,power_law_1.01,0.5873465728759766
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,4,1,power_law_1.2,1.4570918273925781
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,4,1,power_law_1.01,12.163946533203125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,4,1,power_law_1.01,0.7649983978271484
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,4,1,power_law_1.01,1.1389209747314453
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,4,1,power_law_1.01,1.6229338073730468
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,4,1,power_law_1.01,14.85257568359375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,4,1,power_law_1.01,2.354701385498047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,4,1,power_law_1.01,3.075538024902344
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,4,1,power_law_1.01,0.031685121059417724
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,4,1,power_law_1.01,0.07636032104492188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,4,1,power_law_1.01,0.07802175998687744
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,4,1,power_law_1.01,0.07839488029479981
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,4,1,power_law_1.01,0.07891456127166749
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,4,1,power_law_1.01,0.07708096027374267
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,4,1,power_law_1.01,0.07767807960510253
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,4,1,power_law_1.01,0.07919936180114746
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,4,1,power_law_1.01,0.08019583702087403
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,4,1,power_law_1.01,4.1683502197265625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,4,1,power_law_1.01,0.04815807819366455
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,4,1,power_law_1.01,0.05610496044158936
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,4,1,power_law_1.01,0.07964352130889893
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,4,1,power_law_1.01,0.08147071838378907
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,4,1,power_law_1.01,0.090250244140625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,4,1,power_law_1.01,0.0880288028717041
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,4,1,power_law_1.01,0.09419520378112793
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,4,1,power_law_1.01,0.10338111877441407
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,4,1,power_law_1.01,0.12934271812438963
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,4,1,power_law_1.01,0.15380800247192383
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,4,1,power_law_1.01,0.20546239852905274
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,4,1,power_law_1.01,0.2523468780517578
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,4,1,power_law_1.01,0.354516487121582
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,4,1,power_law_1.01,0.6666432189941406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,4,1,power_law_1.01,0.4618630218505859
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,4,1,power_law_1.01,0.8699775695800781
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,4,1,power_law_1.2,0.10393600463867188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,4,1,power_law_1.2,0.12821503639221193
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,4,1,power_law_1.2,0.14077631950378416
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,4,1,power_law_1.2,0.23987264633178712
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,4,1,power_law_1.2,0.2454412841796875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,4,1,power_law_1.2,0.25648000717163083
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,4,1,power_law_1.2,0.25931200027465817
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,4,1,power_law_1.2,0.25478912353515626
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,4,1,power_law_1.2,0.26447999954223633
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,4,1,power_law_1.2,0.2868588829040527
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,4,1,power_law_1.2,0.34954113006591797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,4,1,power_law_1.2,0.37478271484375003
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,4,1,power_law_1.2,0.37889919281005857
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,4,1,power_law_1.2,0.3963814544677734
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,4,1,power_law_1.2,0.42192768096923833
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,4,1,power_law_1.2,0.4689676666259766
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,4,1,power_law_1.01,1.3303295898437502
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,4,1,power_law_1.2,0.5225971221923829
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,4,1,power_law_1.2,0.6694464111328126
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,4,1,power_law_1.2,0.814632339477539
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,4,1,power_law_1.2,1.0982649230957031
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,4,1,power_law_1.2,1.4337799072265625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,4,1,power_law_1.2,2.1068858337402343
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,4,1,power_law_1.01,1.7675225830078127
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,4,1,power_law_1.2,2.8633950805664066
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,4,1,power_law_1.2,4.051026611328124
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,4,1,power_law_1.2,5.4988671875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,4,1,power_law_1.01,2.2206752014160154
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,4,1,power_law_1.2,0.0646950387954712
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,4,1,power_law_1.2,0.07974143981933593
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,4,1,power_law_1.2,0.08749247550964355
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,4,1,power_law_1.2,0.13798975944519043
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,4,1,power_law_1.2,0.1467046356201172
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,4,1,power_law_1.2,0.14987648010253907
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,4,1,power_law_1.2,0.15101951599121094
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,4,1,power_law_1.2,0.15097920417785643
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,4,1,power_law_1.2,0.15294015884399415
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,4,1,power_law_1.2,0.1520627212524414
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,4,1,power_law_1.2,0.15366080284118652
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,4,1,power_law_1.2,0.15700736045837402
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,4,1,power_law_1.2,0.15639552116394043
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,4,1,power_law_1.2,0.17307584762573242
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,4,1,power_law_1.2,0.18306304931640624
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,4,1,power_law_1.2,0.18583423614501954
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,4,1,power_law_1.2,0.20012735366821288
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,4,1,power_law_1.2,0.23080768585205078
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,4,1,power_law_1.2,0.27703487396240234
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,4,1,power_law_1.2,0.3448275375366211
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,4,1,power_law_1.2,9.03734375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,4,1,power_law_1.2,0.42156929016113287
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,4,1,power_law_1.2,0.5876947021484374
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,4,1,power_law_1.2,0.7569657897949218
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,4,1,power_law_1.2,1.1008044433593749
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,4,1,power_law_1.2,11.758851318359374
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,4,1,power_law_1.2,1.5397286987304688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,4,1,power_law_1.2,14.992474365234376
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,4,1,power_law_1.2,0.0316595196723938
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,4,1,power_law_1.2,0.04629695892333984
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,4,1,power_law_1.2,0.05075839996337891
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,4,1,power_law_1.2,0.07649663925170899
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,4,1,power_law_1.2,0.07493631839752198
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,4,1,power_law_1.2,0.07888383865356445
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,4,1,power_law_1.2,0.07888192176818848
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,4,1,power_law_1.2,0.07761600017547607
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,4,1,power_law_1.2,0.07895552158355713
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,4,1,power_law_1.2,0.07972415924072265
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,4,1,power_law_1.2,0.08021120071411134
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,4,1,power_law_1.2,0.08142080307006835
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,4,1,power_law_1.2,0.08112511634826661
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,4,1,power_law_1.2,0.08799679756164551
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,4,1,power_law_1.2,0.09409407615661622
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,4,1,power_law_1.2,0.09706751823425293
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,4,1,power_law_1.2,2.3764857482910156
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,4,1,power_law_1.2,3.151775207519531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,4,1,power_law_1.2,0.10962112426757813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,4,1,power_law_1.2,0.13193535804748535
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,4,1,power_law_1.2,0.2051315116882324
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,4,1,power_law_1.2,0.25608383178710936
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,4,1,power_law_1.2,0.1540294361114502
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,4,1,power_law_1.2,4.123135375976562
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,4,1,power_law_1.2,0.4645497512817383
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,4,1,power_law_1.2,0.3676492691040039
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,4,1,power_law_1.2,0.68494140625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,4,1,power_law_1.2,0.904101791381836
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,1,power_law_1.01,0.0811558437347412
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,1,power_law_1.01,0.10099072456359863
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,1,power_law_1.01,0.14258048057556152
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,1,power_law_1.01,0.2848179244995117
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,1,power_law_1.01,0.3876006317138672
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,1,power_law_1.01,0.5119699096679687
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,1,power_law_1.01,0.694097900390625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,1,power_law_1.01,0.7248697662353516
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,1,power_law_1.01,0.7344153594970704
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,1,power_law_1.01,0.7420537567138672
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,1,power_law_1.01,0.7802687835693359
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,1,power_law_1.01,0.8354438018798828
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,1,power_law_1.01,0.8617772674560547
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,1,power_law_1.01,0.9206310272216797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,1,power_law_1.01,0.9804179382324218
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,1,power_law_1.01,0.9908972930908202
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,4,1,power_law_1.2,1.3991526794433593
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,1,power_law_1.01,1.0224979400634766
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,1,power_law_1.01,1.142065887451172
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,1,power_law_1.01,1.210176010131836
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,1,power_law_1.01,1.4213516235351562
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,1,power_law_1.01,1.612743682861328
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,1,power_law_1.01,2.055699768066406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,1,power_law_1.01,2.596263732910156
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,1,power_law_1.01,3.5241452026367184
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,1,power_law_1.01,4.2625402832031245
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,1,power_law_1.01,6.175715942382813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,4,1,power_law_1.2,1.842512664794922
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,1,power_law_1.01,7.87664794921875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,1,power_law_1.01,0.07246335983276367
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,1,power_law_1.01,0.10416768074035646
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,1,power_law_1.01,0.17341695785522462
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,1,power_law_1.01,0.04816703796386719
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,1,power_law_1.01,0.254956169128418
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,1,power_law_1.01,0.4456742477416992
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,1,power_law_1.01,0.47220222473144535
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,1,power_law_1.01,0.35273151397705077
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,1,power_law_1.01,0.48238529205322267
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,1,power_law_1.01,0.48870399475097653
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,1,power_law_1.01,0.5072844696044922
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,1,power_law_1.01,0.5188000106811523
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,4,1,power_law_1.01,10.2615869140625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,1,power_law_1.01,0.5257990264892578
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,1,power_law_1.01,0.532929916381836
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,1,power_law_1.01,0.5966508865356446
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,1,power_law_1.01,0.6031852722167969
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,1,power_law_1.01,0.629354248046875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,1,power_law_1.01,0.6605657958984376
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,1,power_law_1.01,0.6974956512451171
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,1,power_law_1.01,0.773443832397461
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,1,power_law_1.01,0.8468492889404298
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,1,power_law_1.01,0.9964511871337891
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,1,power_law_1.01,1.1660486602783204
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,1,power_law_1.01,1.49848388671875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,1,power_law_1.01,1.843933410644531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,1,power_law_1.01,2.4914784240722656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,1,power_law_1.01,3.1446017456054687
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,4,1,power_law_1.01,3.8139105224609375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,4,1,power_law_1.2,2.2569349670410155
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,4,1,power_law_1.01,0.0771673583984375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,4,1,power_law_1.01,0.1614009666442871
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,4,1,power_law_1.01,0.06211391925811768
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,4,1,power_law_1.01,0.09173952102661133
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,4,1,power_law_1.01,0.19770240783691406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,4,1,power_law_1.01,0.2691263961791992
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,4,1,power_law_1.01,0.2930406379699707
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,4,1,power_law_1.01,0.28406080245971677
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,4,1,power_law_1.01,0.29622592926025393
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,4,1,power_law_1.01,0.31525760650634765
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,4,1,power_law_1.01,0.32285438537597655
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,4,1,power_law_1.01,0.33088447570800783
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,4,1,power_law_1.01,0.3443833541870117
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,4,1,power_law_1.01,0.3585958480834961
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,4,1,power_law_1.01,0.369444465637207
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,4,1,power_law_1.01,0.39392513275146485
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,4,1,power_law_1.01,0.12277312278747558
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,4,1,power_law_1.01,0.44580608367919916
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,4,1,power_law_1.01,0.6032358551025391
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,4,1,power_law_1.01,0.7152108764648437
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,4,1,power_law_1.01,0.4994015884399414
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,4,1,power_law_1.01,0.9341587066650391
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,4,1,power_law_1.01,1.1608812713623047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,4,1,power_law_1.01,1.6638067626953124
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,4,1,power_law_1.01,2.188579864501953
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,1,power_law_1.2,0.08128959655761718
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,1,power_law_1.2,0.10215167999267578
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,1,power_law_1.2,0.1449843215942383
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,1,power_law_1.2,0.2558982467651367
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,1,power_law_1.2,0.36714881896972656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,1,power_law_1.2,0.4904358291625977
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,1,power_law_1.2,0.6865837097167968
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,1,power_law_1.2,0.7202508544921875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,1,power_law_1.2,0.7304467010498047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,1,power_law_1.2,0.7498566436767578
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,1,power_law_1.2,0.7844249725341796
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,1,power_law_1.2,0.8331404876708983
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,1,power_law_1.2,0.8726796722412109
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,4,1,power_law_1.01,3.2355194091796875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,1,power_law_1.2,0.953658905029297
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,4,1,power_law_1.01,4.089097595214843
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,1,power_law_1.2,0.986495361328125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,1,power_law_1.2,1.002529296875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,1,power_law_1.2,1.0507097625732422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,1,power_law_1.2,1.168609313964844
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,1,power_law_1.2,1.2299116516113282
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,1,power_law_1.2,1.485704345703125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,7168,2048,8,256,4,1,power_law_1.01,5.162235107421875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,1,power_law_1.2,1.6282687377929688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,1,power_law_1.2,2.0285504150390627
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,1,power_law_1.2,2.7089553833007813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,1,power_law_1.2,3.2604205322265627
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,1,power_law_1.2,4.251720275878906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,1,power_law_1.2,0.05057343959808349
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,1,power_law_1.2,0.07204031944274902
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,1,power_law_1.2,0.10353983879089354
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,1,power_law_1.2,5.64699462890625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,1,power_law_1.2,0.16431808471679688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,1,power_law_1.2,0.24453567504882812
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,1,power_law_1.2,0.33985214233398436
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,1,power_law_1.2,0.43923713684082033
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,1,power_law_1.2,0.4572985458374023
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,1,power_law_1.2,0.47315200805664065
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,1,power_law_1.2,8.120008544921875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,1,power_law_1.2,0.4900838470458984
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,1,power_law_1.2,0.49976638793945316
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,1,power_law_1.2,0.5157548904418945
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,1,power_law_1.2,0.5272307205200195
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,1,power_law_1.2,0.544537582397461
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,1,power_law_1.2,0.6045830535888672
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,1,power_law_1.2,0.6224307250976563
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,1,power_law_1.2,0.6387628936767579
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,1,power_law_1.2,0.6853376007080078
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,1,power_law_1.2,0.7245561981201172
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,4,1,power_law_1.2,9.498062133789062
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,1,power_law_1.2,0.8035660552978516
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,1,power_law_1.2,0.8734822082519532
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,1,power_law_1.2,1.045978240966797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,1,power_law_1.2,1.2114937591552735
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,1,power_law_1.2,1.5283654785156249
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,1,power_law_1.2,1.8709400939941407
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,1,power_law_1.2,2.533477172851563
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,4,1,power_law_1.2,0.06359168052673339
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,1,power_law_1.2,3.17517822265625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,4,1,power_law_1.2,0.07682112216949463
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,4,1,power_law_1.2,0.08856063842773437
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,4,1,power_law_1.2,0.11651007652282715
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,4,1,power_law_1.2,0.1511616039276123
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,4,1,power_law_1.2,3.8228857421874998
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,4,1,power_law_1.2,0.1889459228515625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,4,1,power_law_1.2,0.26571840286254883
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,4,1,power_law_1.2,0.2777132797241211
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,4,1,power_law_1.2,0.28843135833740235
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,4,1,power_law_1.2,0.29996480941772463
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,4,1,power_law_1.2,0.3106092834472656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,4,1,power_law_1.2,0.32333118438720704
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,4,1,power_law_1.2,0.33796417236328125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,4,1,power_law_1.2,0.35928256988525387
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,4,1,power_law_1.2,0.3746355056762695
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,4,1,power_law_1.2,0.3893523025512695
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,4,1,power_law_1.2,0.4302195358276367
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,4,1,power_law_1.2,0.5026022338867188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,4,1,power_law_1.2,0.5745299148559571
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,4,1,power_law_1.2,0.7442924499511718
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,4,1,power_law_1.2,1.282476806640625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,4,1,power_law_1.2,0.9182367706298828
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,4,1,power_law_1.2,1.6937152099609374
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,4,1,power_law_1.2,2.506680908203125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,1,power_law_1.01,0.03603775978088379
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,1,power_law_1.01,0.04385536193847656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,1,power_law_1.01,0.050474238395690915
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,1,power_law_1.01,0.057925758361816404
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,1,power_law_1.01,0.07518400192260742
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,1,power_law_1.01,0.07726208209991456
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,1,power_law_1.01,0.0801478385925293
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,1,power_law_1.01,0.08239423751831054
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,1,power_law_1.01,0.08426239967346191
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,1,power_law_1.01,0.08963711738586426
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,1,power_law_1.01,0.0990668773651123
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,1,power_law_1.01,0.10094464302062989
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,1,power_law_1.01,0.10442496299743653
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,1,power_law_1.01,0.10808639526367188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,1,power_law_1.01,0.11080767631530761
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,1,power_law_1.01,0.11828927993774414
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,1,power_law_1.01,0.13643775939941408
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,1,power_law_1.01,0.16092031478881835
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,1,power_law_1.01,0.18976064682006838
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,1,power_law_1.01,0.22362688064575192
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,1,power_law_1.01,0.2913759994506836
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,1,power_law_1.01,0.36055103302001956
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,4,1,power_law_1.2,3.364772338867188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,1,power_law_1.01,0.4973580932617187
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,1,power_law_1.01,0.6298515319824218
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,1,power_law_1.01,0.9077011108398437
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,1,power_law_1.01,1.172899856567383
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,2048,768,8,128,4,1,power_law_1.01,1.445662078857422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,1,power_law_1.01,0.03370431900024414
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,1,power_law_1.01,0.04195007801055908
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,1,power_law_1.01,0.04808000087738037
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,1,power_law_1.01,0.05724671840667724
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,1,power_law_1.01,0.07428927898406983
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,1,power_law_1.01,0.0801734447479248
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,1,power_law_1.01,0.08206463813781738
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,1,power_law_1.01,0.08380543708801269
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,1,power_law_1.01,0.08436991691589354
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,1,power_law_1.01,0.0852128028869629
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,1,power_law_1.01,0.09590784072875977
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,1,power_law_1.01,0.098023681640625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,1,power_law_1.01,0.03209791898727417
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,1,power_law_1.01,0.09968447685241699
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,1,power_law_1.01,0.10267711639404298
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,1,power_law_1.01,0.10412287712097168
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,1,power_law_1.01,0.1107046413421631
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,1,power_law_1.01,0.11732288360595704
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,1,power_law_1.01,0.13278592109680176
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,1,power_law_1.01,0.1551961612701416
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,1,power_law_1.01,0.18007488250732423
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,1,power_law_1.01,0.227512321472168
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,1,power_law_1.01,0.27632383346557615
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,1,power_law_1.01,0.37670528411865234
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,1,power_law_1.01,0.46812160491943355
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,1,power_law_1.01,0.6714694213867187
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,1,power_law_1.01,0.861045150756836
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,2048,768,8,128,4,1,power_law_1.01,1.0596275329589844
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,4,1,power_law_1.01,0.04069759845733643
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,4,1,power_law_1.01,0.047144317626953126
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,4,1,power_law_1.01,0.0505350399017334
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,4,1,power_law_1.01,0.05925183773040772
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,4,1,power_law_1.01,0.06149759769439698
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,4,1,power_law_1.01,0.061987199783325196
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,4,1,power_law_1.01,0.04193151950836181
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,4,1,power_law_1.01,0.063024001121521
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,4,1,power_law_1.01,0.06413695812225342
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,4,1,power_law_1.01,0.06533376216888429
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,4,1,power_law_1.01,0.06665855884552002
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,4,1,power_law_1.01,0.06711552143096924
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,4,1,power_law_1.01,0.07083583831787109
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,4,1,power_law_1.01,0.04292287826538086
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,4,1,power_law_1.01,0.07325183868408203
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,4,1,power_law_1.01,0.07553728103637695
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,4,1,power_law_1.01,0.08003456115722656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,4,1,power_law_1.01,0.1068070411682129
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,4,1,power_law_1.01,0.09276479721069336
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,4,1,power_law_1.01,0.13637375831604004
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,4,1,power_law_1.01,0.16245952606201172
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,4,1,power_law_1.01,0.22499008178710939
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,4,1,power_law_1.01,0.2900927925109863
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,4,1,power_law_1.01,0.5575545501708985
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,4,1,power_law_1.01,0.42432830810546873
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,1,power_law_1.2,0.02801919937133789
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,1,power_law_1.2,0.03492799997329712
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,1,power_law_1.2,0.04215616226196289
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,4,1,power_law_1.01,0.8426457977294921
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,1,power_law_1.2,0.04927999973297119
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,1,power_law_1.2,0.07591040134429931
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,1,power_law_1.2,0.07803711891174317
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,1,power_law_1.2,0.08190336227416992
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,1,power_law_1.2,0.084268159866333
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,4,1,power_law_1.01,1.13608642578125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,1,power_law_1.2,0.08563584327697754
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,1,power_law_1.2,0.1006662368774414
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,1,power_law_1.2,0.10241727828979492
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,1,power_law_1.2,0.10678079605102539
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,1,power_law_1.2,0.056467199325561525
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,1,power_law_1.2,0.10954175949096681
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,1,power_law_1.2,0.11447232246398927
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,1,power_law_1.2,0.11907903671264648
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,1,power_law_1.2,0.08977919578552246
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,1,power_law_1.2,0.13832511901855468
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,1,power_law_1.2,0.16108671188354493
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,1,power_law_1.2,0.19138816833496092
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,1,power_law_1.2,0.2220147132873535
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,1,power_law_1.2,0.2909267234802246
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,1,power_law_1.2,0.3602316665649414
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,2048,768,8,128,4,1,power_law_1.01,1.4376409912109376
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,1,power_law_1.2,0.49508926391601565
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,1,power_law_1.2,0.6343731307983399
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,1,power_law_1.2,0.03134079933166504
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,1,power_law_1.2,0.03338752031326294
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,1,power_law_1.2,0.04024127960205078
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,1,power_law_1.2,0.04698944091796875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,1,power_law_1.2,0.055392642021179196
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,1,power_law_1.2,0.07499775886535645
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,1,power_law_1.2,0.07953728199005126
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,1,power_law_1.2,0.08281920433044435
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,1,power_law_1.2,0.08412672042846679
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,1,power_law_1.2,0.08432640075683594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,1,power_law_1.2,0.08542336463928223
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,1,power_law_1.2,0.0977184009552002
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,1,power_law_1.2,0.09834431648254395
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,1,power_law_1.2,0.10135487556457519
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,1,power_law_1.2,0.10380288124084473
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,1,power_law_1.2,0.10621888160705566
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,1,power_law_1.2,0.11127872467041015
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,1,power_law_1.2,0.11949888229370118
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,1,power_law_1.2,0.13687423706054686
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,1,power_law_1.2,0.1591212844848633
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,1,power_law_1.2,0.18436864852905271
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,1,power_law_1.2,0.22819583892822265
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,1,power_law_1.2,0.27941568374633785
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,1,power_law_1.2,0.3780531311035156
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,1,power_law_1.2,0.9086438751220702
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,1,power_law_1.2,0.4707315063476562
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,1,power_law_1.2,1.1740908813476563
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,1,power_law_1.2,0.6703667449951172
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,2048,768,8,128,4,1,power_law_1.2,1.4451397705078126
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,4,1,power_law_1.2,0.04196479797363281
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,4,1,power_law_1.2,0.04436863899230957
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,4,1,power_law_1.2,0.04639232158660889
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,4,1,power_law_1.2,0.04069375991821289
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,4,1,power_law_1.2,0.04967487812042236
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,4,1,power_law_1.2,0.05948607921600342
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,4,1,power_law_1.2,0.06165760040283204
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,4,1,power_law_1.2,0.06290112018585206
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,4,1,power_law_1.2,0.0637113618850708
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,4,1,power_law_1.2,0.06431935787200928
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,4,1,power_law_1.2,0.06574783802032472
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,4,1,power_law_1.2,0.0684768009185791
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,4,1,power_law_1.2,0.06880703926086426
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,4,1,power_law_1.2,0.07714111804962158
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,4,1,power_law_1.2,0.0809119987487793
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,4,1,power_law_1.2,0.07315072059631347
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,4,1,power_law_1.2,0.08896832466125489
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,1,power_law_1.2,0.8657344055175781
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,4,1,power_law_1.2,0.1294540786743164
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,4,1,power_law_1.2,0.10884160041809081
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,4,1,power_law_1.2,0.1752070426940918
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,4,1,power_law_1.2,0.2218828773498535
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,2048,768,8,128,4,1,power_law_1.2,1.0540480041503906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,4,1,power_law_1.2,0.3217772674560547
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,4,1,power_law_1.2,0.4265580749511718
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,4,1,power_law_1.2,0.6241376113891601
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,4,1,power_law_1.2,0.821905288696289
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,1,power_law_1.01,0.03865792036056519
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,1,power_law_1.01,0.06591296195983887
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,1,power_law_1.01,0.08453887939453125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,1,power_law_1.01,0.11175359725952147
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,1,power_law_1.01,0.13748607635498047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,1,power_law_1.01,0.22411136627197265
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,1,power_law_1.01,0.2244211196899414
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,1,power_law_1.01,0.22803840637207032
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,1,power_law_1.01,0.23041791915893556
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,1,power_law_1.01,0.22978303909301756
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,1,power_law_1.01,0.2435353660583496
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,1,power_law_1.01,0.26460479736328124
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,1,power_law_1.01,0.26761344909667967
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,1,power_law_1.01,0.274455680847168
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,1,power_law_1.01,0.2790707206726074
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,1,power_law_1.01,0.282992000579834
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,1,power_law_1.01,0.2951136016845703
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,1,power_law_1.01,0.36668865203857426
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,1,power_law_1.01,0.39502849578857424
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,1,power_law_1.01,0.4782912063598633
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,1,power_law_1.01,0.5695897674560546
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,4,1,power_law_1.2,1.2276403045654296
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,1,power_law_1.01,0.7485343933105468
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,1,power_law_1.01,0.9301420593261719
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,1,power_law_1.01,1.2922642517089844
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,1,power_law_1.01,1.6566368103027345
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,1,power_law_1.01,2.4269926452636716
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,1,power_law_1.01,3.1244583129882812
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,4,1,power_law_1.01,3.933411254882812
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,1,power_law_1.01,0.04931647777557373
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,1,power_law_1.01,0.06759424209594726
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,1,power_law_1.01,0.08959615707397461
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,1,power_law_1.01,0.11882176399230956
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,1,power_law_1.01,0.15864255905151367
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,1,power_law_1.01,0.15988927841186523
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,1,power_law_1.01,0.16196863174438475
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,1,power_law_1.01,0.16603519439697265
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,1,power_law_1.01,0.1672729682922363
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,1,power_law_1.01,0.17122175216674804
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,1,power_law_1.01,0.03784640073776245
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,1,power_law_1.01,0.19168319702148437
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,1,power_law_1.01,0.19868736267089843
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,1,power_law_1.01,0.20241407394409178
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,1,power_law_1.01,0.20527679443359376
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,1,power_law_1.01,0.21527103424072264
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,1,power_law_1.01,0.22985919952392578
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,1,power_law_1.01,0.2567500877380371
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,1,power_law_1.01,0.19556608200073242
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,1,power_law_1.01,0.2935308837890625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,1,power_law_1.01,0.3498944091796875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,1,power_law_1.01,0.43703872680664063
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,1,power_law_1.01,0.5236371231079102
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,1,power_law_1.01,0.7113401794433594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,1,power_law_1.01,0.8919999694824219
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,1,power_law_1.01,1.2608710479736327
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,1,power_law_1.01,1.6288319396972657
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,4,1,power_law_1.01,1.9830047607421875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,4,1,power_law_1.01,0.044721279144287106
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,4,1,power_law_1.01,0.05290688037872314
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,4,1,power_law_1.01,0.06301631927490234
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,4,1,power_law_1.01,0.03948224067687988
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,4,1,power_law_1.01,0.07246016025543213
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,4,1,power_law_1.01,0.09534144401550293
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,4,1,power_law_1.01,0.10043583869934083
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,4,1,power_law_1.01,0.10271231651306152
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,4,1,power_law_1.2,1.6446528625488281
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,4,1,power_law_1.01,0.10463616371154785
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,4,1,power_law_1.01,0.10644607543945313
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,4,1,power_law_1.01,0.11325440406799317
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,4,1,power_law_1.01,0.11681920051574708
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,4,1,power_law_1.01,0.12265472412109375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,2048,768,8,128,4,1,power_law_1.2,2.0559814453125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,4,1,power_law_1.01,0.11099840164184571
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,4,1,power_law_1.01,0.13596223831176757
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,4,1,power_law_1.01,0.13126784324645996
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,4,1,power_law_1.01,0.15103936195373535
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,4,1,power_law_1.01,0.1816192054748535
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,4,1,power_law_1.01,0.21086847305297854
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,4,1,power_law_1.01,0.2787545585632324
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,4,1,power_law_1.01,0.3462995147705078
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,4,1,power_law_1.01,0.4915955352783203
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,4,1,power_law_1.01,0.6301388931274414
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,4,1,power_law_1.01,0.9389311981201172
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,4,1,power_law_1.01,1.2517804718017578
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,1,power_law_1.2,0.044154877662658694
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,4,1,power_law_1.01,1.8808038330078127
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,1,power_law_1.2,0.08021439552307129
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,1,power_law_1.2,0.06440447807312012
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,1,power_law_1.2,0.10232768058776856
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,4,1,power_law_1.01,2.578608093261719
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,1,power_law_1.2,0.1261836814880371
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,1,power_law_1.2,0.22144575119018556
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,1,power_law_1.2,0.22430463790893557
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,1,power_law_1.2,0.23115327835083005
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,1,power_law_1.2,0.23093439102172852
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,1,power_law_1.2,0.24582847595214843
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,1,power_law_1.2,0.26552831649780273
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,1,power_law_1.2,0.23160127639770506
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,1,power_law_1.2,0.27602943420410153
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,1,power_law_1.2,0.2698080062866211
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,1,power_law_1.2,0.2790438461303711
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,1,power_law_1.2,0.28544704437255863
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,1,power_law_1.2,0.2996063995361328
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,1,power_law_1.2,0.3683232116699219
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,1,power_law_1.2,0.3993824005126953
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,1,power_law_1.2,0.4897785568237305
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,1,power_law_1.2,0.5691212844848633
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,1,power_law_1.2,0.7549971008300782
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,1,power_law_1.2,0.9316422271728516
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,1,power_law_1.2,1.3025657653808593
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,1536,8,128,4,1,power_law_1.01,3.2222686767578126
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,1,power_law_1.2,1.6689555358886718
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,1,power_law_1.2,0.036661760807037355
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,1,power_law_1.2,0.048703360557556155
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,1,power_law_1.2,0.06566656112670899
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,1,power_law_1.2,0.08029567718505859
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,1,power_law_1.2,0.11111680030822754
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,1,power_law_1.2,0.15004735946655273
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,1,power_law_1.2,0.15891712188720702
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,1,power_law_1.2,0.16430528640747072
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,1,power_law_1.2,0.16769535064697266
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,1,power_law_1.2,0.16740863800048827
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,1,power_law_1.2,0.17038400650024416
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,1,power_law_1.2,0.19243520736694336
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,1,power_law_1.2,0.194385929107666
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,1,power_law_1.2,0.20100095748901364
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,1,power_law_1.2,0.20492416381835937
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,1,power_law_1.2,0.20778623580932618
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,1,power_law_1.2,0.21769407272338864
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,1,power_law_1.2,0.23590208053588868
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,1,power_law_1.2,0.26301055908203125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,1,power_law_1.2,0.3061369514465332
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,1,power_law_1.2,0.35167297363281247
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,1,power_law_1.2,0.4408915328979492
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,1,power_law_1.2,0.532426872253418
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,1,power_law_1.2,2.377406005859375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,1,power_law_1.2,0.7171180725097657
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,1,power_law_1.2,0.8915219116210938
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,1,power_law_1.2,3.158186950683594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,1,power_law_1.2,1.2666867065429688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,4,1,power_law_1.2,0.03942271947860718
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,4,1,power_law_1.2,0.04450240135192871
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,4,1,power_law_1.2,0.05254591941833496
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,4,1,power_law_1.2,0.061098241806030275
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,4,1,power_law_1.2,0.07080832004547119
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,4,1,power_law_1.2,0.09642047882080078
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,4,1,power_law_1.2,0.1012979221343994
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,4,1,power_law_1.2,0.10439552307128905
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,4,1,power_law_1.2,0.10666175842285157
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,4,1,power_law_1.2,0.10859264373779295
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,4,1,power_law_1.2,0.11255552291870116
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,4,1,power_law_1.2,0.1172332763671875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,4,1,power_law_1.2,0.12100288391113281
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,4,1,power_law_1.2,0.1307910442352295
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,4,1,power_law_1.2,0.1405247974395752
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,1,power_law_1.2,1.6248524475097657
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,4,1,power_law_1.2,0.14948543548583984
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,4,1,power_law_1.2,0.16955583572387695
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,4,1,power_law_1.2,0.21299840927124025
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,4,1,power_law_1.2,3.849284973144531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,4,1,power_law_1.2,0.25864128112792967
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,4,1,power_law_1.2,0.356759033203125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,4,1,power_law_1.2,1.9865298461914063
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,4,1,power_law_1.2,0.6708415985107422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,4,1,power_law_1.2,0.4551615905761719
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,4,1,power_law_1.2,0.8996025848388672
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,4,1,power_law_1.2,1.3308396911621094
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,4,1,power_law_1.2,1.7644383239746095
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,1,power_law_1.01,0.08414655685424804
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,1,power_law_1.01,0.10067520141601563
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,1,power_law_1.01,0.15222208023071288
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,1,power_law_1.01,0.25888704299926757
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,1,power_law_1.01,0.32960254669189454
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,1,power_law_1.01,0.49314174652099607
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,1,power_law_1.01,0.5200678253173828
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,1,power_law_1.01,0.5442636871337891
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,1,power_law_1.01,0.5699391937255859
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,1,power_law_1.01,0.5817644882202149
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,1,power_law_1.01,0.6091328048706055
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,1,power_law_1.01,0.6536511993408203
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,1,power_law_1.01,0.6758924865722656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,1,power_law_1.01,0.7288108825683594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,1,power_law_1.01,0.7566585540771484
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,1,power_law_1.01,0.7621100616455079
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,1,power_law_1.01,0.8202899169921875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,1,power_law_1.01,0.8750892639160156
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,1,power_law_1.01,0.9680774688720704
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,1,power_law_1.01,1.150096664428711
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,1,power_law_1.01,1.3725260925292968
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,1,power_law_1.01,1.8114303588867187
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,1,power_law_1.01,2.246897888183594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,1,power_law_1.01,3.12786376953125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,1,power_law_1.01,3.885094299316406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,1,power_law_1.01,5.65447021484375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,4,1,power_law_1.2,2.6229089355468753
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,1,power_law_1.01,7.866470947265626
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,1,power_law_1.01,0.05505983829498291
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,1,power_law_1.01,0.07413119792938232
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,1,power_law_1.01,0.11989184379577637
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,1,power_law_1.01,0.15601984024047852
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,1,power_law_1.01,0.2156319999694824
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,1,power_law_1.01,0.3176460838317871
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,1,power_law_1.01,0.33446014404296875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,1,power_law_1.01,0.3426342391967773
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,1,power_law_1.01,0.351212158203125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,1,power_law_1.01,0.355544319152832
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,1,power_law_1.01,0.3631033706665039
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,4,1,power_law_1.01,9.732293701171875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,1,power_law_1.01,0.37436286926269535
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,1,power_law_1.01,0.4176006317138672
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,1,power_law_1.01,0.4351257705688477
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,1,power_law_1.01,0.43709056854248046
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,4,1,power_law_1.2,3.4946490478515626
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,1,power_law_1.01,0.4266604614257813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,1,power_law_1.01,0.4615711975097656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,1,power_law_1.01,0.4926623916625976
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,1,power_law_1.01,0.5237740707397461
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,1,power_law_1.01,0.5940639877319336
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,1,power_law_1.01,0.6910297393798828
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,1,power_law_1.01,0.84557373046875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,1,power_law_1.01,1.0112076568603516
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,1,power_law_1.01,1.3492723083496094
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,1,power_law_1.01,1.713072052001953
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,1,power_law_1.01,2.3097798156738283
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,1,power_law_1.01,2.9689785766601564
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,4,1,power_law_1.01,3.6323028564453126
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,4,1,power_law_1.01,0.04799871921539307
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,4,1,power_law_1.01,0.10160832405090332
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,4,1,power_law_1.01,0.0656774377822876
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,4,1,power_law_1.01,0.1267155170440674
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,4,1,power_law_1.01,0.18839359283447266
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,4,1,power_law_1.01,0.19799104690551758
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,4,1,power_law_1.01,0.20703487396240233
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,4,1,power_law_1.01,0.2125382423400879
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,4,1,power_law_1.01,0.2159859275817871
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,4,1,power_law_1.01,0.22380863189697267
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,4,1,power_law_1.01,0.23237056732177735
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,4,1,power_law_1.01,0.23405632019042968
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,4,1,power_law_1.01,0.24954624176025392
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,4,1,power_law_1.01,0.2564326477050781
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,4,1,power_law_1.01,0.2622208023071289
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,4,1,power_law_1.01,0.0792409610748291
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,1536,8,128,4,1,power_law_1.2,4.358596496582031
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,4,1,power_law_1.01,0.28778432846069335
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,4,1,power_law_1.01,0.3376684951782226
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,4,1,power_law_1.01,0.38223934173583984
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,4,1,power_law_1.01,0.4791360092163086
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,4,1,power_law_1.01,0.5909222412109375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,4,1,power_law_1.01,0.8101388549804687
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,4,1,power_law_1.01,1.0339373016357423
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,4,1,power_law_1.01,1.5022624206542967
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,4,1,power_law_1.01,1.9776864624023438
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,1,power_law_1.2,0.08404800415039063
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,1,power_law_1.2,0.10074687957763673
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,1,power_law_1.2,0.15025792121887208
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,1,power_law_1.2,0.23091007232666017
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,1,power_law_1.2,0.3152364730834961
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,1,power_law_1.2,0.4885894393920898
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,4,1,power_law_1.01,2.9558770751953127
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,1,power_law_1.2,0.5106444931030273
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,1,power_law_1.2,0.5518310546875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,1,power_law_1.2,0.5804467010498047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,1,power_law_1.2,0.6036486434936523
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,1,power_law_1.2,0.6290566253662109
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,1,power_law_1.2,0.680097885131836
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,1,power_law_1.2,0.7186681365966796
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,1,power_law_1.2,0.7533663940429688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,1,power_law_1.2,0.7680448150634765
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,1,power_law_1.2,0.7806323242187501
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,1,power_law_1.2,0.8146176147460938
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,1,power_law_1.2,0.9034809875488282
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,1,power_law_1.2,0.9712268829345703
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,1,power_law_1.2,1.1665945434570313
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,1,power_law_1.2,1.3618605041503906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,1,power_law_1.2,1.8748077392578124
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,1,power_law_1.2,2.258672637939453
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,1,power_law_1.2,3.1429171752929688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,1,power_law_1.2,3.9586868286132812
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,2560,8,160,4,1,power_law_1.01,4.964942626953125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,1,power_law_1.2,5.632964477539063
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,1,power_law_1.2,0.05312640190124511
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,1,power_law_1.2,0.07491136074066163
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,1,power_law_1.2,0.11181759834289551
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,1,power_law_1.2,0.14711551666259765
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,1,power_law_1.2,0.20937471389770507
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,4,1,power_law_1.01,3.9627923583984375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,1,power_law_1.2,0.3135603141784668
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,1,power_law_1.2,0.3463033676147461
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,1,power_law_1.2,0.3329094314575195
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,1,power_law_1.2,0.35683841705322267
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,1,power_law_1.2,0.36155712127685546
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,1,power_law_1.2,0.3728550338745117
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,1,power_law_1.2,0.37927745819091796
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,1,power_law_1.2,0.4355295944213867
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,1,power_law_1.2,0.43818943023681645
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,1,power_law_1.2,0.4519865417480469
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,1,power_law_1.2,0.4600352096557617
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,1,power_law_1.2,0.47644096374511713
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,1,power_law_1.2,0.49709823608398435
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,1,power_law_1.2,0.5404390335083008
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,1,power_law_1.2,0.6188147354125977
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,1,power_law_1.2,0.6987712097167968
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,1,power_law_1.2,0.8693894195556642
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,1,power_law_1.2,1.027105255126953
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,1,power_law_1.2,1.3677862548828126
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,1,power_law_1.2,7.6723193359375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,1,power_law_1.2,1.6837107849121093
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,1,power_law_1.2,2.3413183593749998
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,4,1,power_law_1.2,0.04846784114837647
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,4,1,power_law_1.2,0.062447361946105964
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,4,1,power_law_1.2,0.07720064163208008
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,4,1,power_law_1.2,0.09625408172607422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,4,1,power_law_1.2,0.1210048007965088
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,1,power_law_1.2,2.9712811279296876
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,4,1,power_law_1.2,0.1856185531616211
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,4,1,power_law_1.2,0.19619264602661132
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,4,1,power_law_1.2,0.2159116744995117
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,4,1,power_law_1.2,0.2077107238769531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,4,1,power_law_1.2,0.22077888488769531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,4,1,power_law_1.2,0.23041215896606446
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,4,1,power_law_1.2,9.817191162109376
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,4,1,power_law_1.2,0.23612031936645508
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,4,1,power_law_1.2,0.27617727279663085
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,4,1,power_law_1.2,0.2577907180786133
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,4,1,power_law_1.2,0.29007232666015625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,4,1,power_law_1.2,3.627646179199219
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,4,1,power_law_1.2,0.24655296325683596
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,4,1,power_law_1.2,0.3849497604370117
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,4,1,power_law_1.2,0.45807361602783203
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,4,1,power_law_1.2,0.3230662536621094
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,4,1,power_law_1.2,0.6140595245361328
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,4,1,power_law_1.2,0.7823865509033203
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,4,1,power_law_1.2,1.151476516723633
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,4,1,power_law_1.2,1.5352890014648437
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,4,1,power_law_1.2,2.342481994628906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,4,1,power_law_1.2,3.1627392578125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,1,power_law_1.01,0.08871487617492677
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,1,power_law_1.01,0.10511103630065918
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,1,power_law_1.01,0.1458956813812256
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,1,power_law_1.01,0.2861996841430664
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,1,power_law_1.01,0.4701830291748047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,1,power_law_1.01,0.6288095855712891
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,1,power_law_1.01,0.7329081726074219
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,1,power_law_1.01,0.9246361541748047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,1,power_law_1.01,0.9648767852783203
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,1,power_law_1.01,1.006492156982422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,1,power_law_1.01,1.048312301635742
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,1,power_law_1.01,1.0635040283203125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,4,1,power_law_1.2,4.843110961914062
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,1,power_law_1.01,1.1135174560546877
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,1,power_law_1.01,1.1903814697265624
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,1,power_law_1.01,1.247864990234375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,1,power_law_1.01,1.3406752014160157
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,1,power_law_1.01,1.3469247436523437
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,1,power_law_1.01,1.4728268432617189
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,1,power_law_1.01,1.5431173706054688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,1,power_law_1.01,1.7741049194335936
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,1,power_law_1.01,1.925977020263672
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,1,power_law_1.01,2.4764472961425783
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,1,power_law_1.01,2.883436279296875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,1,power_law_1.01,3.7494354248046875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,1,power_law_1.01,4.513276062011719
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,1,power_law_1.01,8.219232788085938
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,1,power_law_1.01,0.052614398002624505
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,1,power_law_1.01,0.07546688079833984
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,1,power_law_1.01,0.10731648445129394
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,1,power_law_1.01,0.17705984115600587
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,1,power_law_1.01,0.3019935989379883
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,4,1,power_law_1.01,9.902843627929688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,1,power_law_1.01,0.42334590911865233
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,1,power_law_1.01,0.5051001739501954
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,1,power_law_1.01,0.6365887832641601
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,1,power_law_1.01,0.6094496154785156
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,1,power_law_1.01,0.6625331115722657
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,1,power_law_1.01,0.6771263885498047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,1,power_law_1.01,0.6883155059814453
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,1,power_law_1.01,0.7066860961914062
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,1,power_law_1.01,0.7322418975830078
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,1,power_law_1.01,0.7357574462890625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,4,1,power_law_1.2,6.5077972412109375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,1,power_law_1.01,0.7649024200439454
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,1,power_law_1.01,0.8440262603759766
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,1,power_law_1.01,0.9150137329101563
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,1,power_law_1.01,0.8781215667724609
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,1,power_law_1.01,0.9982022094726564
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,1,power_law_1.01,1.080255355834961
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,1,power_law_1.01,1.240600357055664
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,1,power_law_1.01,1.4084165954589845
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,1,power_law_1.01,1.7537055969238282
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,1,power_law_1.01,2.090168304443359
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,1,power_law_1.01,2.7815155029296874
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,1,power_law_1.01,3.4367218017578125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,4,1,power_law_1.01,4.209856567382813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,4,1,power_law_1.01,0.07818624019622802
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,4,1,power_law_1.01,0.08096768379211425
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,4,1,power_law_1.01,0.07769599914550782
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,4,1,power_law_1.01,0.08545920372009277
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,4,1,power_law_1.01,0.09697983741760255
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,4,1,power_law_1.01,0.10433216094970703
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,4,1,power_law_1.01,0.11857279777526855
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,4,1,power_law_1.01,0.1291155242919922
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,4,1,power_law_1.01,0.13762432098388672
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,4,1,power_law_1.01,0.1586092758178711
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,4,1,power_law_1.01,0.07763584136962891
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,4,1,power_law_1.01,0.19922367095947266
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,4,1,power_law_1.01,0.2194905662536621
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,4,1,power_law_1.01,0.2684838485717774
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,4,1,power_law_1.01,0.3245926284790039
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,4,1,power_law_1.01,0.46689086914062494
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,4,1,power_law_1.01,0.37133121490478516
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,4,1,power_law_1.01,0.6647020721435547
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,4,1,power_law_1.01,0.8678463745117189
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,4,1,power_law_1.01,1.3018130493164062
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,4,1,power_law_1.01,1.704400634765625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,4,1,power_law_1.01,3.3280084228515627
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,4,1,power_law_1.01,2.5251916503906253
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,4,1,power_law_1.01,4.952561950683593
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,1,power_law_1.2,0.12224255561828615
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,1,power_law_1.2,0.14745408058166504
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,1,power_law_1.2,0.29337663650512696
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,1,power_law_1.2,0.41549057006835943
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,1,power_law_1.2,0.6869197082519531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,1,power_law_1.2,0.9225126647949219
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,1,power_law_1.2,0.9489920043945312
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,1,power_law_1.2,0.9726156616210938
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,1,power_law_1.2,1.013565444946289
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,1,power_law_1.2,1.0743571472167968
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,1,power_law_1.2,1.114065933227539
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,1,power_law_1.2,1.1902982330322265
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,1,power_law_1.2,1.269722900390625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,1,power_law_1.2,1.364627227783203
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,1,power_law_1.2,1.3725062561035155
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,1,power_law_1.2,1.5025106811523439
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,1,power_law_1.2,1.6161407470703124
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,1,power_law_1.2,1.8376428222656251
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,1,power_law_1.2,2.1212294006347654
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,1,power_law_1.2,2.563157653808594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,1,power_law_1.2,2.99698486328125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,1,power_law_1.2,3.803428344726562
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,1,power_law_1.2,4.640096130371093
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,1,power_law_1.2,6.383035888671875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,1,power_law_1.2,8.435134887695312
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,1,power_law_1.2,0.052602238655090325
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,1,power_law_1.2,0.07565760135650634
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,1,power_law_1.2,0.10721920013427735
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,1,power_law_1.2,0.17696063995361327
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,1,power_law_1.2,0.27111040115356444
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,1,power_law_1.2,0.4718649673461914
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,4,1,power_law_1.2,10.002560424804688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,1,power_law_1.2,0.6007609558105469
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,1,power_law_1.2,0.6379270553588867
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,1,power_law_1.2,0.6152070236206055
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,1,power_law_1.2,0.6591295623779297
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,1,power_law_1.2,0.6791283416748046
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,1,power_law_1.2,0.6989273834228515
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,1,power_law_1.2,0.7238227081298828
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,1,power_law_1.2,0.7360646057128906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,1,power_law_1.2,0.7602848052978516
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,1,power_law_1.2,0.8454220581054688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,1,power_law_1.2,0.8936460876464845
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,1,power_law_1.2,0.941470718383789
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,1,power_law_1.2,1.0299507141113282
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,1,power_law_1.2,0.3904070281982422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,1,power_law_1.2,1.1262169647216798
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,1,power_law_1.2,1.2944953918457032
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,1,power_law_1.2,1.4625523376464844
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,1,power_law_1.2,1.811078338623047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,1,power_law_1.2,2.151336975097656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,1,power_law_1.2,2.825640869140625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,1,power_law_1.2,3.520467834472656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,4,1,power_law_1.2,0.07944384098052978
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,4,1,power_law_1.2,0.0809830379486084
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,4,1,power_law_1.2,4.227299194335937
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,4,1,power_law_1.2,0.08545087814331055
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,4,1,power_law_1.2,0.10431296348571777
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,4,1,power_law_1.2,0.11715904235839844
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,4,1,power_law_1.2,0.12933055877685545
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,4,1,power_law_1.2,0.13904704093933104
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,4,1,power_law_1.2,0.15904704093933106
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,4,1,power_law_1.2,0.19611263275146484
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,4,1,power_law_1.2,0.07713535785675049
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,4,1,power_law_1.2,0.2195724868774414
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,4,1,power_law_1.2,0.09535167694091798
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,4,1,power_law_1.2,0.07748991966247558
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,4,1,power_law_1.2,0.2669100761413574
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,4,1,power_law_1.2,0.3261113739013672
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,4,1,power_law_1.2,0.36986049652099606
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,4,1,power_law_1.2,0.46825729370117186
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,4,1,power_law_1.2,0.6637004852294922
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,4,1,power_law_1.2,0.8640729522705077
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,4,1,power_law_1.2,1.2995622253417969
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,4,1,power_law_1.2,1.7027737426757814
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,4,1,power_law_1.2,2.5194003295898435
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,4,1,power_law_1.2,3.331697998046875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,4,1,power_law_1.2,4.964075622558594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,1,power_law_1.01,0.07698880195617676
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,1,power_law_1.01,0.0832915210723877
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,1,power_law_1.01,0.10471424102783203
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,1,power_law_1.01,0.165467529296875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,1,power_law_1.01,0.17121471405029295
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,1,power_law_1.01,0.17586240768432618
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,1,power_law_1.01,0.1780793571472168
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,1,power_law_1.01,0.18029056549072267
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,1,power_law_1.01,0.18999231338500977
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,1,power_law_1.01,0.2018284797668457
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,1,power_law_1.01,0.20468927383422852
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,1,power_law_1.01,0.20997951507568358
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,1,power_law_1.01,0.21028095245361328
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,1,power_law_1.01,0.2159199905395508
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,1,power_law_1.01,0.2472915267944336
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,1,power_law_1.01,0.26711679458618165
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,1,power_law_1.01,0.28905023574829103
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,1,power_law_1.01,0.34632896423339843
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,1,power_law_1.01,0.39800769805908204
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,1,power_law_1.01,0.5102355194091797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,1,power_law_1.01,0.6285542297363281
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,1,power_law_1.01,0.8608902740478517
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,1,power_law_1.01,1.0919308471679687
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,1,power_law_1.01,1.5517138671875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,1,power_law_1.01,2.0143443298339845
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,2688,1856,6,128,4,1,power_law_1.01,2.4965382385253907
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,1,power_law_1.2,0.04515520095825195
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,1,power_law_1.2,0.05967423915863037
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,1,power_law_1.2,0.07734272003173828
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,1,power_law_1.2,0.098156156539917
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,1,power_law_1.2,0.16462656021118166
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,1,power_law_1.2,0.17140352249145507
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,1,power_law_1.2,0.17591808319091798
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,1,power_law_1.2,0.17778560638427734
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,1,power_law_1.2,0.029655680656433103
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,1,power_law_1.2,0.18195072174072266
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,1,power_law_1.2,0.19245119094848634
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,1,power_law_1.2,0.20389184951782227
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,1,power_law_1.2,0.213055362701416
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,1,power_law_1.2,0.2141708755493164
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,1,power_law_1.2,0.20382272720336916
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,1,power_law_1.2,0.23408767700195315
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,1,power_law_1.2,0.2705971145629883
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,1,power_law_1.2,0.289051513671875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,1,power_law_1.2,0.4030335998535156
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,1,power_law_1.2,0.5164838409423828
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,1,power_law_1.2,0.630263671875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,1,power_law_1.2,0.34668033599853515
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,1,power_law_1.2,0.8666092681884766
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,1,power_law_1.2,1.1006175994873046
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,1,power_law_1.2,1.5539833068847657
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,1,power_law_1.2,2.0252198791503906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,2688,1856,6,128,4,1,power_law_1.2,2.483906555175781
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,1,power_law_1.01,0.17388608932495117
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,1,power_law_1.01,0.09730688095092774
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,1,power_law_1.01,0.287578239440918
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,1,power_law_1.01,0.4677471923828125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,1,power_law_1.01,1.0203091430664064
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,1,power_law_1.01,0.6404121398925782
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,1,power_law_1.01,1.0935648345947266
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,1,power_law_1.01,1.1296896362304687
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,1,power_law_1.01,1.1763097381591796
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,1,power_law_1.01,1.2112716674804687
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,1,power_law_1.01,1.2742777252197266
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,1,power_law_1.01,1.3952000427246094
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,1,power_law_1.01,1.4576966857910156
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,1,power_law_1.01,1.6562425231933595
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,1,power_law_1.01,1.844865264892578
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,1,power_law_1.01,1.9903456115722655
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,1,power_law_1.01,2.3527001953125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,1,power_law_1.01,1.89359619140625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,1,power_law_1.01,2.1010150146484374
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,1,power_law_1.01,2.494446105957031
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,1,power_law_1.01,3.082006530761719
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,1,power_law_1.01,3.842711791992188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,1,power_law_1.01,4.586069030761719
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,1,power_law_1.01,6.852421264648437
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,1,power_law_1.01,8.821510620117188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,1,power_law_1.01,0.07968255996704102
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,1,power_law_1.01,0.11846783638000488
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,1,power_law_1.01,0.19263423919677733
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,1,power_law_1.01,12.0189697265625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,1,power_law_1.01,0.4212012863159179
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,1,power_law_1.01,0.6563974761962891
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,1,power_law_1.01,0.29443071365356444
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,1,power_law_1.01,0.6951551818847657
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,1,power_law_1.01,0.7521247863769531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,1,power_law_1.01,15.554498291015625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,1,power_law_1.01,0.7580345916748047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,1,power_law_1.01,0.7820761871337891
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,1,power_law_1.01,0.7264466857910156
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,1,power_law_1.01,0.8062214660644532
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,1,power_law_1.01,0.814263687133789
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,1,power_law_1.01,0.857922592163086
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,1,power_law_1.01,0.8513945770263671
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,1,power_law_1.01,0.883978271484375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,2688,22,512,4,1,power_law_1.01,20.017314453125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,1,power_law_1.01,0.9383955383300782
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,1,power_law_1.01,1.0834451293945313
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,1,power_law_1.01,1.1568959808349608
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,1,power_law_1.01,1.339563446044922
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,1,power_law_1.01,1.852156219482422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,1,power_law_1.01,2.179536590576172
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,1,power_law_1.01,1.5175059509277342
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,1,power_law_1.01,3.0106970214843747
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,1,power_law_1.01,3.6189727783203125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,1,power_law_1.01,4.97674560546875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,1,power_law_1.01,6.42733154296875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,2688,22,512,4,1,power_law_1.01,8.258156738281249
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,4,1,power_law_1.01,0.11276479721069337
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,4,1,power_law_1.01,0.112805118560791
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,4,1,power_law_1.01,0.11764479637145997
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,4,1,power_law_1.01,0.12657664299011231
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,4,1,power_law_1.01,0.13444479942321777
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,4,1,power_law_1.01,0.11248831748962403
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,4,1,power_law_1.01,0.14345343589782716
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,4,1,power_law_1.01,0.1515123176574707
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,4,1,power_law_1.01,0.17501440048217773
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,4,1,power_law_1.01,0.15890560150146485
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,4,1,power_law_1.01,0.11416255950927734
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,4,1,power_law_1.01,0.21313407897949216
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,4,1,power_law_1.01,0.22800575256347658
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,4,1,power_law_1.01,0.25944704055786133
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,4,1,power_law_1.01,0.3178009605407715
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,4,1,power_law_1.01,0.3549401473999024
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,4,1,power_law_1.01,0.44167041778564453
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,4,1,power_law_1.01,0.615080337524414
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,4,1,power_law_1.01,0.7866560363769531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,4,1,power_law_1.01,1.1353094482421875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,4,1,power_law_1.01,1.4803251647949218
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,4,1,power_law_1.01,2.169810485839844
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,4,1,power_law_1.01,2.850348205566406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,4,1,power_law_1.01,4.232371215820312
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,1,power_law_1.2,0.10585087776184081
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,1,power_law_1.2,0.17445568084716795
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,4,1,power_law_1.01,5.594918212890625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,1,power_law_1.2,0.2863180732727051
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,1,power_law_1.2,0.5863948822021484
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,1,power_law_1.2,1.0083481597900392
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,1,power_law_1.2,1.088801956176758
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,1,power_law_1.2,1.1204946899414063
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,1,power_law_1.2,0.43104385375976567
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,1,power_law_1.2,1.1827359771728516
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,1,power_law_1.2,1.30334716796875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,1,power_law_1.2,1.2270015716552733
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,1,power_law_1.2,1.4185932922363282
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,1,power_law_1.2,1.4856166076660156
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,1,power_law_1.2,1.6910488891601563
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,1,power_law_1.2,1.8565151977539063
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,1,power_law_1.2,2.0450956726074216
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,1,power_law_1.2,2.4135110473632815
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,1,power_law_1.2,1.879996795654297
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,1,power_law_1.2,2.137132110595703
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,1,power_law_1.2,2.585260009765625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,1,power_law_1.2,2.9725531005859374
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,1,power_law_1.2,3.8153863525390626
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,1,power_law_1.2,4.700767211914062
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,1,power_law_1.2,6.491900634765625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,1,power_law_1.2,8.413391723632813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,1,power_law_1.2,12.192017822265624
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,1,power_law_1.2,0.07949120044708252
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,1,power_law_1.2,0.11861568450927736
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,1,power_law_1.2,0.19359935760498045
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,1,power_law_1.2,0.2800799942016602
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,1,power_law_1.2,0.38713409423828127
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,1,power_law_1.2,0.646030044555664
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,1,power_law_1.2,0.6957676696777344
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,1,power_law_1.2,0.7136211395263672
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,1,power_law_1.2,0.7462028503417969
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,1,power_law_1.2,0.767628173828125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,1,power_law_1.2,0.7869657897949219
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,1,power_law_1.2,0.8204914855957032
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,1,power_law_1.2,0.8352832031249999
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,1,power_law_1.2,0.8878189086914062
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,1,power_law_1.2,0.8885388946533203
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,1,power_law_1.2,0.9182054138183593
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,1,power_law_1.2,0.9723686218261719
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,1,power_law_1.2,1.1231756591796875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,1,power_law_1.2,15.529310302734377
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,1,power_law_1.2,1.2012576293945312
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,1,power_law_1.2,1.540115203857422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,1,power_law_1.2,1.8889645385742189
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,1,power_law_1.2,1.3682496643066406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,1,power_law_1.2,2.2301402282714844
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,1,power_law_1.2,2.9201779174804687
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,1,power_law_1.2,3.6604312133789065
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,1,power_law_1.2,5.042826843261719
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,4,1,power_law_1.2,0.11311231613159181
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,4,1,power_law_1.2,0.11281791687011719
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,4,1,power_law_1.2,0.11293824195861817
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,4,1,power_law_1.2,0.11346943855285645
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,4,1,power_law_1.2,0.11765760421752929
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,4,1,power_law_1.2,0.1262828826904297
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,2688,22,512,4,1,power_law_1.2,19.26326904296875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,4,1,power_law_1.2,0.13447423934936525
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,4,1,power_law_1.2,0.15045120239257812
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,4,1,power_law_1.2,0.15917440414428713
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,4,1,power_law_1.2,0.17400127410888672
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,1,power_law_1.2,6.449158935546874
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,4,1,power_law_1.2,0.21827327728271487
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,4,1,power_law_1.2,0.22689151763916016
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,4,1,power_law_1.2,0.2681017684936523
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,4,1,power_law_1.2,0.35537281036376955
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,4,1,power_law_1.2,0.14154687881469727
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,2688,22,512,4,1,power_law_1.2,8.124884033203125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,4,1,power_law_1.2,0.44429759979248046
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,4,1,power_law_1.2,0.3179667282104492
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,4,1,power_law_1.2,1.1360217285156249
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,4,1,power_law_1.2,0.613267822265625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,4,1,power_law_1.2,1.4809893798828124
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,4,1,power_law_1.2,0.7858207702636719
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,4,1,power_law_1.2,2.176260528564453
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,4,1,power_law_1.2,2.859749755859375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,4,1,power_law_1.2,4.224410095214844
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,4,1,power_law_1.2,5.597164306640625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,8,1,power_law_1.01,0.07455872058868408
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,8,1,power_law_1.01,0.07203264236450195
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,8,1,power_law_1.01,0.08683967590332031
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,8,1,power_law_1.01,0.11988224029541014
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,8,1,power_law_1.01,0.12112959861755371
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,8,1,power_law_1.01,0.12219391822814943
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,8,1,power_law_1.01,0.12213248252868651
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,8,1,power_law_1.01,0.12281791687011719
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,8,1,power_law_1.01,0.12421440124511718
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,8,1,power_law_1.01,0.12783040046691893
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,8,1,power_law_1.01,0.12940095901489257
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,8,1,power_law_1.01,0.1341766357421875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,8,1,power_law_1.01,0.1407430362701416
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,8,1,power_law_1.01,0.1456441593170166
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,8,1,power_law_1.01,0.17392959594726562
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,8,1,power_law_1.01,0.21116672515869142
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,8,1,power_law_1.01,0.2442188835144043
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,8,1,power_law_1.01,0.29009408950805665
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,8,1,power_law_1.01,0.37625473022460937
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,8,1,power_law_1.01,0.46177726745605463
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,8,1,power_law_1.01,0.6419865417480468
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,8,1,power_law_1.01,0.8181305694580079
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,8,1,power_law_1.01,1.1760428619384764
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,8,1,power_law_1.01,1.5749472045898438
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,8,1,power_law_1.01,2.3368006896972657
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,8,1,power_law_1.01,0.0370086407661438
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,8,1,power_law_1.01,0.04440383911132813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,8,1,power_law_1.01,0.04801727771759033
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,8,1,power_law_1.01,0.05933440208435059
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,8,1,power_law_1.01,0.06834432125091552
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,8,1,power_law_1.01,0.06363903999328613
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,8,1,power_law_1.01,0.06881343841552734
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,8,1,power_law_1.01,0.06459904193878174
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,8,1,power_law_1.01,0.06871679782867432
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,8,1,power_law_1.01,0.06616767883300781
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,8,1,power_law_1.01,0.07158783912658692
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,8,1,power_law_1.01,0.06785664081573486
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,8,1,power_law_1.01,0.07392064094543457
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,8,1,power_law_1.01,0.07713088035583496
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,8,1,power_law_1.01,0.08832063674926757
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,8,1,power_law_1.01,0.09470272064208984
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,8,1,power_law_1.01,0.09840255737304687
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,8,1,power_law_1.01,0.10281855583190917
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,8,1,power_law_1.01,0.12476287841796876
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,8,1,power_law_1.01,0.15308671951293945
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,8,1,power_law_1.01,0.24473087310791017
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,8,1,power_law_1.01,0.1797977638244629
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,8,1,power_law_1.01,0.318525447845459
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,8,1,power_law_1.01,0.4412403106689453
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,8,1,power_law_1.01,3.8925439453125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,8,1,power_law_1.01,0.5733574295043946
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,8,1,power_law_1.01,0.8219993591308594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,8,1,power_law_1.01,1.0687615966796875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,8,1,power_law_1.01,0.019123200178146362
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,8,1,power_law_1.01,0.03224767923355103
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,8,1,power_law_1.01,0.03348864078521728
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,8,1,power_law_1.01,0.03523904085159302
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,8,1,power_law_1.01,0.03532799959182739
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,8,1,power_law_1.01,0.035385599136352544
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,8,1,power_law_1.01,0.035436799526214605
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,8,1,power_law_1.01,0.03611776113510132
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,8,1,power_law_1.01,0.023457920551300047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,8,1,power_law_1.01,0.03611200094223023
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,8,1,power_law_1.01,1.3153427124023438
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,8,1,power_law_1.01,0.025610239505767823
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,8,1,power_law_1.01,0.03788160085678101
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,8,1,power_law_1.01,0.03660351991653442
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,8,1,power_law_1.01,0.04130943775177002
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,8,1,power_law_1.01,0.04003456115722656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,8,1,power_law_1.01,0.044405121803283695
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,8,1,power_law_1.01,0.04826432228088379
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,8,1,power_law_1.01,0.05711679935455323
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,8,1,power_law_1.01,0.06633408069610595
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,8,1,power_law_1.01,0.08503680229187012
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,8,1,power_law_1.01,0.10313471794128418
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,8,1,power_law_1.01,0.14049983978271485
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,8,1,power_law_1.01,0.18125120162963868
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,8,1,power_law_1.01,0.2611935997009277
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,8,1,power_law_1.01,0.34445121765136716
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,8,1,power_law_1.2,0.05814911842346191
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,8,1,power_law_1.2,0.06840064048767089
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,8,1,power_law_1.2,0.07047039985656738
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,8,1,power_law_1.2,0.08861568450927734
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,8,1,power_law_1.2,0.11959424018859863
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,8,1,power_law_1.2,0.1214732837677002
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,8,1,power_law_1.2,0.12231424331665039
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,8,1,power_law_1.2,0.12266880035400392
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,8,1,power_law_1.2,0.12365056037902833
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,8,1,power_law_1.2,0.12486592292785645
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,8,1,power_law_1.2,0.12888383865356445
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,8,1,power_law_1.2,0.1301375961303711
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,8,1,power_law_1.2,0.1342355155944824
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,8,1,power_law_1.2,0.1419264030456543
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,8,1,power_law_1.2,0.14580544471740722
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,8,1,power_law_1.2,0.18341247558593748
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,8,1,power_law_1.2,0.21295743942260742
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,8,1,power_law_1.01,0.5027302551269531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,8,1,power_law_1.2,0.24541248321533202
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,8,1,power_law_1.2,0.2940518379211426
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,8,1,power_law_1.2,0.38097854614257814
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,8,1,power_law_1.2,0.4623558425903321
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,8,1,power_law_1.2,0.6453421020507812
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,8,1,power_law_1.01,0.6682118225097656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,8,1,power_law_1.2,0.816542739868164
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,8,1,power_law_1.2,1.1722994995117186
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,8,1,power_law_1.2,1.5367109680175781
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,8,1,power_law_1.01,0.8240620422363282
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,8,1,power_law_1.2,0.03983295917510986
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,8,1,power_law_1.2,0.04678336143493652
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,8,1,power_law_1.2,0.04841279983520508
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,8,1,power_law_1.2,0.05979904174804688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,8,1,power_law_1.2,0.061820158958435054
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,8,1,power_law_1.2,0.0634988784790039
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,8,1,power_law_1.2,0.06478208065032959
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,8,1,power_law_1.2,0.06561600208282471
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,8,1,power_law_1.2,0.06612736225128174
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,8,1,power_law_1.2,0.06648704051971435
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,8,1,power_law_1.2,0.06772287845611571
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,8,1,power_law_1.2,0.06892672061920166
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,8,1,power_law_1.2,0.06986623764038086
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,8,1,power_law_1.2,0.0779641580581665
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,8,1,power_law_1.2,0.08654784202575684
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,8,1,power_law_1.2,0.09763135910034179
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,8,1,power_law_1.2,0.09516032218933106
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,8,1,power_law_1.2,0.10201408386230468
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,8,1,power_law_1.2,0.12260864257812501
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,8,1,power_law_1.2,0.14812416076660156
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,8,1,power_law_1.2,2.349426574707031
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,8,1,power_law_1.2,0.18284608840942382
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,8,1,power_law_1.2,0.2467718315124512
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,8,1,power_law_1.2,0.31618623733520507
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,8,1,power_law_1.2,0.44582401275634764
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,8,1,power_law_1.2,3.0181658935546873
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,8,1,power_law_1.2,0.5691961669921876
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,8,1,power_law_1.2,3.7469415283203125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,8,1,power_law_1.2,0.8254592132568359
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,8,1,power_law_1.2,0.023508479595184324
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,8,1,power_law_1.2,0.024333438873291015
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,8,1,power_law_1.2,0.034454400539398196
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,8,1,power_law_1.2,0.03415679931640625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,8,1,power_law_1.2,0.03647808074951172
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,8,1,power_law_1.2,0.036145920753479006
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,8,1,power_law_1.2,0.03581183910369873
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,8,1,power_law_1.2,0.019433599710464478
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,8,1,power_law_1.2,0.0366483211517334
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,8,1,power_law_1.2,0.037309439182281495
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,8,1,power_law_1.2,0.03722815990447998
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,8,1,power_law_1.2,0.03746687889099121
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,8,1,power_law_1.2,0.03986047983169556
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,8,1,power_law_1.2,1.0725202941894532
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,8,1,power_law_1.2,0.037292160987854005
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,8,1,power_law_1.2,0.04238143920898437
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,8,1,power_law_1.2,1.3067277526855468
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,8,1,power_law_1.2,0.04607615947723388
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,8,1,power_law_1.2,0.04854527950286865
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,8,1,power_law_1.2,0.058348798751831056
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,8,1,power_law_1.2,0.06697792053222656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,8,1,power_law_1.2,0.08616767883300781
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,8,1,power_law_1.2,0.10462528228759767
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,8,1,power_law_1.2,0.14713088035583496
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,8,1,power_law_1.2,0.18595903396606445
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,8,1,power_law_1.2,0.27290687561035154
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,8,1,power_law_1.2,0.3544857788085937
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,8,1,power_law_1.01,0.07719615936279298
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,8,1,power_law_1.01,0.0920019245147705
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,8,1,power_law_1.01,0.09880127906799316
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,8,1,power_law_1.01,0.1280787181854248
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,8,1,power_law_1.01,0.1644767951965332
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,8,1,power_law_1.01,0.1662841606140137
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,8,1,power_law_1.01,0.16759103775024414
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,8,1,power_law_1.01,0.16713600158691405
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,8,1,power_law_1.01,0.16786367416381837
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,8,1,power_law_1.01,0.17143295288085938
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,8,1,power_law_1.01,0.1814873504638672
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,8,1,power_law_1.01,0.1905779266357422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,8,1,power_law_1.01,0.206167049407959
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,8,1,power_law_1.01,0.22186624526977541
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,8,1,power_law_1.01,0.23417984008789064
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,8,1,power_law_1.01,0.28375232696533204
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,8,1,power_law_1.01,0.3140755271911621
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,8,1,power_law_1.01,0.3815750503540039
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,8,1,power_law_1.01,0.4535526275634766
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,8,1,power_law_1.01,0.601333122253418
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,8,1,power_law_1.2,0.5221926498413085
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,8,1,power_law_1.01,0.7488620758056641
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,8,1,power_law_1.01,1.0456902313232423
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,8,1,power_law_1.01,1.3526354980468749
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,8,1,power_law_1.01,1.9924467468261717
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,8,1,power_law_1.01,2.6526470947265626
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,8,1,power_law_1.01,3.944721984863281
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,8,1,power_law_1.2,0.6867295837402343
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,8,1,power_law_1.01,0.05100927829742432
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,8,1,power_law_1.01,0.0594432020187378
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,8,1,power_law_1.01,0.06316160202026368
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,8,1,power_law_1.01,0.0838047981262207
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,8,1,power_law_1.01,0.09045568466186524
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,8,1,power_law_1.01,0.09079360008239747
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,8,1,power_law_1.01,0.09187007904052734
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,8,1,power_law_1.01,0.09229503631591797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,8,1,power_law_1.01,0.09236224174499512
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,8,1,power_law_1.01,0.09374527931213379
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,8,1,power_law_1.01,0.09529536247253419
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,8,1,power_law_1.01,0.09579839706420898
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,8,1,power_law_1.01,0.09892992019653321
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,8,1,power_law_1.01,0.11198719978332519
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,8,1,power_law_1.01,0.11752703666687012
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,8,1,power_law_1.01,0.11933888435363768
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,8,1,power_law_1.01,0.12689791679382326
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,8,1,power_law_1.01,0.1431116771697998
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,8,1,power_law_1.01,0.17870271682739258
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,8,1,power_law_1.01,0.22571008682250976
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,8,1,power_law_1.01,0.2595052719116211
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,8,1,power_law_1.01,5.393845825195312
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,8,1,power_law_1.01,0.3619801712036133
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,8,1,power_law_1.01,0.459986572265625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,8,1,power_law_1.01,0.6434745788574219
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,8,1,power_law_1.2,0.8601721954345702
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,8,1,power_law_1.01,0.8446208190917968
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,8,1,power_law_1.01,6.694886474609374
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,8,1,power_law_1.01,1.2229222106933595
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,8,1,power_law_1.01,0.02526144027709961
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,8,1,power_law_1.01,1.674784698486328
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,8,1,power_law_1.01,0.03873087882995606
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,8,1,power_law_1.01,0.04730175971984864
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,8,1,power_law_1.01,0.04804543972015381
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,8,1,power_law_1.01,0.04805312156677246
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,8,1,power_law_1.01,0.04777472019195557
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,8,1,power_law_1.01,0.047800321578979496
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,8,1,power_law_1.01,0.04884031772613525
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,8,1,power_law_1.01,0.04925759792327881
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,8,1,power_law_1.01,0.03130880117416382
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,8,1,power_law_1.01,2.091905975341797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,8,1,power_law_1.01,0.050880641937255854
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,8,1,power_law_1.01,0.0475436782836914
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,8,1,power_law_1.01,0.05136640071868896
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,8,1,power_law_1.01,0.05649536132812501
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,8,1,power_law_1.01,0.05904704093933105
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,8,1,power_law_1.01,0.06575359821319579
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,8,1,power_law_1.01,0.08059136390686035
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,8,1,power_law_1.01,0.06029376029968262
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,8,1,power_law_1.01,0.09585984230041504
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,8,1,power_law_1.01,0.12185471534729005
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,8,1,power_law_1.01,0.1489510440826416
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,8,1,power_law_1.01,0.21018943786621094
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,8,1,power_law_1.01,0.27050048828125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,8,1,power_law_1.01,0.3882304000854492
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,8,1,power_law_1.01,0.508416633605957
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,8,1,power_law_1.2,0.07782336235046386
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,8,1,power_law_1.2,0.09056320190429687
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,8,1,power_law_1.2,0.09393471717834473
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,8,1,power_law_1.2,0.13011263847351073
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,8,1,power_law_1.2,0.16223743438720703
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,8,1,power_law_1.2,0.16509759902954102
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,8,1,power_law_1.2,0.166309757232666
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,8,1,power_law_1.2,0.16671680450439452
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,8,1,power_law_1.2,0.16797119140625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,8,1,power_law_1.2,0.1705183982849121
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,8,1,power_law_1.2,0.17976768493652345
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,8,1,power_law_1.2,0.19211519241333008
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,8,1,power_law_1.2,0.20279872894287107
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,8,1,power_law_1.2,0.22218559265136717
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,8,1,power_law_1.2,0.23546560287475585
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,8,1,power_law_1.2,0.2965702438354492
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,8,1,power_law_1.2,0.3115129661560059
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,8,1,power_law_1.2,0.38144256591796877
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,8,1,power_law_1.2,0.45826942443847657
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,8,1,power_law_1.2,0.6050912094116211
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,8,1,power_law_1.2,0.7453369903564453
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,8,1,power_law_1.2,1.0473190307617188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,8,1,power_law_1.01,1.0014105224609375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,8,1,power_law_1.01,0.7558598327636719
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,8,1,power_law_1.2,1.3618861389160155
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,8,1,power_law_1.01,1.2428678131103514
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,8,1,power_law_1.2,1.964378204345703
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,8,1,power_law_1.2,2.662034606933594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,8,1,power_law_1.2,0.05143424034118652
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,8,1,power_law_1.2,0.058613758087158206
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,8,1,power_law_1.2,0.06085055828094482
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,8,1,power_law_1.2,0.08381183624267578
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,8,1,power_law_1.2,0.08806528091430664
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,8,1,power_law_1.2,0.09071295738220216
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,8,1,power_law_1.2,0.09167488098144531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,8,1,power_law_1.2,0.09148351669311523
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,8,1,power_law_1.2,0.09307968139648437
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,8,1,power_law_1.2,0.0936793613433838
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,8,1,power_law_1.2,0.09508480072021484
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,8,1,power_law_1.2,0.09670463562011719
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,8,1,power_law_1.2,0.09760191917419433
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,8,1,power_law_1.2,3.9886712646484375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,8,1,power_law_1.2,0.11276672363281251
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,8,1,power_law_1.2,0.12278143882751465
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,8,1,power_law_1.2,0.12374655723571777
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,8,1,power_law_1.2,0.13177408218383788
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,8,1,power_law_1.2,0.1798316764831543
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,8,1,power_law_1.2,0.1435807991027832
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,8,1,power_law_1.2,0.21669952392578126
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,8,1,power_law_1.2,0.26204160690307615
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,8,1,power_law_1.2,0.3635513687133789
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,8,1,power_law_1.2,0.46231231689453123
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,8,1,power_law_1.2,5.17682373046875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,8,1,power_law_1.2,0.6463680267333984
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,8,1,power_law_1.2,6.777063598632813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,8,1,power_law_1.2,0.8434815979003906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,8,1,power_law_1.2,0.02476032018661499
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,8,1,power_law_1.2,0.03024512052536011
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,8,1,power_law_1.2,0.03332607984542847
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,8,1,power_law_1.2,0.046637439727783205
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,8,1,power_law_1.2,0.04691904067993164
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,8,1,power_law_1.2,0.047895040512084965
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,8,1,power_law_1.2,0.047887358665466305
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,8,1,power_law_1.2,0.047619199752807616
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,8,1,power_law_1.2,0.04792511940002441
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,8,1,power_law_1.2,0.04844927787780762
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,8,1,power_law_1.2,0.04889344215393067
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,8,1,power_law_1.2,1.2624742126464843
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,8,1,power_law_1.2,0.051431040763854984
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,8,1,power_law_1.2,0.05647232055664062
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,8,1,power_law_1.2,1.73233154296875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,8,1,power_law_1.2,0.05094016075134278
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,8,1,power_law_1.2,0.0606220817565918
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,8,1,power_law_1.2,0.061915521621704106
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,8,1,power_law_1.2,2.068694458007813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,8,1,power_law_1.2,0.08126655578613282
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,8,1,power_law_1.2,0.06718016147613526
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,8,1,power_law_1.2,0.21673088073730468
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,8,1,power_law_1.2,0.09579008102416993
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,8,1,power_law_1.2,0.12467071533203125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,8,1,power_law_1.2,0.15225536346435548
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,8,1,power_law_1.2,0.27630271911621096
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,8,1,power_law_1.2,0.4013478469848633
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,8,1,power_law_1.2,0.5283475112915039
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,1,power_law_1.01,0.06589888095855713
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,1,power_law_1.01,0.08454976081848145
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,1,power_law_1.01,0.10584447860717774
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,1,power_law_1.01,0.14999232292175294
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,1,power_law_1.01,0.22193983078002927
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,1,power_law_1.01,0.3162528038024902
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,1,power_law_1.01,0.39501888275146485
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,1,power_law_1.01,0.40695873260498044
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,1,power_law_1.01,0.41439041137695315
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,1,power_law_1.01,0.41829761505126956
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,1,power_law_1.01,0.4414112091064453
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,1,power_law_1.01,0.4763558578491211
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,1,power_law_1.01,0.49720062255859376
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,1,power_law_1.01,0.5289977645874023
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,1,power_law_1.01,0.5726969528198242
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,1,power_law_1.01,0.6179481506347656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,1,power_law_1.01,0.6804608154296875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,1,power_law_1.01,0.7170425415039062
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,1,power_law_1.01,0.7543878173828125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,1,power_law_1.01,0.8421727752685546
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,1,power_law_1.01,0.9788800048828126
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,1,power_law_1.01,1.1883033752441405
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,1,power_law_1.01,1.4304953002929688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,8,1,power_law_1.2,0.769675521850586
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,1,power_law_1.01,1.9032672119140623
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,1,power_law_1.01,2.3668115234375002
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,1,power_law_1.01,3.3219793701171874
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,1,power_law_1.01,4.263358154296875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,1,power_law_1.01,0.04725696086883545
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,1,power_law_1.01,0.05521215915679931
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,8,1,power_law_1.01,5.2336621093749995
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,1,power_law_1.01,0.083372802734375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,1,power_law_1.01,0.12523839950561524
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,1,power_law_1.01,0.19111679077148438
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,1,power_law_1.01,0.25608896255493163
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,1,power_law_1.01,0.3231302261352539
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,1,power_law_1.01,0.33329727172851564
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,1,power_law_1.01,0.3384288024902344
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,1,power_law_1.01,0.34233856201171875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,1,power_law_1.01,0.35467712402343754
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,1,power_law_1.01,0.3650035095214844
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,1,power_law_1.01,0.36994369506835934
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,1,power_law_1.01,0.3785696029663086
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,1,power_law_1.01,0.4272281646728516
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,1,power_law_1.01,0.4332729721069336
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,1,power_law_1.01,0.45423553466796873
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,1,power_law_1.01,0.48692672729492187
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,1,power_law_1.01,0.5138272094726563
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,1,power_law_1.01,0.5684422302246094
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,1,power_law_1.01,0.6367225646972656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,1,power_law_1.01,0.75468994140625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,1,power_law_1.01,0.8886943817138672
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,1,power_law_1.01,1.156803207397461
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,1,power_law_1.01,1.4377658081054687
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,1,power_law_1.01,1.9683596801757812
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,1,power_law_1.01,2.4832896423339843
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,8,1,power_law_1.01,3.0328799438476564
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,8,1,power_law_1.2,1.0310822296142579
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,8,1,power_law_1.01,0.06329023838043213
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,8,1,power_law_1.01,0.10311936378479003
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,8,1,power_law_1.01,0.1283091163635254
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,8,1,power_law_1.01,0.1520672035217285
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,8,1,power_law_1.01,0.19230783462524415
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,8,1,power_law_1.01,0.06856768131256104
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,8,1,power_law_1.01,0.20141632080078126
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,8,1,power_law_1.2,1.261333770751953
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,8,1,power_law_1.01,0.20902271270751954
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,8,1,power_law_1.01,0.2105036735534668
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,8,1,power_law_1.01,0.2278879928588867
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,8,1,power_law_1.01,0.231856632232666
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,8,1,power_law_1.01,0.24212223052978515
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,8,1,power_law_1.01,0.25441024780273436
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,8,1,power_law_1.01,0.2637171173095703
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,8,1,power_law_1.01,0.2843391990661621
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,8,1,power_law_1.01,0.0825260829925537
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,8,1,power_law_1.01,0.21913024902343753
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,8,1,power_law_1.01,0.3737881469726562
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,8,1,power_law_1.01,0.3283891296386719
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,8,1,power_law_1.01,0.5686592102050781
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,8,1,power_law_1.01,0.475676155090332
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,8,1,power_law_1.01,0.764532470703125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,8,1,power_law_1.01,0.9946361541748047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,8,1,power_law_1.01,1.4614617919921875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,8,1,power_law_1.01,2.853102722167969
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,1,power_law_1.2,0.07011199951171875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,1,power_law_1.2,0.08486528396606445
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,1,power_law_1.2,0.10506879806518554
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,1,power_law_1.2,0.14210304260253906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,1,power_law_1.2,0.20459327697753907
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,1,power_law_1.2,0.29668672561645504
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,1,power_law_1.2,0.38534782409667967
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,8,1,power_law_1.01,1.913561553955078
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,1,power_law_1.2,0.3997721481323242
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,8,1,power_law_1.01,3.7444882202148433
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,1,power_law_1.2,0.4091449737548828
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,1,power_law_1.2,0.4213216018676758
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,1,power_law_1.2,0.4443027114868164
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,1,power_law_1.2,0.47970497131347656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,1,power_law_1.2,0.5015731048583985
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,1,power_law_1.2,0.5439686584472656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,1,power_law_1.2,0.6121343994140626
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,1,power_law_1.2,0.6563814544677735
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,1,power_law_1.2,0.6852339172363282
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,1,power_law_1.2,0.7247212982177734
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,1,power_law_1.2,0.7612019348144531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,1,power_law_1.2,0.8756684875488281
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,1,power_law_1.2,0.9904032135009766
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,1,power_law_1.2,1.2191391754150391
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,1,power_law_1.2,1.4531430053710939
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,1,power_law_1.2,1.9127693176269531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,1,power_law_1.2,2.3922412109374998
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,1,power_law_1.2,3.355047607421875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,1,power_law_1.2,4.293629455566406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,1,power_law_1.2,0.04690432071685791
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,1,power_law_1.2,0.054976000785827636
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,1,power_law_1.2,0.11962431907653807
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,1,power_law_1.2,0.1836140823364258
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,8,1,power_law_1.2,5.263970947265625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,1,power_law_1.2,0.24724992752075195
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,1,power_law_1.2,0.08457728385925292
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,1,power_law_1.2,0.319804801940918
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,1,power_law_1.2,0.3291603088378906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,1,power_law_1.2,0.33522945404052734
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,1,power_law_1.2,0.342872314453125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,1,power_law_1.2,0.35213504791259764
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,1,power_law_1.2,0.36212417602539065
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,1,power_law_1.2,0.3686470413208008
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,1,power_law_1.2,0.38658241271972654
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,1,power_law_1.2,0.43404479980468746
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,1,power_law_1.2,0.4482892990112305
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,1,power_law_1.2,0.4646540832519531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,1,power_law_1.2,0.501971206665039
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,7168,2048,8,256,8,1,power_law_1.01,4.585073852539063
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,1,power_law_1.2,0.5281919860839844
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,1,power_law_1.2,0.5927513504028321
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,1,power_law_1.2,0.6582323455810547
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,1,power_law_1.2,0.7922041320800781
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,1,power_law_1.2,0.9313401794433593
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,1,power_law_1.2,1.1727021026611328
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,1,power_law_1.2,1.4601593017578125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,8,1,power_law_1.2,0.06280831813812256
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,1,power_law_1.2,1.981922607421875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,1,power_law_1.2,2.526769866943359
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,8,1,power_law_1.2,0.07047808170318604
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,8,1,power_law_1.2,0.07992959976196289
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,8,1,power_law_1.2,0.09970239639282227
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,8,1,power_law_1.2,3.042111511230469
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,8,1,power_law_1.2,0.12246015548706055
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,8,1,power_law_1.2,0.14666751861572264
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,8,1,power_law_1.2,0.19110591888427736
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,8,1,power_law_1.2,0.19862207412719726
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,8,1,power_law_1.2,0.20574720382690428
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,8,1,power_law_1.2,0.21098304748535157
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,8,1,power_law_1.2,0.22094783782958985
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,8,1,power_law_1.2,0.2298099136352539
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,8,1,power_law_1.2,0.23709440231323242
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,8,1,power_law_1.2,0.2529030418395996
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,8,1,power_law_1.2,0.2673151969909668
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,8,1,power_law_1.2,0.28031423568725583
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,8,1,power_law_1.2,0.3785907363891602
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,8,1,power_law_1.2,0.44460990905761716
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,8,1,power_law_1.2,0.3170835113525391
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,8,1,power_law_1.2,0.6086310577392579
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,8,1,power_law_1.2,0.7615814208984375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,8,1,power_law_1.2,1.1069087982177734
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,8,1,power_law_1.2,2.165333709716797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,8,1,power_law_1.2,1.4508146667480468
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,1,power_law_1.01,0.032934401035308834
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,8,1,power_law_1.01,0.06663296222686768
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,1,power_law_1.01,0.07752895832061768
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,1,power_law_1.01,0.09373760223388672
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,1,power_law_1.01,0.12656448364257813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,1,power_law_1.01,0.12533503532409668
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,1,power_law_1.01,0.13039487838745117
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,1,power_law_1.01,0.13503616333007812
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,1,power_law_1.01,0.13888959884643554
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,1,power_law_1.01,0.14981439590454101
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,1,power_law_1.01,0.16764991760253906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,1,power_law_1.01,0.1709676742553711
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,1,power_law_1.01,0.1780147171020508
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,1,power_law_1.01,0.1826963233947754
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,1,power_law_1.01,0.1872467231750488
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,1,power_law_1.01,0.19883840560913085
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,1,power_law_1.01,0.23371456146240233
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,1,power_law_1.01,0.28651391983032226
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,1,power_law_1.01,0.332210578918457
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,1,power_law_1.01,0.3945945739746094
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,1,power_law_1.01,0.5187430572509766
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,1,power_law_1.01,0.6403308868408203
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,1,power_law_1.01,0.8879481506347655
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,1,power_law_1.01,1.132552947998047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,8,1,power_law_1.2,2.992794189453125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,1,power_law_1.01,1.6250956726074217
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,1,power_law_1.01,2.104646453857422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,8,1,power_law_1.01,2.6037374877929684
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,1,power_law_1.01,0.037496960163116454
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,1,power_law_1.01,0.04442880153656006
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,1,power_law_1.01,0.058764162063598635
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,1,power_law_1.01,0.07305984020233154
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,1,power_law_1.01,0.09560383796691894
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,1,power_law_1.01,0.12969663619995117
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,1,power_law_1.01,0.1298745632171631
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,1,power_law_1.01,0.13378623962402342
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,1,power_law_1.01,0.1322822380065918
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,1,power_law_1.01,0.13632384300231934
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,1,power_law_1.01,0.13722304344177244
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,1,power_law_1.01,0.15443455696105957
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,1,power_law_1.01,0.15631487846374512
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,1,power_law_1.01,0.16280319213867187
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,1,power_law_1.01,0.16415103912353515
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,1,power_law_1.01,0.16751039505004883
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,1,power_law_1.01,0.17572544097900392
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,1,power_law_1.01,0.19256704330444335
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,1,power_law_1.01,0.21747648239135745
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,1,power_law_1.01,0.2521785545349121
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,1,power_law_1.01,0.2954668807983398
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,1,power_law_1.01,0.3749542236328125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,1,power_law_1.01,0.45341056823730475
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,1,power_law_1.01,0.621822738647461
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,1,power_law_1.01,0.7756761932373047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,1,power_law_1.01,1.113235855102539
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,1,power_law_1.01,1.4312889099121093
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,8,1,power_law_1.01,1.7621849060058594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,8,1,power_law_1.01,0.042558717727661136
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,8,1,power_law_1.01,0.05060607910156249
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,8,1,power_law_1.01,0.057934718132019045
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,8,1,power_law_1.01,0.06593023777008057
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,8,1,power_law_1.01,0.08183487892150879
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,8,1,power_law_1.01,0.08580096244812012
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,8,1,power_law_1.01,0.04576767921447754
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,8,1,power_law_1.01,0.08783552169799805
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,8,1,power_law_1.01,0.09059200286865235
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,8,1,power_law_1.01,0.0943609619140625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,8,1,power_law_1.01,0.09752256393432616
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,8,1,power_law_1.01,0.09892160415649413
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,8,1,power_law_1.01,0.1060371208190918
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,8,1,power_law_1.01,0.08978303909301757
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,8,1,power_law_1.01,0.11113727569580079
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,8,1,power_law_1.01,0.11734656333923339
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,8,1,power_law_1.01,0.12763968467712403
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,8,1,power_law_1.01,0.15574527740478517
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,8,1,power_law_1.01,0.1823321533203125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,8,1,power_law_1.01,0.2446009635925293
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,8,1,power_law_1.01,0.30269952774047854
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,8,1,power_law_1.01,0.424194564819336
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,8,1,power_law_1.01,0.5508512115478515
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,8,1,power_law_1.01,1.0860749053955079
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,8,1,power_law_1.01,0.8222035217285157
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,1,power_law_1.2,0.033780479431152345
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,8,1,power_law_1.2,0.05289472103118896
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,8,1,power_law_1.2,0.06479680061340332
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,1,power_law_1.2,0.07485055923461914
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,1,power_law_1.2,0.08979264259338379
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,1,power_law_1.2,0.12958144187927245
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,1,power_law_1.2,0.12776512145996094
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,1,power_law_1.2,0.13422271728515625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,8,1,power_law_1.01,1.6496275329589842
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,1,power_law_1.2,0.13956671714782715
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,1,power_law_1.2,0.1515129566192627
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,8,1,power_law_1.01,2.223132781982422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,1,power_law_1.2,0.1688627243041992
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,1,power_law_1.2,0.1738547134399414
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,1,power_law_1.2,0.17945087432861329
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,1,power_law_1.2,0.18439359664916993
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,1,power_law_1.2,0.18890432357788084
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,1,power_law_1.2,0.200762882232666
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,1,power_law_1.2,0.14135616302490234
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,1,power_law_1.2,0.24578752517700195
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,1,power_law_1.2,0.28656448364257814
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,1,power_law_1.2,0.3341657638549805
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,1,power_law_1.2,0.39547134399414063
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,1,power_law_1.2,0.5153343963623047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,1,power_law_1.2,0.6445081329345703
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,1,power_law_1.2,0.8882924652099609
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,1536,8,128,8,1,power_law_1.01,2.8000881958007815
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,1,power_law_1.2,1.1305535888671874
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,1,power_law_1.2,0.037407360076904296
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,1,power_law_1.2,0.046277761459350586
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,1,power_law_1.2,0.05706816196441651
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,1,power_law_1.2,0.06875648021697997
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,1,power_law_1.2,0.09144512176513672
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,1,power_law_1.2,0.12904767990112304
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,1,power_law_1.2,0.12847423553466797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,1,power_law_1.2,0.13150912284851074
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,1,power_law_1.2,0.13404352188110352
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,1,power_law_1.2,0.13457152366638184
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,1,power_law_1.2,0.13677696228027342
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,1,power_law_1.2,0.15626751899719238
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,1,power_law_1.2,0.1571782398223877
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,1,power_law_1.2,0.1629599952697754
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,1,power_law_1.2,0.16619840621948243
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,1,power_law_1.2,0.16874048233032227
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,1,power_law_1.2,0.17772544860839842
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,1,power_law_1.2,0.19543104171752929
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,1,power_law_1.2,0.2225472068786621
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,1,power_law_1.2,0.25947711944580076
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,1,power_law_1.2,0.3002943992614746
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,1,power_law_1.2,0.37545536041259764
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,1,power_law_1.2,0.4586233520507813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,1,power_law_1.2,1.6286207580566405
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,1,power_law_1.2,0.6223155212402344
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,1,power_law_1.2,0.7814118194580078
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,1,power_law_1.2,1.1100985717773437
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,1,power_law_1.2,2.1189573669433592
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,8,1,power_law_1.2,0.04279615879058838
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,8,1,power_law_1.2,0.04650752067565918
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,8,1,power_law_1.2,0.05049791812896729
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,8,1,power_law_1.2,0.056384639739990236
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,8,1,power_law_1.2,0.06403776168823241
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,8,1,power_law_1.2,2.6047647094726565
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,1,power_law_1.2,1.4382540893554687
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,8,1,power_law_1.2,0.08344127655029296
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,8,1,power_law_1.2,0.08950719833374024
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,8,1,power_law_1.2,0.09151616096496581
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,8,1,power_law_1.2,0.0930361557006836
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,8,1,power_law_1.2,0.09627840042114258
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,8,1,power_law_1.2,0.10012415885925292
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,8,1,power_law_1.2,0.10364800453186036
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,8,1,power_law_1.2,0.11944255828857422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,8,1,power_law_1.2,0.12750656127929688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,8,1,power_law_1.2,0.08714879989624023
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,8,1,power_law_1.2,0.11061568260192871
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,8,1,power_law_1.2,0.14483776092529296
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,8,1,power_law_1.2,0.1838707160949707
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,8,1,power_law_1.2,0.22464576721191407
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,8,1,power_law_1.2,0.31287872314453125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,8,1,power_law_1.2,0.4038828659057617
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,8,1,power_law_1.2,0.5901939010620116
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,8,1,power_law_1.2,1.7587129211425783
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,8,1,power_law_1.2,1.1738540649414062
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,8,1,power_law_1.2,0.7919718170166016
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,8,1,power_law_1.2,1.5600025939941407
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,1,power_law_1.01,0.065797119140625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,1,power_law_1.01,0.08279359817504883
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,1,power_law_1.01,0.10438272476196289
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,1,power_law_1.01,0.1437772846221924
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,1,power_law_1.01,0.1909222412109375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,1,power_law_1.01,0.313690242767334
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,1,power_law_1.01,0.32421375274658204
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,1,power_law_1.01,0.331580810546875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,1,power_law_1.01,0.333647346496582
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,1,power_law_1.01,0.33576961517333986
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,1,power_law_1.01,0.3464352035522461
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,1,power_law_1.01,0.376165771484375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,1,power_law_1.01,0.40891456604003906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,1,power_law_1.01,0.42278209686279294
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,1,power_law_1.01,0.4274156951904297
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,1,power_law_1.01,0.4396819305419922
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,1,power_law_1.01,0.5039923095703125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,1,power_law_1.01,0.5591142272949219
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,1,power_law_1.01,0.5951359939575196
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,1,power_law_1.01,0.6909785461425781
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,1,power_law_1.01,0.8187564849853516
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,8,1,power_law_1.2,2.3266009521484374
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,1,power_law_1.01,1.0517676544189452
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,1,power_law_1.01,1.2832633972167968
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,1,power_law_1.01,1.7730975341796875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,1,power_law_1.01,2.2680844116210936
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,1,power_law_1.01,3.2373889160156253
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,1,power_law_1.01,4.218985595703125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,1,power_law_1.01,0.04723519802093506
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,1,power_law_1.01,0.062037119865417475
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,1,power_law_1.01,0.08571136474609375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,8,1,power_law_1.01,5.164251098632812
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,1,power_law_1.01,0.11486335754394532
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,1,power_law_1.01,0.23776960372924805
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,1,power_law_1.01,0.2386412811279297
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,1,power_law_1.01,0.24391807556152342
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,1,power_law_1.01,0.24956031799316408
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,1,power_law_1.01,0.164516487121582
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,1,power_law_1.01,0.2522444725036621
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,1,power_law_1.01,0.25676992416381833
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,1,power_law_1.01,0.26375423431396483
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,1,power_law_1.01,0.3045715141296387
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,1,power_law_1.01,0.3195347213745117
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,1,power_law_1.01,0.3223001480102539
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,1,power_law_1.01,0.31118207931518554
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,1,power_law_1.01,0.34753406524658204
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,1,power_law_1.01,0.3703577423095703
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,1,power_law_1.01,0.3950271987915039
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,1,power_law_1.01,0.46249729156494135
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,1,power_law_1.01,0.528590087890625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,1,power_law_1.01,0.6670867156982422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,1,power_law_1.01,0.8033715057373048
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,1,power_law_1.01,1.073218536376953
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,1,power_law_1.01,1.3542239379882812
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,8,1,power_law_1.2,3.0962808227539065
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,1,power_law_1.01,1.885894470214844
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,1,power_law_1.01,2.4334629821777343
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,8,1,power_law_1.01,2.971049499511719
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,8,1,power_law_1.01,0.04733695983886719
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,8,1,power_law_1.01,0.08174976348876953
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,8,1,power_law_1.01,0.05411392211914062
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,8,1,power_law_1.01,0.09872960090637208
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,8,1,power_law_1.01,0.13828800201416017
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,8,1,power_law_1.01,0.14428223609924315
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,8,1,power_law_1.01,0.14860544204711915
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,8,1,power_law_1.01,0.1532761573791504
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,8,1,power_law_1.01,0.15532992362976075
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,8,1,power_law_1.01,0.1612940788269043
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,8,1,power_law_1.01,0.1666431999206543
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,8,1,power_law_1.01,0.17131200790405274
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,8,1,power_law_1.01,0.06737919807434081
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,8,1,power_law_1.01,0.18171903610229492
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,8,1,power_law_1.01,0.18969600677490234
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,8,1,power_law_1.01,0.19406591415405275
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,8,1,power_law_1.01,0.21786304473876955
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,8,1,power_law_1.01,0.25862144470214843
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,8,1,power_law_1.01,0.29627519607543945
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,8,1,power_law_1.01,0.4758246231079101
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,8,1,power_law_1.01,0.38188991546630857
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,8,1,power_law_1.01,0.6554265594482421
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,1536,8,128,8,1,power_law_1.2,3.875368957519531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,8,1,power_law_1.01,0.8479872131347657
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,8,1,power_law_1.01,1.249947509765625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,8,1,power_law_1.01,1.6659622192382812
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,8,1,power_law_1.01,2.502813415527344
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,1,power_law_1.2,0.08265151977539062
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,1,power_law_1.2,0.10331711769104004
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,1,power_law_1.2,0.13494208335876465
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,1,power_law_1.2,0.1826144027709961
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,1,power_law_1.2,0.06531455993652344
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,1,power_law_1.2,0.31050432205200196
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,1,power_law_1.2,0.3155859184265137
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,1,power_law_1.2,0.3282227325439453
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,1,power_law_1.2,0.3347808074951172
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,1,power_law_1.2,0.34053054809570316
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,1,power_law_1.2,0.3559481430053711
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,1,power_law_1.2,0.38918846130371093
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,1,power_law_1.2,0.4132537460327148
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,1,power_law_1.2,0.4480947113037109
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,1,power_law_1.2,0.4219942474365235
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,1,power_law_1.2,0.4703558349609375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,1,power_law_1.2,0.5168889617919922
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,1,power_law_1.2,0.5602316665649414
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,1,power_law_1.2,0.5967654418945313
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,1,power_law_1.2,0.6996800231933593
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,1,power_law_1.2,0.8230982208251953
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,1,power_law_1.2,1.0605209350585938
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,8,1,power_law_1.01,3.3627697753906247
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,1,power_law_1.2,1.3006796264648437
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,1,power_law_1.2,1.778991394042969
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,2560,8,160,8,1,power_law_1.01,4.232560729980468
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,1,power_law_1.2,2.2650674438476566
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,1,power_law_1.2,0.04765952110290527
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,1,power_law_1.2,0.06274816036224365
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,1,power_law_1.2,0.08303487777709961
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,1,power_law_1.2,0.1070470428466797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,1,power_law_1.2,0.15881216049194335
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,1,power_law_1.2,0.225860481262207
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,1,power_law_1.2,0.2384870338439941
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,1,power_law_1.2,0.2465388870239258
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,1,power_law_1.2,0.2515616035461426
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,1,power_law_1.2,3.242947692871094
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,1,power_law_1.2,0.25628223419189455
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,1,power_law_1.2,0.26166528701782227
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,1,power_law_1.2,0.2665247917175293
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,1,power_law_1.2,0.3222028732299805
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,1,power_law_1.2,0.3235667037963867
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,1,power_law_1.2,0.34204097747802736
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,1,power_law_1.2,0.3355622482299805
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,1,power_law_1.2,0.35751678466796877
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,1,power_law_1.2,0.3758176040649414
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,1,power_law_1.2,0.40366977691650396
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,1,power_law_1.2,0.4753120040893554
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,1,power_law_1.2,0.5459711837768555
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,1,power_law_1.2,0.6864556884765625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,1,power_law_1.2,0.8149894714355469
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,1,power_law_1.2,1.088046112060547
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,1,power_law_1.2,1.3602117919921874
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,1,power_law_1.2,4.200789184570313
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,1,power_law_1.2,1.9100070190429688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,8,1,power_law_1.2,5.180214233398438
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,1,power_law_1.2,2.4475801086425784
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,8,1,power_law_1.2,0.05447487831115723
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,8,1,power_law_1.2,0.06587584018707275
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,8,1,power_law_1.2,0.07807168006896972
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,8,1,power_law_1.2,0.09568256378173828
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,8,1,power_law_1.2,0.13568960189819337
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,8,1,power_law_1.2,0.14526399612426757
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,8,1,power_law_1.2,0.15082048416137694
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,8,1,power_law_1.2,0.15644351959228514
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,8,1,power_law_1.2,0.15898752212524414
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,8,1,power_law_1.2,0.16662208557128905
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,8,1,power_law_1.2,0.17111936569213868
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,8,1,power_law_1.2,0.17945856094360352
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,8,1,power_law_1.2,0.18982912063598634
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,8,1,power_law_1.2,0.21900096893310547
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,8,1,power_law_1.2,0.04743807792663574
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,8,1,power_law_1.2,2.9816244506835936
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,8,1,power_law_1.2,0.2067027282714844
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,8,1,power_law_1.2,0.24581888198852536
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,8,1,power_law_1.2,0.3019820785522461
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,8,1,power_law_1.2,0.3645568084716797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,8,1,power_law_1.2,0.4997273635864258
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,8,1,power_law_1.2,0.6450982666015624
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,8,1,power_law_1.2,0.9625651550292968
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,8,1,power_law_1.2,2.011649322509766
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,8,1,power_law_1.2,1.3000090026855469
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,8,1,power_law_1.2,2.7300384521484373
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,1,power_law_1.01,0.07168575763702392
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,1,power_law_1.01,0.0874176025390625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,1,power_law_1.01,0.10936575889587404
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,1,power_law_1.01,0.155861759185791
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,1,power_law_1.01,0.288536319732666
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,1,power_law_1.01,0.3880281448364258
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,1,power_law_1.01,0.4217689514160156
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,1,power_law_1.01,0.529375991821289
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,1,power_law_1.01,0.5498009490966796
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,1,power_law_1.01,0.5682854461669922
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,1,power_law_1.01,0.5841388702392578
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,1,power_law_1.01,0.5916211318969726
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,1,power_law_1.01,0.6193254470825196
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,1,power_law_1.01,0.6686943817138672
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,1,power_law_1.01,0.7023693084716797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,1,power_law_1.01,0.7501881408691407
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,1,power_law_1.01,0.8378681945800782
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,1,power_law_1.01,0.8775718688964844
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,1,power_law_1.01,0.9871289825439453
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,1,power_law_1.01,1.0728141021728514
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,1,power_law_1.01,1.1595852661132813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,1,power_law_1.01,1.4012396240234375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,1,power_law_1.01,1.7255807495117188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,1,power_law_1.01,2.1476666259765627
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,1,power_law_1.01,3.5748001098632813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,8,1,power_law_1.2,4.204866638183594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,1,power_law_1.01,4.586295166015625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,1,power_law_1.01,0.05092480182647705
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,1,power_law_1.01,0.061134719848632814
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,1,power_law_1.01,0.08737215995788575
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,1,power_law_1.01,0.13129280090332032
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,8,1,power_law_1.2,5.666599731445312
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,1,power_law_1.01,0.21572799682617189
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,1,power_law_1.01,0.36893566131591793
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,1,power_law_1.01,0.44049022674560545
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,1,power_law_1.01,0.31139904022216797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,1,power_law_1.01,0.4534745788574218
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,1,power_law_1.01,0.4665836715698243
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,8,1,power_law_1.01,5.535947875976563
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,1,power_law_1.01,0.4786713409423828
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,1,power_law_1.01,0.4863590240478516
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,1,power_law_1.01,0.5027193450927735
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,1,power_law_1.01,0.5248761749267578
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,1,power_law_1.01,0.5209830474853516
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,1,power_law_1.01,0.5464863967895508
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,1,power_law_1.01,0.6126374435424805
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,1,power_law_1.01,0.6358585739135743
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,1,power_law_1.01,0.6670841979980469
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,1,power_law_1.01,0.7298016357421875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,1,power_law_1.01,0.8034681701660157
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,1,power_law_1.01,0.9335302734375001
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,1,power_law_1.01,1.0699967956542968
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,1,power_law_1.01,1.3525421142578125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,1,power_law_1.01,1.6296153259277344
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,1,power_law_1.01,2.191834259033203
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,1,power_law_1.01,2.7359661865234375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,8,1,power_law_1.01,3.303575134277344
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,8,1,power_law_1.01,0.07881216049194337
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,8,1,power_law_1.01,0.0787987184524536
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,8,1,power_law_1.01,0.08122816085815429
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,8,1,power_law_1.01,0.08442496299743653
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,8,1,power_law_1.01,0.09106687545776368
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,8,1,power_law_1.01,0.09791872024536133
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,8,1,power_law_1.01,0.10491135597229004
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,8,1,power_law_1.01,0.11278335571289062
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,8,1,power_law_1.01,0.11837247848510742
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,8,1,power_law_1.01,0.08003904342651366
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,8,1,power_law_1.01,0.13899968147277833
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,8,1,power_law_1.01,0.18202175140380858
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,8,1,power_law_1.01,0.2469094467163086
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,8,1,power_law_1.01,0.20052032470703124
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,8,1,power_law_1.01,0.29862911224365235
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,8,1,power_law_1.01,0.34396991729736326
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,8,1,power_law_1.01,0.43503360748291015
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,8,1,power_law_1.01,0.7972006225585938
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,8,1,power_law_1.01,0.6209926223754882
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,8,1,power_law_1.01,1.1660838317871094
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,8,1,power_law_1.01,1.5459603881835937
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,8,1,power_law_1.01,2.309853515625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,8,1,power_law_1.01,3.0543698120117186
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,1,power_law_1.2,0.06332543849945069
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,1,power_law_1.2,0.08798527717590332
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,1,power_law_1.2,0.10963199615478517
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,1,power_law_1.2,0.15568767547607423
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,1,power_law_1.2,0.24318336486816405
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,1,power_law_1.2,0.3672076797485352
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,1,power_law_1.2,0.4029593658447266
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,8,1,power_law_1.01,4.542143859863281
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,1,power_law_1.2,0.5265983963012696
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,1,power_law_1.2,0.5522175979614258
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,1,power_law_1.2,0.565994873046875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,1,power_law_1.2,0.5762879943847656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,1,power_law_1.2,0.5963372802734375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,1,power_law_1.2,0.622996482849121
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,1,power_law_1.2,0.6736441802978516
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,1,power_law_1.2,0.7151232147216797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,1,power_law_1.2,0.7613107299804687
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,1,power_law_1.2,0.8366374206542968
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,1,power_law_1.2,0.9106034851074218
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,1,power_law_1.2,0.9932575988769532
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,1,power_law_1.2,1.084189453125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,1,power_law_1.2,1.2018867492675782
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,1,power_law_1.2,1.4427488708496095
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,1,power_law_1.2,1.6976576232910154
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,1,power_law_1.2,2.1617356872558595
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,1,power_law_1.2,2.638155517578125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,1,power_law_1.2,3.6116952514648437
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,1,power_law_1.2,4.581480407714844
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,1,power_law_1.2,0.050321922302246094
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,1,power_law_1.2,0.061687040328979495
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,1,power_law_1.2,0.08642175674438476
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,1,power_law_1.2,0.12811967849731445
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,1,power_law_1.2,0.20139455795288086
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,8,1,power_law_1.2,5.578059692382813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,1,power_law_1.2,0.28774848937988284
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,1,power_law_1.2,0.3547097778320313
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,1,power_law_1.2,0.43628288269042964
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,1,power_law_1.2,0.4569977569580078
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,1,power_law_1.2,0.4481203079223633
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,1,power_law_1.2,0.4689120101928711
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,1,power_law_1.2,0.48417537689208984
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,1,power_law_1.2,0.4979238510131836
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,1,power_law_1.2,0.5171379089355469
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,1,power_law_1.2,0.5267583847045898
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,1,power_law_1.2,0.5441376113891602
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,1,power_law_1.2,0.6118656158447265
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,1,power_law_1.2,0.65010498046875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,1,power_law_1.2,0.6787916564941406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,1,power_law_1.2,0.760975341796875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,1,power_law_1.2,0.8306495666503906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,1,power_law_1.2,0.9777337646484374
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,1,power_law_1.2,1.1218489837646484
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,1,power_law_1.2,1.3944947814941406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,1,power_law_1.2,1.6691513061523438
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,1,power_law_1.2,2.2308216857910157
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,1,power_law_1.2,2.791100158691406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,8,1,power_law_1.2,0.07879615783691406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,8,1,power_law_1.2,3.3426132202148438
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,8,1,power_law_1.2,0.0797331190109253
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,8,1,power_law_1.2,0.08450112342834473
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,8,1,power_law_1.2,0.08111743927001953
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,8,1,power_law_1.2,0.091396484375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,8,1,power_law_1.2,0.09865983963012695
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,8,1,power_law_1.2,0.10514687538146972
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,8,1,power_law_1.2,0.11275135993957519
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,8,1,power_law_1.2,0.1196723175048828
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,8,1,power_law_1.2,0.17748863220214844
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,8,1,power_law_1.2,0.13598336219787596
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,8,1,power_law_1.2,0.200183048248291
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,8,1,power_law_1.2,0.24233856201171874
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,8,1,power_law_1.2,0.34481857299804686
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,8,1,power_law_1.2,0.07836287975311279
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,8,1,power_law_1.2,0.2986848068237305
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,8,1,power_law_1.2,0.6191584014892578
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,8,1,power_law_1.2,0.43623168945312496
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,8,1,power_law_1.2,0.7979673767089844
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,8,1,power_law_1.2,1.1629631805419922
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,8,1,power_law_1.2,1.537838134765625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,8,1,power_law_1.2,3.05291259765625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,8,1,power_law_1.2,2.3154080200195315
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,8,1,power_law_1.2,4.538206787109376
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,8,1,power_law_1.01,0.07330175876617431
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,8,1,power_law_1.01,0.08548095703125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,8,1,power_law_1.01,0.10568384170532226
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,8,1,power_law_1.01,0.15393856048583984
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,8,1,power_law_1.01,0.16252159118652343
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,8,1,power_law_1.01,0.16887424468994142
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,8,1,power_law_1.01,0.17513023376464845
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,8,1,power_law_1.01,0.18300928115844725
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,8,1,power_law_1.01,0.20261632919311526
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,8,1,power_law_1.01,0.2205830383300781
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,8,1,power_law_1.01,0.2246009635925293
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,8,1,power_law_1.01,0.23117439270019532
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,8,1,power_law_1.01,0.23693311691284177
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,8,1,power_law_1.01,0.2457356834411621
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,8,1,power_law_1.01,0.26727167129516605
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,8,1,power_law_1.01,0.31203071594238285
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,8,1,power_law_1.01,0.34383808135986327
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,8,1,power_law_1.01,0.4030182266235352
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,8,1,power_law_1.01,0.47787136077880865
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,8,1,power_law_1.01,0.6063904190063477
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,8,1,power_law_1.01,0.7484941101074218
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,8,1,power_law_1.01,1.0224813079833983
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,8,1,power_law_1.01,1.2962789916992188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,8,1,power_law_1.01,2.3880986022949218
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,2688,1856,6,128,8,1,power_law_1.01,2.9271731567382813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,8,1,power_law_1.01,1.8404197692871094
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,8,1,power_law_1.2,0.04738111972808838
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,8,1,power_law_1.2,0.06281343936920167
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,8,1,power_law_1.2,0.0794604778289795
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,8,1,power_law_1.2,0.1004697608947754
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,8,1,power_law_1.2,0.15392127990722657
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,8,1,power_law_1.2,0.17217983245849608
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,8,1,power_law_1.2,0.17995712280273438
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,8,1,power_law_1.2,0.18896575927734377
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,8,1,power_law_1.2,0.03081664085388184
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,8,1,power_law_1.2,0.2065715217590332
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,8,1,power_law_1.2,0.22201919555664062
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,8,1,power_law_1.2,0.2309734344482422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,8,1,power_law_1.2,0.23947263717651368
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,8,1,power_law_1.2,0.24602239608764648
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,8,1,power_law_1.2,0.2560895919799805
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,8,1,power_law_1.2,0.285468807220459
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,8,1,power_law_1.2,0.31922111511230467
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,8,1,power_law_1.2,0.34699775695800783
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,8,1,power_law_1.2,0.16167936325073243
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,8,1,power_law_1.2,0.41385726928710936
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,8,1,power_law_1.2,0.6137971115112305
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,8,1,power_law_1.2,0.7430687713623046
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,8,1,power_law_1.2,1.0244486236572266
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,8,1,power_law_1.2,0.47836929321289057
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,8,1,power_law_1.2,1.2978700256347655
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,8,1,power_law_1.2,1.8365843200683591
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,8,1,power_law_1.2,2.3961497497558595
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,8,1,power_law_1.01,0.08026687622070312
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,8,1,power_law_1.01,0.10630784034729004
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,8,1,power_law_1.01,0.18124351501464844
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,8,1,power_law_1.01,0.27105215072631833
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,2688,1856,6,128,8,1,power_law_1.2,2.9337472534179687
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,8,1,power_law_1.01,0.38121406555175785
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,8,1,power_law_1.01,0.6209260940551757
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,8,1,power_law_1.01,0.6431571197509766
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,8,1,power_law_1.01,0.6688473510742188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,8,1,power_law_1.01,0.69425537109375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,8,1,power_law_1.01,0.7299372863769531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,8,1,power_law_1.01,0.7840512084960938
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,8,1,power_law_1.01,0.8312928009033204
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,8,1,power_law_1.01,0.9469222259521484
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,8,1,power_law_1.01,1.017936019897461
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,8,1,power_law_1.01,0.5885260772705079
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,8,1,power_law_1.01,1.1245113372802735
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,8,1,power_law_1.01,1.3335192871093748
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,8,1,power_law_1.01,1.143132781982422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,8,1,power_law_1.01,1.0279315185546873
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,8,1,power_law_1.01,1.3622886657714843
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,8,1,power_law_1.01,2.13863037109375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,8,1,power_law_1.01,1.6859519958496094
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,8,1,power_law_1.01,2.585008544921875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,8,1,power_law_1.01,3.575727233886719
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,8,1,power_law_1.01,4.622593994140625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,8,1,power_law_1.01,6.658651733398438
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,8,1,power_law_1.01,0.08613632202148438
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,8,1,power_law_1.01,8.949779052734375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,8,1,power_law_1.01,0.13424575805664063
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,8,1,power_law_1.01,0.0591321611404419
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,8,1,power_law_1.01,0.2932595252990723
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,8,1,power_law_1.01,0.47471168518066403
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,8,1,power_law_1.01,0.49390911102294927
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,8,1,power_law_1.01,0.5121887969970703
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,8,1,power_law_1.01,0.21275583267211914
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,8,1,power_law_1.01,0.4530976104736328
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,8,1,power_law_1.01,0.5188889694213867
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,2688,22,512,8,1,power_law_1.01,10.784567871093751
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,8,1,power_law_1.01,0.530261116027832
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,8,1,power_law_1.01,0.551844482421875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,8,1,power_law_1.01,0.5582271957397461
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,8,1,power_law_1.01,0.581135368347168
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,8,1,power_law_1.01,0.5897695922851562
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,8,1,power_law_1.01,0.6079455947875976
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,8,1,power_law_1.01,0.6445216369628907
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,8,1,power_law_1.01,0.7376921844482422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,8,1,power_law_1.01,0.7935155487060547
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,8,1,power_law_1.01,0.9238886260986329
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,8,1,power_law_1.01,1.0575392150878906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,8,1,power_law_1.01,1.32901123046875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,8,1,power_law_1.01,1.5907066345214844
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,8,1,power_law_1.01,2.1352621459960934
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,8,1,power_law_1.01,2.6660186767578127
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,8,1,power_law_1.01,3.7503359985351565
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,8,1,power_law_1.01,4.830377502441406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,2688,22,512,8,1,power_law_1.01,5.925245971679688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,8,1,power_law_1.2,0.07959231853485108
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,8,1,power_law_1.2,0.10686847686767578
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,8,1,power_law_1.2,0.18079296112060547
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,8,1,power_law_1.2,0.2632358360290527
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,8,1,power_law_1.2,0.3429318237304687
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,8,1,power_law_1.2,0.5756800079345703
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,8,1,power_law_1.2,0.6200569534301759
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,8,1,power_law_1.2,0.6386905670166015
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,8,1,power_law_1.2,0.6711910247802735
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,8,1,power_law_1.2,0.7116678619384766
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,8,1,power_law_1.2,0.7410931396484375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,8,1,power_law_1.2,0.8084966278076171
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,8,1,power_law_1.2,0.8574617767333985
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,8,1,power_law_1.2,0.97670654296875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,8,1,power_law_1.2,1.0540806579589845
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,8,1,power_law_1.2,1.158627166748047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,8,1,power_law_1.2,1.3640237426757813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,8,1,power_law_1.2,1.0454176330566407
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,8,1,power_law_1.2,1.1526841735839843
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,8,1,power_law_1.2,1.4048959350585937
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,8,1,power_law_1.2,1.6998323059082032
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,8,1,power_law_1.2,2.1249203491210937
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,8,1,power_law_1.2,2.5976959228515626
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,8,1,power_law_1.2,3.5998785400390623
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,8,1,power_law_1.2,4.568778381347657
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,8,1,power_law_1.2,6.6172900390625005
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,8,1,power_law_1.2,8.6721728515625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,8,1,power_law_1.2,0.05849472045898437
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,8,1,power_law_1.2,0.08723775863647461
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,8,1,power_law_1.2,0.13406784057617188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,8,1,power_law_1.2,0.2692576026916504
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,8,1,power_law_1.2,0.4405036926269531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,8,1,power_law_1.2,0.4749887847900391
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,8,1,power_law_1.2,0.4887033462524414
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,8,1,power_law_1.2,0.5103251266479492
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,8,1,power_law_1.2,0.5232640075683593
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,8,1,power_law_1.2,0.5363142395019531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,8,1,power_law_1.2,0.5592403030395507
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,8,1,power_law_1.2,0.5711001586914063
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,8,1,power_law_1.2,0.19087039947509765
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,8,1,power_law_1.2,0.6080684661865234
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,8,1,power_law_1.2,0.6004486465454102
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,8,1,power_law_1.2,0.631233901977539
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,8,1,power_law_1.2,0.668287353515625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,8,1,power_law_1.2,0.767633285522461
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,8,1,power_law_1.2,0.8268192291259766
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,8,1,power_law_1.2,0.9549113464355468
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,8,1,power_law_1.2,1.088260498046875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,8,1,power_law_1.2,1.3556192016601563
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,8,1,power_law_1.2,1.6200601196289064
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,8,1,power_law_1.2,2.6947552490234377
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,8,1,power_law_1.2,2.1534739685058595
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,2688,22,512,8,1,power_law_1.2,10.9146484375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,8,1,power_law_1.2,3.792869262695312
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,8,1,power_law_1.2,4.848238220214844
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,2688,22,512,8,1,power_law_1.2,5.9352917480468745
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,16,1,power_law_1.01,0.05258240222930908
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,16,1,power_law_1.01,0.05517375946044922
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,16,1,power_law_1.01,0.06477952003479004
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,16,1,power_law_1.01,0.0929312038421631
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,16,1,power_law_1.01,0.09411711692810058
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,16,1,power_law_1.01,0.09535231590270996
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,16,1,power_law_1.01,0.09576191902160644
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,16,1,power_law_1.01,0.09682175636291504
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,16,1,power_law_1.01,0.09788607597351075
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,16,1,power_law_1.01,0.10133440017700195
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,16,1,power_law_1.01,0.10217791557312013
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,16,1,power_law_1.01,0.10516736030578613
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,16,1,power_law_1.01,0.10880640029907226
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,16,1,power_law_1.01,0.11217151641845705
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,16,1,power_law_1.01,0.12042048454284668
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,16,1,power_law_1.01,0.12421055793762208
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,16,1,power_law_1.01,0.1412268829345703
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,16,1,power_law_1.01,0.17990144729614257
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,16,1,power_law_1.01,0.22954368591308594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,16,1,power_law_1.01,0.27801151275634767
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,16,1,power_law_1.01,0.3762438583374023
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,16,1,power_law_1.01,0.47292736053466794
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,16,1,power_law_1.01,0.6717132568359375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,16,1,power_law_1.01,0.8694054412841797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,16,1,power_law_1.01,1.2587814331054688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,16,1,power_law_1.01,0.032126080989837644
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,16,1,power_law_1.01,0.035291519165039066
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,16,1,power_law_1.01,0.03654848098754883
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,16,1,power_law_1.01,0.039863679409027096
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,16,1,power_law_1.01,0.04392576217651367
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,16,1,power_law_1.01,0.045724802017211914
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,16,1,power_law_1.01,0.04562367916107178
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,16,1,power_law_1.01,0.046819200515747075
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,16,1,power_law_1.01,0.0468281602859497
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,16,1,power_law_1.01,0.0472441577911377
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,16,1,power_law_1.01,0.049640960693359375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,16,1,power_law_1.01,0.04932544231414795
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,16,1,power_law_1.01,0.050852479934692385
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,16,1,power_law_1.01,0.05693376064300537
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,16,1,power_law_1.01,0.061078400611877436
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,16,1,power_law_1.01,0.0628281593322754
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,16,1,power_law_1.01,0.06472320079803466
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,16,1,power_law_1.01,0.07308351993560791
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,16,1,power_law_1.01,0.0903769588470459
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,16,1,power_law_1.01,0.11365056037902832
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,16,1,power_law_1.01,0.12737215995788573
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,16,1,power_law_1.01,0.17318016052246094
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,16,1,power_law_1.01,0.22277248382568357
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,16,1,power_law_1.01,2.034954833984375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,16,1,power_law_1.01,0.30402687072753903
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,16,1,power_law_1.01,0.4023526382446289
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,16,1,power_law_1.01,0.5690796661376953
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,16,1,power_law_1.01,0.7436819458007813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,16,1,power_law_1.01,0.01808127999305725
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,16,1,power_law_1.01,0.020714879035949707
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,16,1,power_law_1.01,0.020585598945617675
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,16,1,power_law_1.01,0.02139775991439819
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,16,1,power_law_1.01,0.02139775991439819
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,16,1,power_law_1.01,0.02178623914718628
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,16,1,power_law_1.01,0.02176383972167969
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,16,1,power_law_1.01,0.021823360919952392
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,16,1,power_law_1.01,0.02179327964782715
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,16,1,power_law_1.01,0.01877951979637146
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,16,1,power_law_1.01,0.9210310363769532
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,16,1,power_law_1.01,0.02020224094390869
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,16,1,power_law_1.01,0.023725440502166746
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,16,1,power_law_1.01,0.023405439853668213
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,16,1,power_law_1.01,0.02637696027755737
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,16,1,power_law_1.01,0.029370880126953124
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,16,1,power_law_1.01,0.03241535902023315
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,16,1,power_law_1.01,0.03555520057678223
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,16,1,power_law_1.01,0.042028799057006835
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,16,1,power_law_1.01,0.04859776020050049
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,16,1,power_law_1.01,0.06266816139221192
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,16,1,power_law_1.01,0.0759987211227417
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,16,1,power_law_1.01,0.10087552070617675
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,16,1,power_law_1.01,0.12721407890319825
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,16,1,power_law_1.01,0.18267776489257812
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,16,1,power_law_1.01,0.23941120147705078
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,16,1,power_law_1.2,0.04675712108612061
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,16,1,power_law_1.2,0.05337344169616699
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,16,1,power_law_1.2,0.05391104221343994
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,16,1,power_law_1.2,0.06395455837249756
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,16,1,power_law_1.2,0.09361791610717773
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,16,1,power_law_1.2,0.09518272399902343
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,16,1,power_law_1.2,0.09614656448364259
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,16,1,power_law_1.2,0.09675904273986816
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,16,1,power_law_1.2,0.09778752326965331
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,16,1,power_law_1.2,0.09906047821044922
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,16,1,power_law_1.2,0.10086336135864257
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,16,1,power_law_1.2,0.10295743942260742
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,16,1,power_law_1.2,0.10485119819641113
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,16,1,power_law_1.2,0.10997695922851562
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,16,1,power_law_1.2,0.11372608184814452
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,16,1,power_law_1.2,0.12286463737487793
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,16,1,power_law_1.2,0.1242527961730957
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,16,1,power_law_1.2,0.14226880073547363
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,16,1,power_law_1.2,0.18168767929077148
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,16,1,power_law_1.2,0.23102783203125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,16,1,power_law_1.01,0.35309505462646484
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,16,1,power_law_1.2,0.2762041664123535
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,16,1,power_law_1.2,0.3777254486083984
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,16,1,power_law_1.2,0.47462337493896484
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,16,1,power_law_1.01,0.46379966735839845
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,16,1,power_law_1.2,0.6723884582519531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,16,1,power_law_1.01,0.5750425720214845
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,16,1,power_law_1.2,0.8666662597656251
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,16,1,power_law_1.2,0.03441152095794678
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,16,1,power_law_1.2,0.036983680725097653
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,16,1,power_law_1.2,0.03875711917877197
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,16,1,power_law_1.2,0.03864576101303101
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,16,1,power_law_1.2,0.043937277793884275
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,16,1,power_law_1.2,0.04621503829956054
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,16,1,power_law_1.2,0.04559487819671631
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,16,1,power_law_1.2,0.046464638710021974
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,16,1,power_law_1.2,0.047226881980895995
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,16,1,power_law_1.2,0.04849023818969726
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,16,1,power_law_1.2,0.04868800163269043
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,16,1,power_law_1.2,0.050828161239624026
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,16,1,power_law_1.2,0.05131135940551758
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,16,1,power_law_1.2,0.058140802383422854
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,16,1,power_law_1.2,0.06174272060394287
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,16,1,power_law_1.2,0.06636352062225341
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,16,1,power_law_1.2,0.06643648147583008
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,16,1,power_law_1.2,0.07222591876983643
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,16,1,power_law_1.2,0.0903321647644043
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,16,1,power_law_1.2,0.11188672065734864
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,16,1,power_law_1.2,0.12653568267822266
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,16,1,power_law_1.2,0.174006404876709
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,16,1,power_law_1.2,1.2515392303466797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,16,1,power_law_1.2,0.22429119110107423
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,16,1,power_law_1.2,0.3050150489807129
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,16,1,power_law_1.2,1.6857337951660156
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,16,1,power_law_1.2,0.40277374267578125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,16,1,power_law_1.2,2.0194975280761716
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,16,1,power_law_1.2,0.01894719958305359
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,16,1,power_law_1.2,0.01934720039367676
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,16,1,power_law_1.2,0.0202239990234375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,16,1,power_law_1.2,0.02139967918395996
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,16,1,power_law_1.2,0.021480960845947267
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,16,1,power_law_1.2,0.02149120092391968
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,16,1,power_law_1.2,0.021437439918518066
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,16,1,power_law_1.2,0.021680641174316406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,16,1,power_law_1.2,0.021838719844818118
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,16,1,power_law_1.2,0.021867520809173584
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,16,1,power_law_1.2,0.022135679721832273
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,16,1,power_law_1.2,0.023479681015014648
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,16,1,power_law_1.2,0.02465791940689087
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,16,1,power_law_1.2,0.026394879817962645
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,16,1,power_law_1.2,0.02925184011459351
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,16,1,power_law_1.2,0.033263359069824215
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,16,1,power_law_1.2,0.573955192565918
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,16,1,power_law_1.2,0.0358515191078186
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,16,1,power_law_1.2,0.04277056217193603
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,16,1,power_law_1.2,0.04977280139923096
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,16,1,power_law_1.2,0.7500460815429688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,16,1,power_law_1.2,0.10467967987060547
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,16,1,power_law_1.2,0.06298240184783935
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,16,1,power_law_1.2,0.07723392009735107
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,16,1,power_law_1.2,0.9179341125488282
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,16,1,power_law_1.2,0.1325126361846924
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,16,1,power_law_1.2,0.1895167922973633
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,16,1,power_law_1.2,0.24571008682250978
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,16,1,power_law_1.01,0.06358975887298583
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,16,1,power_law_1.01,0.07169727802276611
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,16,1,power_law_1.01,0.07535999774932861
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,16,1,power_law_1.01,0.0887667179107666
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,16,1,power_law_1.01,0.12875136375427246
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,16,1,power_law_1.01,0.12940287590026855
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,16,1,power_law_1.01,0.13031552314758302
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,16,1,power_law_1.01,0.1314515209197998
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,16,1,power_law_1.01,0.13187840461730957
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,16,1,power_law_1.01,0.1344703960418701
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,16,1,power_law_1.01,0.1403116798400879
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,16,1,power_law_1.01,0.1458784008026123
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,16,1,power_law_1.01,0.1509017562866211
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,16,1,power_law_1.01,0.15476096153259278
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,16,1,power_law_1.01,0.15888511657714843
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,16,1,power_law_1.01,0.1663609504699707
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,16,1,power_law_1.01,0.17578943252563478
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,16,1,power_law_1.01,0.22374912261962893
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,16,1,power_law_1.01,0.29315839767456053
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,16,1,power_law_1.01,0.3561139297485352
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,16,1,power_law_1.01,0.4302918243408203
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,16,1,power_law_1.01,0.592314224243164
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,16,1,power_law_1.01,0.7505856323242187
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,16,1,power_law_1.2,0.3654073715209961
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,16,1,power_law_1.01,1.0683148956298827
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,16,1,power_law_1.01,1.3904698181152344
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,16,1,power_law_1.01,2.039862365722656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,16,1,power_law_1.2,0.4833599853515625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,16,1,power_law_1.01,0.04044032096862793
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,16,1,power_law_1.01,0.04926464080810547
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,16,1,power_law_1.01,0.050615677833557124
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,16,1,power_law_1.01,0.06260223865509033
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,16,1,power_law_1.01,0.06835008144378663
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,16,1,power_law_1.01,0.0695743989944458
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,16,1,power_law_1.01,0.07047232151031493
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,16,1,power_law_1.01,0.07059711933135986
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,16,1,power_law_1.01,0.07125823974609376
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,16,1,power_law_1.01,0.07269824028015137
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,16,1,power_law_1.01,0.07413119792938232
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,16,1,power_law_1.01,0.07457856178283692
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,16,1,power_law_1.01,0.07717184066772462
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,16,1,power_law_1.01,0.07993663787841797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,16,1,power_law_1.01,0.08283519744873047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,16,1,power_law_1.01,0.08555520057678223
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,16,1,power_law_1.01,0.0873158359527588
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,16,1,power_law_1.01,0.09901823997497558
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,16,1,power_law_1.01,0.1322169589996338
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,16,1,power_law_1.01,0.15328831672668458
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,16,1,power_law_1.01,0.1771718406677246
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,16,1,power_law_1.01,0.2480031967163086
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,16,1,power_law_1.01,2.668751220703125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,16,1,power_law_1.01,0.31515520095825195
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,16,1,power_law_1.01,3.3624288940429685
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,16,1,power_law_1.2,0.5945849609375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,16,1,power_law_1.01,0.4516569519042969
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,16,1,power_law_1.01,0.5735961532592773
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,16,1,power_law_1.01,0.8354214477539064
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,16,1,power_law_1.01,1.0860979461669922
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,16,1,power_law_1.01,0.020216960906982422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,16,1,power_law_1.01,0.02764607906341553
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,16,1,power_law_1.01,0.02908096075057983
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,16,1,power_law_1.01,0.02924864053726196
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,16,1,power_law_1.01,0.02929215908050537
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,16,1,power_law_1.01,0.02914752006530762
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,16,1,power_law_1.01,0.029336960315704347
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,16,1,power_law_1.01,0.03005120038986206
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,16,1,power_law_1.01,0.03087104082107544
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,16,1,power_law_1.01,0.023959040641784668
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,16,1,power_law_1.01,1.3502386474609376
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,16,1,power_law_1.01,0.024356479644775393
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,16,1,power_law_1.01,0.03171008110046387
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,16,1,power_law_1.01,0.03393856048583984
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,16,1,power_law_1.01,0.041579518318176266
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,16,1,power_law_1.01,0.03964031934738159
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,16,1,power_law_1.01,0.043206400871276855
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,16,1,power_law_1.01,0.047608962059021
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,16,1,power_law_1.01,0.055646719932556155
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,16,1,power_law_1.01,0.08551103591918945
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,16,1,power_law_1.01,0.06867775917053223
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,16,1,power_law_1.01,0.1010086441040039
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,16,1,power_law_1.01,0.14042495727539062
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,16,1,power_law_1.01,0.18229440689086912
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,16,1,power_law_1.01,0.257061767578125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,16,1,power_law_1.01,0.33822654724121093
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,16,1,power_law_1.2,0.06304255962371827
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,16,1,power_law_1.2,0.07078591823577882
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,16,1,power_law_1.2,0.07257408142089844
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,16,1,power_law_1.2,0.08884927749633789
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,16,1,power_law_1.2,0.1282271957397461
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,16,1,power_law_1.2,0.12902591705322267
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,16,1,power_law_1.2,0.130828161239624
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,16,1,power_law_1.2,0.13126272201538086
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,16,1,power_law_1.2,0.13238911628723143
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,16,1,power_law_1.2,0.13507072448730467
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,16,1,power_law_1.2,0.14044032096862794
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,16,1,power_law_1.2,0.14651328086853027
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,16,1,power_law_1.2,0.15036864280700685
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,16,1,power_law_1.2,0.15538816452026366
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,16,1,power_law_1.2,0.15964096069335937
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,16,1,power_law_1.2,0.16888959884643556
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,16,1,power_law_1.2,0.17711231231689453
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,16,1,power_law_1.2,0.22693567276000975
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,16,1,power_law_1.01,0.4984441757202148
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,16,1,power_law_1.2,0.2975584030151367
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,16,1,power_law_1.2,0.3561593627929688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,16,1,power_law_1.2,0.4346425628662109
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,16,1,power_law_1.2,0.5977190399169923
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,16,1,power_law_1.2,0.7568243408203126
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,16,1,power_law_1.01,0.6624185943603516
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,16,1,power_law_1.2,1.069518051147461
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,16,1,power_law_1.01,0.8175968170166016
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,16,1,power_law_1.2,1.3857843017578124
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,16,1,power_law_1.2,0.04070911884307861
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,16,1,power_law_1.2,0.04887040138244629
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,16,1,power_law_1.2,0.04934656143188477
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,16,1,power_law_1.2,0.06257855892181396
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,16,1,power_law_1.2,0.06801216125488281
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,16,1,power_law_1.2,0.06949312210083007
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,16,1,power_law_1.2,0.07027135848999023
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,16,1,power_law_1.2,0.07037312030792237
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,16,1,power_law_1.2,0.07162559986114501
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,16,1,power_law_1.2,0.07186367988586426
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,16,1,power_law_1.2,0.07348608016967774
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,16,1,power_law_1.2,0.07446591854095459
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,16,1,power_law_1.2,0.0760972785949707
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,16,1,power_law_1.2,0.07971583843231202
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,16,1,power_law_1.2,0.08265088081359864
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,16,1,power_law_1.2,0.08557184219360352
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,16,1,power_law_1.2,0.08961919784545899
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,16,1,power_law_1.2,0.09893376350402831
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,16,1,power_law_1.2,0.13255231857299804
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,16,1,power_law_1.2,0.15249407768249512
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,16,1,power_law_1.2,2.040838470458984
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,16,1,power_law_1.2,0.17871231079101563
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,16,1,power_law_1.2,0.24846271514892576
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,16,1,power_law_1.2,0.3146879959106445
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,16,1,power_law_1.2,2.7444436645507815
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,16,1,power_law_1.2,0.4517670440673828
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,16,1,power_law_1.2,3.3910208129882813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,16,1,power_law_1.2,0.5745158386230469
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,16,1,power_law_1.2,0.8360940551757812
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,16,1,power_law_1.2,0.02366080045700073
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,16,1,power_law_1.2,0.024352641105651857
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,16,1,power_law_1.2,0.027580161094665528
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,16,1,power_law_1.2,0.028406400680541993
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,16,1,power_law_1.2,0.029481599330902098
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,16,1,power_law_1.2,0.029614078998565673
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,16,1,power_law_1.2,0.029609599113464356
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,16,1,power_law_1.2,0.029592320919036862
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,16,1,power_law_1.2,0.030259199142456054
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,16,1,power_law_1.2,0.030646400451660154
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,16,1,power_law_1.2,0.0206060791015625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,16,1,power_law_1.2,0.03358464002609253
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,16,1,power_law_1.2,0.040052480697631836
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,16,1,power_law_1.2,0.03172096014022827
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,16,1,power_law_1.2,0.044767360687255855
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,16,1,power_law_1.2,1.091827163696289
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,16,1,power_law_1.2,1.3431558227539062
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,16,1,power_law_1.2,0.04383935928344727
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,16,1,power_law_1.2,0.0486521577835083
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,16,1,power_law_1.2,0.05655360221862793
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,16,1,power_law_1.2,0.06877056121826172
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,16,1,power_law_1.2,0.10402367591857911
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,16,1,power_law_1.2,0.1465497589111328
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,16,1,power_law_1.2,0.08638591766357422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,16,1,power_law_1.2,0.1854035186767578
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,16,1,power_law_1.2,0.26686208724975585
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,16,1,power_law_1.2,0.3511475372314453
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,16,1,power_law_1.01,0.04692863941192627
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,16,1,power_law_1.01,0.06588799953460693
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,16,1,power_law_1.01,0.08649472236633302
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,16,1,power_law_1.01,0.11072064399719239
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,16,1,power_law_1.01,0.14157119750976563
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,16,1,power_law_1.01,0.1844339179992676
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,16,1,power_law_1.01,0.23968576431274413
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,16,1,power_law_1.01,0.24782335281372073
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,16,1,power_law_1.01,0.2512332725524902
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,16,1,power_law_1.01,0.2557977676391602
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,16,1,power_law_1.01,0.26931903839111326
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,16,1,power_law_1.01,0.29796480178833007
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,16,1,power_law_1.01,0.317127685546875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,16,1,power_law_1.01,0.3384960174560547
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,16,1,power_law_1.01,0.38362560272216795
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,16,1,power_law_1.01,0.3925376129150391
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,16,1,power_law_1.01,0.41560447692871094
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,16,1,power_law_1.01,0.46545665740966796
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,16,1,power_law_1.01,0.5525491333007813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,16,1,power_law_1.01,0.6205433654785156
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,16,1,power_law_1.01,0.6912940979003906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,16,1,power_law_1.01,0.861413116455078
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,16,1,power_law_1.01,1.0438387298583984
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,16,1,power_law_1.01,1.3986015319824219
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,16,1,power_law_1.01,1.7540914916992185
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,16,1,power_law_1.01,2.4614259338378908
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,16,1,power_law_1.2,0.5187750244140625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,16,1,power_law_1.01,3.161015625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,16,1,power_law_1.01,0.04637311935424805
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,16,1,power_law_1.01,0.05178815841674804
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,16,1,power_law_1.01,0.07331264019012451
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,16,1,power_law_1.01,3.8658892822265627
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,16,1,power_law_1.01,0.10503680229187011
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,16,1,power_law_1.01,0.2066374397277832
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,16,1,power_law_1.01,0.2603654479980469
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,16,1,power_law_1.01,0.2678214454650879
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,16,1,power_law_1.01,0.27195648193359373
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,16,1,power_law_1.01,0.2758899116516113
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,16,1,power_law_1.01,0.2843411254882812
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,16,1,power_law_1.01,0.29268608093261717
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,16,1,power_law_1.01,0.29679040908813475
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,16,1,power_law_1.01,0.3039411163330078
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,16,1,power_law_1.01,0.33662654876708986
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,16,1,power_law_1.01,0.1473536014556885
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,16,1,power_law_1.01,0.3422649765014648
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,16,1,power_law_1.01,0.3551174545288086
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,16,1,power_law_1.01,0.3778035354614258
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,16,1,power_law_1.01,0.40595649719238286
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,16,1,power_law_1.01,0.4559711837768555
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,16,1,power_law_1.01,0.5248659133911133
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,16,1,power_law_1.01,0.6325894546508789
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,16,1,power_law_1.01,0.7393740844726563
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,16,1,power_law_1.01,0.9680883026123046
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,16,1,power_law_1.01,1.211499557495117
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,16,1,power_law_1.2,0.6832377624511718
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,16,1,power_law_1.01,1.6763002014160158
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,16,1,power_law_1.01,2.125445709228516
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,16,1,power_law_1.01,2.58363525390625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,16,1,power_law_1.01,0.06830399990081787
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,16,1,power_law_1.01,0.0974726390838623
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,16,1,power_law_1.01,0.08336447715759278
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,16,1,power_law_1.01,0.11873151779174804
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,16,1,power_law_1.2,0.8443647766113281
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,16,1,power_law_1.01,0.13662783622741698
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,16,1,power_law_1.01,0.16990848541259768
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,16,1,power_law_1.01,0.17986688613891602
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,16,1,power_law_1.01,0.1839263916015625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,16,1,power_law_1.01,0.191843204498291
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,16,1,power_law_1.01,0.19840768814086912
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,16,1,power_law_1.01,0.07102015972137452
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,16,1,power_law_1.01,0.2031839942932129
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,16,1,power_law_1.01,0.21194112777709959
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,16,1,power_law_1.01,0.22922880172729493
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,16,1,power_law_1.01,0.22284927368164062
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,16,1,power_law_1.01,0.2524985694885254
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,16,1,power_law_1.01,0.17642623901367188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,16,1,power_law_1.01,0.2932953643798828
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,16,1,power_law_1.01,0.431025276184082
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,16,1,power_law_1.01,0.33297534942626955
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,16,1,power_law_1.01,0.5248543930053711
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,16,1,power_law_1.01,0.7036390686035157
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,16,1,power_law_1.01,0.930206756591797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,16,1,power_law_1.01,1.2969100952148438
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,16,1,power_law_1.01,1.804258575439453
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,16,1,power_law_1.2,0.04696512222290039
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,16,1,power_law_1.2,0.06713088035583496
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,16,1,power_law_1.2,0.08660160064697266
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,16,1,power_law_1.01,2.6729638671875002
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,16,1,power_law_1.2,0.10760831832885742
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,16,1,power_law_1.2,0.17972608566284182
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,16,1,power_law_1.2,0.2318252754211426
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,16,1,power_law_1.2,0.2450227165222168
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,16,1,power_law_1.2,0.2486425590515137
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,16,1,power_law_1.01,3.3675436401367187
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,16,1,power_law_1.2,0.25530303955078126
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,16,1,power_law_1.2,0.3023961639404297
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,16,1,power_law_1.2,0.2698419189453125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,16,1,power_law_1.2,0.31976703643798826
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,16,1,power_law_1.2,0.3456076812744141
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,16,1,power_law_1.2,0.38438465118408205
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,16,1,power_law_1.2,0.39980926513671877
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,16,1,power_law_1.2,0.13457728385925294
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,16,1,power_law_1.2,0.41937534332275395
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,16,1,power_law_1.2,0.4863686370849609
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,16,1,power_law_1.2,0.6237849426269532
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,16,1,power_law_1.2,0.7039724731445313
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,16,1,power_law_1.2,0.8895014190673829
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,16,1,power_law_1.2,0.5631475067138672
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,16,1,power_law_1.2,1.0577145385742188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,16,1,power_law_1.2,1.4014694213867187
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,16,1,power_law_1.2,1.7622227478027344
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,16,1,power_law_1.2,2.470777587890625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,16,1,power_law_1.2,0.0460422420501709
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,16,1,power_law_1.2,0.052595200538635256
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,16,1,power_law_1.2,0.07355135917663574
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,16,1,power_law_1.2,3.167789306640625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,16,1,power_law_1.2,0.10153535842895507
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,16,1,power_law_1.2,0.1411334419250488
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,16,1,power_law_1.2,0.2003993606567383
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,16,1,power_law_1.2,0.25860095977783204
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,16,1,power_law_1.2,0.26544063568115234
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,16,1,power_law_1.2,0.2699193572998047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,16,1,power_law_1.2,0.27637887954711915
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,16,1,power_law_1.2,0.2829177665710449
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,16,1,power_law_1.2,3.8818817138671875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,16,1,power_law_1.2,0.29112192153930666
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,16,1,power_law_1.2,0.29725311279296873
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,16,1,power_law_1.2,0.3074265670776367
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,16,1,power_law_1.2,0.3397983932495117
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,16,1,power_law_1.2,0.3484928131103516
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,16,1,power_law_1.2,0.36117504119873045
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,16,1,power_law_1.2,0.39190719604492186
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,16,1,power_law_1.2,0.41615039825439454
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,16,1,power_law_1.2,0.47758209228515625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,16,1,power_law_1.2,0.5416313552856445
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,16,1,power_law_1.2,0.647570571899414
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,16,1,power_law_1.2,0.775955810546875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,16,1,power_law_1.2,0.9821273803710937
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,16,1,power_law_1.2,1.220907516479492
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,16,1,power_law_1.2,1.6914349365234376
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,7168,2048,8,256,16,1,power_law_1.01,4.361089172363281
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,16,1,power_law_1.2,0.06828927993774414
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,16,1,power_law_1.2,2.156035766601563
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,16,1,power_law_1.2,0.07428864002227784
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,16,1,power_law_1.2,2.5915066528320314
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,16,1,power_law_1.2,0.08195136070251466
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,16,1,power_law_1.2,0.09477888107299805
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,16,1,power_law_1.2,0.16861888885498047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,16,1,power_law_1.2,0.13181056022644042
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,16,1,power_law_1.2,0.11385408401489258
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,16,1,power_law_1.2,0.1748531150817871
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,16,1,power_law_1.2,0.18013887405395507
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,16,1,power_law_1.2,0.18498559951782229
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,16,1,power_law_1.2,0.19303680419921876
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,16,1,power_law_1.2,0.20081855773925783
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,16,1,power_law_1.2,0.22101823806762697
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,16,1,power_law_1.2,0.2082828712463379
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,16,1,power_law_1.2,0.23451200485229493
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,16,1,power_law_1.2,0.24684864044189453
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,16,1,power_law_1.2,0.2761484718322754
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,16,1,power_law_1.2,0.33710079193115233
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,16,1,power_law_1.2,0.5445721435546875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,16,1,power_law_1.2,0.40777217864990234
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,16,1,power_law_1.2,0.7069062042236328
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,16,1,power_law_1.2,1.042437744140625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,16,1,power_law_1.2,1.3909683227539062
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,16,1,power_law_1.2,2.0589913940429687
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,16,1,power_law_1.01,0.04603519916534424
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,16,1,power_law_1.01,0.05803647994995117
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,16,1,power_law_1.01,0.06681536197662354
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,16,1,power_law_1.01,0.07725376129150391
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,16,1,power_law_1.01,0.09330944061279298
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,16,1,power_law_1.01,0.09654335975646973
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,16,1,power_law_1.01,0.09896703720092773
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,16,1,power_law_1.01,0.10247296333312987
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,16,1,power_law_1.01,0.1053279972076416
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,16,1,power_law_1.01,0.11154303550720215
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,16,1,power_law_1.01,0.14780096054077146
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,16,1,power_law_1.01,0.15206080436706543
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,16,1,power_law_1.01,0.15649663925170898
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,16,1,power_law_1.01,0.16007360458374023
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,16,1,power_law_1.01,0.1638310432434082
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,16,1,power_law_1.01,0.17364864349365233
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,16,1,power_law_1.01,0.1941606330871582
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,16,1,power_law_1.01,0.21183296203613283
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,16,1,power_law_1.2,2.8507403564453124
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,16,1,power_law_1.01,0.25166336059570316
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,16,1,power_law_1.01,0.2974150466918945
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,16,1,power_law_1.01,0.41500030517578124
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,16,1,power_law_1.01,0.49574913024902345
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,16,1,power_law_1.01,0.6973919677734375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,16,1,power_law_1.01,0.8779379272460938
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,16,1,power_law_1.01,1.2661587524414064
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,16,1,power_law_1.01,1.6357542419433593
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,16,1,power_law_1.01,0.03665791988372803
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,16,1,power_law_1.01,2.030431365966797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,16,1,power_law_1.01,0.052250881195068356
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,16,1,power_law_1.01,0.042007040977478025
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,16,1,power_law_1.01,0.06346752166748046
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,16,1,power_law_1.01,0.0822860813140869
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,16,1,power_law_1.01,0.10020992279052734
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,16,1,power_law_1.01,0.09770303726196289
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,16,1,power_law_1.01,0.10267392158508301
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,16,1,power_law_1.01,0.10610624313354493
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,16,1,power_law_1.01,0.10555839538574219
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,16,1,power_law_1.01,0.10976639747619629
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,16,1,power_law_1.01,0.12248319625854491
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,16,1,power_law_1.01,0.12450688362121583
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,16,1,power_law_1.01,0.12978303909301758
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,16,1,power_law_1.01,0.13110783576965332
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,16,1,power_law_1.01,0.13402432441711426
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,16,1,power_law_1.01,0.14231040000915526
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,16,1,power_law_1.01,0.15675711631774902
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,16,1,power_law_1.01,0.18211904525756836
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,16,1,power_law_1.01,0.21641536712646486
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,16,1,power_law_1.01,0.2511545562744141
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,16,1,power_law_1.01,0.31673919677734375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,16,1,power_law_1.01,0.391383056640625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,16,1,power_law_1.01,0.5370399856567383
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,16,1,power_law_1.01,0.6634931182861328
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,16,1,power_law_1.01,0.9539405059814452
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,16,1,power_law_1.01,1.2334336090087892
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,16,1,power_law_1.01,1.5202975463867188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,16,1,power_law_1.01,0.0451423978805542
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,16,1,power_law_1.01,0.05169023990631103
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,16,1,power_law_1.01,0.05785600185394287
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,16,1,power_law_1.01,0.06353600025177002
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,16,1,power_law_1.01,0.08015680313110352
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,16,1,power_law_1.01,0.08629952430725098
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,16,1,power_law_1.01,0.08797760009765625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,16,1,power_law_1.01,0.047239041328430174
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,16,1,power_law_1.01,0.08949312210083007
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,16,1,power_law_1.01,0.09196479797363281
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,16,1,power_law_1.01,0.09761216163635253
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,16,1,power_law_1.01,0.10310400009155274
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,16,1,power_law_1.01,0.08171072006225585
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,16,1,power_law_1.01,0.11269696235656738
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,16,1,power_law_1.01,0.09273280143737793
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,16,1,power_law_1.01,0.11785728454589843
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,16,1,power_law_1.01,0.12799360275268554
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,16,1,power_law_1.01,0.18116928100585938
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,16,1,power_law_1.01,0.1527622413635254
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,16,1,power_law_1.01,0.24008256912231446
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,16,1,power_law_1.01,0.29964351654052734
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,16,1,power_law_1.01,0.4178438568115235
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,16,1,power_law_1.01,0.5419430541992187
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,16,1,power_law_1.01,1.0697376251220703
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,16,1,power_law_1.01,0.7991929626464843
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,16,1,power_law_1.2,0.05098495960235596
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,16,1,power_law_1.2,0.057487998008728024
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,16,1,power_law_1.01,1.6187341308593752
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,16,1,power_law_1.2,0.06403327941894531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,16,1,power_law_1.2,0.09399040222167969
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,16,1,power_law_1.2,0.09671808242797851
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,16,1,power_law_1.2,0.10008576393127441
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,16,1,power_law_1.2,0.1044108772277832
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,16,1,power_law_1.2,0.1068166446685791
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,16,1,power_law_1.2,0.03506367921829224
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,16,1,power_law_1.2,0.11229503631591797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,16,1,power_law_1.2,0.147827844619751
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,16,1,power_law_1.2,0.1566662406921387
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,16,1,power_law_1.2,0.15205247879028322
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,16,1,power_law_1.2,0.1595494365692139
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,16,1,power_law_1.2,0.16440767288208008
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,16,1,power_law_1.2,0.1739334487915039
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,16,1,power_law_1.2,0.19460800170898437
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,16,1,power_law_1.2,0.21224576950073243
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,16,1,power_law_1.2,0.25274879455566407
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,16,1,power_law_1.2,0.30204160690307613
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,16,1,power_law_1.2,0.41610496520996093
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,16,1,power_law_1.2,0.07452928066253663
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,16,1,power_law_1.2,0.5003238296508788
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,16,1,power_law_1.2,0.6988089752197266
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,16,1,power_law_1.2,0.8831334686279296
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,16,1,power_law_1.2,1.2699417877197265
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,16,1,power_law_1.2,0.036780800819396976
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,16,1,power_law_1.2,0.04038271903991699
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,16,1,power_law_1.2,0.051976318359374996
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,16,1,power_law_1.2,0.06137792110443115
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,16,1,power_law_1.2,0.07676928043365479
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,16,1,power_law_1.2,0.09862272262573242
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,16,1,power_law_1.2,0.10306559562683107
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,16,1,power_law_1.2,0.10378496170043945
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,16,1,power_law_1.2,0.10827327728271485
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,16,1,power_law_1.2,0.10672703742980957
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,16,1,power_law_1.2,0.11114175796508789
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,16,1,power_law_1.2,0.1242579174041748
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,16,1,power_law_1.2,0.12949119567871095
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,16,1,power_law_1.2,0.13159040451049803
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,16,1,power_law_1.2,0.13345536231994629
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,16,1,power_law_1.2,0.14508352279663086
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,16,1,power_law_1.2,0.16171136856079102
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,16,1,power_law_1.2,0.18789823532104494
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,16,1,power_law_1.2,0.222159366607666
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,16,1,power_law_1.2,0.2549600028991699
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,16,1,power_law_1.2,0.13170623779296875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,16,1,power_law_1.2,0.3180134391784668
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,16,1,power_law_1.2,0.39034496307373046
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,16,1,power_law_1.2,0.5371174240112305
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,16,1,power_law_1.2,0.6587238311767578
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,16,1,power_law_1.2,1.6517050170898435
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,16,1,power_law_1.2,0.9504550170898437
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,16,1,power_law_1.2,0.045146241188049316
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,16,1,power_law_1.2,0.04727168083190918
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,16,1,power_law_1.2,0.05162879943847656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,16,1,power_law_1.2,0.055959677696228026
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,16,1,power_law_1.2,0.06208703994750977
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,16,1,power_law_1.2,0.08122112274169921
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,16,1,power_law_1.2,2.027864990234375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,16,1,power_law_1.2,0.08533184051513672
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,16,1,power_law_1.2,0.0867193603515625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,16,1,power_law_1.2,0.08934335708618164
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,16,1,power_law_1.2,0.0899187183380127
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,16,1,power_law_1.2,0.09417344093322753
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,16,1,power_law_1.2,0.09670975685119629
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,16,1,power_law_1.2,0.10133567810058594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,16,1,power_law_1.2,0.11909952163696289
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,16,1,power_law_1.2,0.12660799980163576
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,16,1,power_law_1.2,0.11324159622192383
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,16,1,power_law_1.2,0.18242624282836914
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,16,1,power_law_1.2,0.14577664375305174
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,16,1,power_law_1.2,1.2304351806640625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,16,1,power_law_1.2,0.22294847488403322
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,16,1,power_law_1.2,0.3054252815246582
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,16,1,power_law_1.2,0.39042751312255863
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,16,1,power_law_1.2,0.5795609664916992
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,16,1,power_law_1.2,0.7701165008544921
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,16,1,power_law_1.2,1.1384178924560548
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,16,1,power_law_1.2,1.5043475341796875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,16,1,power_law_1.2,1.5233197021484375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,16,1,power_law_1.01,0.04654784202575683
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,16,1,power_law_1.01,0.07137856006622315
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,16,1,power_law_1.01,0.08773183822631836
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,16,1,power_law_1.01,0.10550463676452637
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,16,1,power_law_1.01,0.12892288208007813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,16,1,power_law_1.01,0.18264768600463868
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,16,1,power_law_1.01,0.19030527114868162
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,16,1,power_law_1.01,0.19735807418823242
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,16,1,power_law_1.01,0.2044256019592285
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,16,1,power_law_1.01,0.21130495071411132
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,16,1,power_law_1.01,0.22649280548095704
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,16,1,power_law_1.01,0.24838464736938476
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,16,1,power_law_1.01,0.26006336212158204
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,16,1,power_law_1.01,0.26971456527709964
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,16,1,power_law_1.01,0.2747711944580078
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,16,1,power_law_1.01,0.28044095993041995
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,16,1,power_law_1.01,0.30091968536376956
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,16,1,power_law_1.01,0.34896961212158206
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,16,1,power_law_1.01,0.429425277709961
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,16,1,power_law_1.01,0.4975308990478515
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,16,1,power_law_1.01,0.5767289733886718
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,16,1,power_law_1.01,0.7617171478271485
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,16,1,power_law_1.01,0.9174848175048828
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,16,1,power_law_1.01,1.2623379516601563
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,16,1,power_law_1.01,1.605914306640625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,16,1,power_law_1.2,2.256929931640625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,16,1,power_law_1.01,2.29015869140625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,16,1,power_law_1.01,2.985462951660156
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,16,1,power_law_1.01,0.04523519992828369
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,16,1,power_law_1.01,0.051927042007446286
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,16,1,power_law_1.01,0.07462016105651856
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,16,1,power_law_1.01,0.09809344291687012
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,16,1,power_law_1.01,3.666534423828125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,16,1,power_law_1.01,0.13247103691101075
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,16,1,power_law_1.01,0.18777408599853515
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,16,1,power_law_1.01,0.19137088775634764
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,16,1,power_law_1.01,0.1963270378112793
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,16,1,power_law_1.01,0.1999724769592285
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,16,1,power_law_1.01,0.20370624542236326
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,16,1,power_law_1.01,0.20691776275634766
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,16,1,power_law_1.01,0.21280128479003907
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,16,1,power_law_1.01,0.246374397277832
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,16,1,power_law_1.01,0.26270336151123047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,16,1,power_law_1.01,0.26335424423217774
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,16,1,power_law_1.01,0.2531443214416504
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,16,1,power_law_1.01,0.2868601608276367
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,16,1,power_law_1.01,0.3110374450683594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,16,1,power_law_1.01,0.33243392944335937
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,16,1,power_law_1.01,0.38346817016601564
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,16,1,power_law_1.01,0.4599238586425781
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,16,1,power_law_1.01,0.5783699035644532
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,16,1,power_law_1.01,0.6873702239990235
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,16,1,power_law_1.01,0.9419046020507812
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,16,1,power_law_1.01,1.1841011047363281
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,16,1,power_law_1.01,1.6576165771484377
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,16,1,power_law_1.01,2.1418885803222656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,16,1,power_law_1.01,2.6287109375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,16,1,power_law_1.2,2.9863269042968748
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,16,1,power_law_1.01,0.05178112030029297
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,16,1,power_law_1.01,0.07899199962615967
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,16,1,power_law_1.01,0.09157055854797364
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,16,1,power_law_1.01,0.12383872032165529
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,16,1,power_law_1.01,0.13190143585205077
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,16,1,power_law_1.01,0.13695743560791016
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,16,1,power_law_1.01,0.14131903648376465
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,16,1,power_law_1.01,0.14353535652160646
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,16,1,power_law_1.01,0.14804927825927733
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,16,1,power_law_1.01,0.056380801200866706
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,16,1,power_law_1.01,0.15505151748657225
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,16,1,power_law_1.01,0.15835264205932617
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,16,1,power_law_1.01,0.16696895599365233
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,16,1,power_law_1.01,0.17530303955078125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,16,1,power_law_1.01,0.06742720127105713
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,16,1,power_law_1.01,0.17814783096313475
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,16,1,power_law_1.01,0.2723590469360352
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,16,1,power_law_1.01,0.2364998435974121
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,1536,8,128,16,1,power_law_1.2,3.740285949707031
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,16,1,power_law_1.01,0.1983673667907715
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,16,1,power_law_1.01,0.35207744598388674
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,16,1,power_law_1.01,0.4380031967163086
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,16,1,power_law_1.01,0.6072556686401367
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,16,1,power_law_1.01,0.7923238372802734
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,16,1,power_law_1.01,1.1623980712890625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,16,1,power_law_1.01,1.5359744262695312
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,16,1,power_law_1.2,0.041900801658630374
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,16,1,power_law_1.2,0.06974207878112793
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,16,1,power_law_1.2,0.08694911956787109
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,16,1,power_law_1.2,0.10203200340270997
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,16,1,power_law_1.2,0.12848575592041017
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,16,1,power_law_1.2,0.19409727096557616
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,16,1,power_law_1.2,0.18698495864868164
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,16,1,power_law_1.2,0.20163391113281248
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,16,1,power_law_1.2,0.20984767913818358
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,16,1,power_law_1.2,0.22048511505126953
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,16,1,power_law_1.01,2.3191935729980466
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,16,1,power_law_1.2,0.23124864578247067
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,16,1,power_law_1.2,0.2608531188964844
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,16,1,power_law_1.2,0.26411584854125975
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,16,1,power_law_1.2,0.2722649574279785
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,16,1,power_law_1.2,0.2781017684936523
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,16,1,power_law_1.2,0.285295352935791
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,16,1,power_law_1.2,0.3084460830688477
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,16,1,power_law_1.2,0.3625881576538086
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,16,1,power_law_1.2,0.43924736022949223
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,16,1,power_law_1.2,0.49804351806640623
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,16,1,power_law_1.2,0.581578254699707
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,16,1,power_law_1.2,0.7609932708740235
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,16,1,power_law_1.2,0.9243987274169921
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,16,1,power_law_1.2,1.26565185546875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,16,1,power_law_1.2,1.6140768432617185
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,16,1,power_law_1.2,2.3001676940917966
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,16,1,power_law_1.01,3.1212518310546877
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,16,1,power_law_1.2,0.0470579195022583
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,16,1,power_law_1.2,0.06034239768981934
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,16,1,power_law_1.2,0.07347392082214356
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,16,1,power_law_1.2,0.09220416069030761
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,16,1,power_law_1.2,0.12397120475769044
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,16,1,power_law_1.2,0.18669631958007812
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,16,1,power_law_1.2,0.19056127548217774
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,16,1,power_law_1.2,0.1976063919067383
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,2560,8,160,16,1,power_law_1.01,3.9174130249023436
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,16,1,power_law_1.2,0.20245759963989257
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,16,1,power_law_1.2,0.21073791503906253
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,16,1,power_law_1.2,0.20435903549194334
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,16,1,power_law_1.2,0.2148102378845215
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,16,1,power_law_1.2,0.2616211128234863
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,16,1,power_law_1.2,0.26420671463012696
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,16,1,power_law_1.2,0.2741087913513184
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,16,1,power_law_1.2,0.2937900733947754
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,16,1,power_law_1.2,0.2827872085571289
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,16,1,power_law_1.2,0.3120128059387207
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,16,1,power_law_1.2,0.3395564651489258
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,16,1,power_law_1.2,0.3911040115356445
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,16,1,power_law_1.2,0.46606529235839844
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,16,1,power_law_1.2,0.5911296081542969
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,16,1,power_law_1.2,0.7000153350830078
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,16,1,power_law_1.2,2.997888488769531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,16,1,power_law_1.2,0.9436275482177734
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,16,1,power_law_1.2,1.190149154663086
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,16,1,power_law_1.2,1.6708038330078125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,16,1,power_law_1.2,0.05169280052185059
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,16,1,power_law_1.2,0.057257599830627434
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,16,1,power_law_1.2,0.06454783916473389
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,16,1,power_law_1.2,0.07383808135986328
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,16,1,power_law_1.2,0.08844672203063965
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,16,1,power_law_1.2,0.12486144065856934
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,16,1,power_law_1.2,0.13074624061584472
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,16,1,power_law_1.2,0.1380345630645752
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,16,1,power_law_1.2,3.6772570800781255
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,16,1,power_law_1.2,2.146992034912109
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,16,1,power_law_1.2,0.1430348777770996
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,16,1,power_law_1.2,0.15397567749023439
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,16,1,power_law_1.2,0.15745599746704103
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,16,1,power_law_1.2,0.16387264251708983
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,16,1,power_law_1.2,0.17467136383056642
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,16,1,power_law_1.2,2.6429153442382813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,16,1,power_law_1.2,0.14721856117248536
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,16,1,power_law_1.2,0.1922572708129883
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,16,1,power_law_1.2,0.1995743942260742
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,16,1,power_law_1.2,0.2767974472045899
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,16,1,power_law_1.2,0.3356492614746094
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,16,1,power_law_1.2,0.46229118347167975
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,16,1,power_law_1.2,0.22885503768920898
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,16,1,power_law_1.2,0.5960038375854493
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,16,1,power_law_1.2,0.8957958221435547
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,16,1,power_law_1.2,1.2109785461425782
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,16,1,power_law_1.2,1.8770495605468749
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,16,1,power_law_1.2,2.5646444702148434
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,16,1,power_law_1.01,0.05218111991882324
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,16,1,power_law_1.01,0.050044798851013185
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,16,1,power_law_1.01,0.09002047538757324
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,16,1,power_law_1.01,0.11406399726867675
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,16,1,power_law_1.01,0.1582636833190918
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,16,1,power_law_1.01,0.22412799835205077
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,16,1,power_law_1.01,0.2638662338256836
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,16,1,power_law_1.2,3.9337362670898437
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,16,1,power_law_1.01,0.33619583129882813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,16,1,power_law_1.01,0.34846656799316406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,16,1,power_law_1.01,0.3607340621948242
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,16,1,power_law_1.01,0.3720166397094727
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,16,1,power_law_1.01,0.37973312377929686
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,16,1,power_law_1.01,0.3928678512573242
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,16,1,power_law_1.01,0.41205310821533203
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,16,1,power_law_1.01,0.42817214965820316
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,16,1,power_law_1.01,0.45989246368408204
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,16,1,power_law_1.01,0.6171200180053711
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,16,1,power_law_1.01,0.6457753753662109
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,16,1,power_law_1.01,0.6733222198486328
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,16,1,power_law_1.01,0.7973567962646484
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,16,1,power_law_1.01,0.8869471740722655
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,16,1,power_law_1.01,1.029151382446289
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,16,1,power_law_1.01,1.206368637084961
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,16,1,power_law_1.01,1.5758650207519531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,16,1,power_law_1.01,1.9266995239257814
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,16,1,power_law_1.2,5.31177734375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,16,1,power_law_1.01,2.661125793457031
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,16,1,power_law_1.01,0.053034238815307624
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,16,1,power_law_1.01,0.05544191837310791
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,16,1,power_law_1.01,3.380881958007812
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,16,1,power_law_1.01,0.06828864097595215
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,16,1,power_law_1.01,0.16380416870117187
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,16,1,power_law_1.01,0.25251327514648436
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,16,1,power_law_1.01,4.123777160644531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,16,1,power_law_1.01,0.1097862434387207
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,16,1,power_law_1.01,0.30400064468383786
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,16,1,power_law_1.01,0.3447987365722656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,16,1,power_law_1.01,0.35504257202148437
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,16,1,power_law_1.01,0.3659743881225586
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,16,1,power_law_1.01,0.37625022888183596
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,16,1,power_law_1.01,0.381420783996582
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,16,1,power_law_1.01,0.39487422943115236
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,16,1,power_law_1.01,0.4138764953613281
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,16,1,power_law_1.01,0.4165958404541016
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,16,1,power_law_1.01,0.4332857513427735
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,16,1,power_law_1.01,0.48530815124511717
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,16,1,power_law_1.01,0.5062649536132813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,16,1,power_law_1.01,0.5321247863769532
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,16,1,power_law_1.01,0.5906867218017579
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,16,1,power_law_1.01,0.6425145721435547
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,16,1,power_law_1.01,0.7655238342285157
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,16,1,power_law_1.01,0.8918131256103516
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,16,1,power_law_1.01,1.136262435913086
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,16,1,power_law_1.01,1.3750105285644532
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,16,1,power_law_1.01,1.8593772888183593
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,16,1,power_law_1.01,2.3350341796875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,16,1,power_law_1.01,0.08414912223815918
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,16,1,power_law_1.01,0.08586624145507812
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,16,1,power_law_1.01,0.08493696212768555
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,16,1,power_law_1.01,0.08964096069335938
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,16,1,power_law_1.01,0.08416959762573242
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,16,1,power_law_1.01,0.09634112358093262
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,16,1,power_law_1.01,0.10340671539306641
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,16,1,power_law_1.01,0.10985728263854981
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,16,1,power_law_1.01,0.11667776107788086
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,16,1,power_law_1.01,0.12344511985778808
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,16,1,power_law_1.01,0.1367584037780762
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,16,1,power_law_1.01,0.1627692794799805
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,16,1,power_law_1.01,0.19146879196166994
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,16,1,power_law_1.01,0.23434175491333006
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,16,1,power_law_1.01,0.3341657638549805
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,16,1,power_law_1.01,0.29501760482788086
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,16,1,power_law_1.01,0.4244473648071289
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,16,1,power_law_1.01,0.5978092956542969
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,16,1,power_law_1.01,0.7724896240234375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,16,1,power_law_1.01,1.1283052825927735
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,16,1,power_law_1.01,1.4773983764648437
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,16,1,power_law_1.01,2.185405426025391
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,16,1,power_law_1.01,2.9145971679687497
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,16,1,power_law_1.2,0.05214975833892822
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,16,1,power_law_1.2,0.06566463947296143
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,16,1,power_law_1.2,0.09005503654479981
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,16,1,power_law_1.2,0.11399935722351073
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,16,1,power_law_1.01,4.337164306640625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,16,1,power_law_1.2,0.15168704032897948
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,16,1,power_law_1.2,0.24747327804565428
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,16,1,power_law_1.2,0.33094337463378903
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,16,1,power_law_1.2,0.2164537620544434
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,16,1,power_law_1.2,0.345464973449707
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,16,1,power_law_1.2,0.3514668655395508
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,16,1,power_law_1.2,0.3653683090209961
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,16,1,power_law_1.2,0.37852542877197265
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,16,1,power_law_1.2,0.3927916717529297
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,16,1,power_law_1.2,0.4096646499633789
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,16,1,power_law_1.2,0.43206336975097653
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,16,1,power_law_1.2,0.4658483123779297
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,16,1,power_law_1.2,0.6154649734497071
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,16,1,power_law_1.2,0.6429138946533203
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,16,1,power_law_1.2,0.6780025482177734
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,16,1,power_law_1.2,0.813815689086914
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,16,1,power_law_1.2,0.893337631225586
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,16,1,power_law_1.2,1.0380063629150391
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,16,1,power_law_1.2,1.2194297790527344
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,16,1,power_law_1.2,1.5871717834472656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,16,1,power_law_1.2,1.9482637023925782
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,16,1,power_law_1.2,2.6751806640624998
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,16,1,power_law_1.2,3.4040231323242187
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,16,1,power_law_1.2,0.04975103855133057
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,16,1,power_law_1.2,0.05507199764251709
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,16,1,power_law_1.2,0.07718463897705077
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,16,1,power_law_1.2,4.131729431152344
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,16,1,power_law_1.2,0.10979711532592773
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,16,1,power_law_1.2,0.15628864288330077
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,16,1,power_law_1.2,0.23361536026000979
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,16,1,power_law_1.2,0.29419008255004886
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,16,1,power_law_1.2,0.3400787353515625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,16,1,power_law_1.2,0.35018497467041015
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,16,1,power_law_1.2,0.3572864151000977
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,16,1,power_law_1.2,0.3670399856567383
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,16,1,power_law_1.2,0.37931583404541014
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,16,1,power_law_1.2,0.39175872802734374
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,16,1,power_law_1.2,0.40869632720947263
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,16,1,power_law_1.2,0.41645568847656256
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,16,1,power_law_1.2,0.4301593780517578
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,16,1,power_law_1.2,0.4821926498413086
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,16,1,power_law_1.2,0.5154624176025391
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,16,1,power_law_1.2,0.5431974411010743
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,16,1,power_law_1.2,0.6020716857910157
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,16,1,power_law_1.2,0.6653523254394531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,16,1,power_law_1.2,0.8017874908447264
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,16,1,power_law_1.2,0.9186406707763671
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,16,1,power_law_1.2,1.1613471984863282
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,16,1,power_law_1.2,1.4094931030273439
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,16,1,power_law_1.2,1.8852140808105466
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,16,1,power_law_1.2,2.3717887878417967
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,16,1,power_law_1.2,0.08375743865966798
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,16,1,power_law_1.2,2.8551339721679687
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,16,1,power_law_1.2,0.0840608024597168
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,16,1,power_law_1.2,0.08526911735534667
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,16,1,power_law_1.2,0.08630016326904297
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,16,1,power_law_1.2,0.08934911727905273
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,16,1,power_law_1.2,0.09631296157836913
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,16,1,power_law_1.2,0.1032857608795166
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,16,1,power_law_1.2,0.10972928047180175
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,16,1,power_law_1.2,0.11650303840637206
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,16,1,power_law_1.2,0.12309311866760253
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,16,1,power_law_1.2,0.1367187213897705
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,16,1,power_law_1.2,0.1656595230102539
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,16,1,power_law_1.2,0.23396095275878906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,16,1,power_law_1.2,0.2933657646179199
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,16,1,power_law_1.2,0.3337305450439453
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,16,1,power_law_1.2,0.5975142288208007
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,16,1,power_law_1.2,0.18773311614990235
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,16,1,power_law_1.2,0.7728108978271484
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,16,1,power_law_1.2,0.4226777648925781
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,16,1,power_law_1.2,1.4759564208984375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,16,1,power_law_1.2,2.1947975158691406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,16,1,power_law_1.2,1.1266867065429689
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,16,1,power_law_1.2,2.9087231445312502
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,16,1,power_law_1.2,4.338466491699219
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,16,1,power_law_1.01,0.06088255882263184
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,16,1,power_law_1.01,0.06766143798828125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,16,1,power_law_1.01,0.08169280052185059
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,16,1,power_law_1.01,0.11163392066955566
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,16,1,power_law_1.01,0.11881535530090333
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,16,1,power_law_1.01,0.12327360153198241
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,16,1,power_law_1.01,0.12733375549316406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,16,1,power_law_1.01,0.13161664009094237
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,16,1,power_law_1.01,0.1400044822692871
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,16,1,power_law_1.01,0.1709209632873535
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,16,1,power_law_1.01,0.18625215530395506
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,16,1,power_law_1.01,0.19643199920654297
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,16,1,power_law_1.01,0.1981273651123047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,16,1,power_law_1.01,0.20226112365722657
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,16,1,power_law_1.01,0.21271360397338865
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,16,1,power_law_1.01,0.2333344078063965
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,16,1,power_law_1.01,0.2552128028869629
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,16,1,power_law_1.01,0.3122444725036621
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,16,1,power_law_1.01,0.3631603240966797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,16,1,power_law_1.01,0.46804992675781254
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,16,1,power_law_1.01,0.5739699172973632
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,16,1,power_law_1.01,0.7727436828613281
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,16,1,power_law_1.01,0.9915251159667969
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,16,1,power_law_1.01,1.4061158752441405
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,16,1,power_law_1.01,1.827879638671875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,2688,1856,6,128,16,1,power_law_1.01,2.2531231689453124
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,16,1,power_law_1.2,0.031226239204406737
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,16,1,power_law_1.2,0.0418720006942749
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,16,1,power_law_1.2,0.05290048122406006
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,16,1,power_law_1.2,0.06349823951721192
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,16,1,power_law_1.2,0.07767615795135499
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,16,1,power_law_1.2,0.11873023986816407
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,16,1,power_law_1.2,0.13339967727661134
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,16,1,power_law_1.2,0.1418841552734375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,16,1,power_law_1.2,0.18988800048828125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,16,1,power_law_1.2,0.19252159118652343
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,16,1,power_law_1.2,0.19858112335205078
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,16,1,power_law_1.2,0.20015871047973635
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,16,1,power_law_1.2,0.11965824127197267
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,16,1,power_law_1.2,0.12416383743286133
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,16,1,power_law_1.2,0.20538623809814452
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,16,1,power_law_1.2,0.2113555145263672
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,16,1,power_law_1.2,0.23161151885986325
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,16,1,power_law_1.2,0.25619264602661135
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,16,1,power_law_1.2,0.3204927825927734
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,16,1,power_law_1.2,0.3628153610229492
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,16,1,power_law_1.2,0.47036289215087895
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,16,1,power_law_1.2,0.5724140930175781
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,16,1,power_law_1.2,0.7743385314941407
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,16,1,power_law_1.2,0.10784640312194824
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,16,1,power_law_1.2,0.9889875030517578
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,16,1,power_law_1.2,1.4070553588867187
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,16,1,power_law_1.2,1.837890625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,16,1,power_law_1.01,0.07721727848052978
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,16,1,power_law_1.01,0.11653759956359863
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,16,1,power_law_1.01,0.18035455703735354
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,16,1,power_law_1.01,0.28828672409057615
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,16,1,power_law_1.01,0.42159168243408207
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,16,1,power_law_1.01,0.6521612548828125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,16,1,power_law_1.01,0.6959257507324219
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,16,1,power_law_1.01,0.7283513641357422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,16,1,power_law_1.01,0.7531629180908203
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,16,1,power_law_1.01,0.7723513793945312
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,16,1,power_law_1.01,0.8354585266113281
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,16,1,power_law_1.01,0.904180450439453
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,2688,1856,6,128,16,1,power_law_1.2,2.256573486328125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,16,1,power_law_1.01,0.963301773071289
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,16,1,power_law_1.01,1.106742401123047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,16,1,power_law_1.01,1.3283769226074218
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,16,1,power_law_1.01,1.5749824523925782
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,16,1,power_law_1.01,1.1889778900146486
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,16,1,power_law_1.01,1.2800775146484376
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,16,1,power_law_1.01,1.4192498779296874
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,16,1,power_law_1.01,1.736876220703125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,16,1,power_law_1.01,2.0640325927734375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,16,1,power_law_1.01,2.6975347900390623
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,16,1,power_law_1.01,3.3056243896484374
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,16,1,power_law_1.01,4.548235473632813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,16,1,power_law_1.01,5.809195556640625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,16,1,power_law_1.01,8.339794921875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,16,1,power_law_1.01,10.881003417968751
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,2688,22,512,16,1,power_law_1.01,13.3845947265625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,16,1,power_law_1.01,0.3099020767211914
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,16,1,power_law_1.01,0.9312850952148437
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,16,1,power_law_1.01,0.44756351470947264
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,16,1,power_law_1.01,1.544469757080078
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,16,1,power_law_1.01,2.5588441467285157
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,16,1,power_law_1.01,3.8552410888671873
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,16,1,power_law_1.01,3.6567584228515626
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,16,1,power_law_1.01,4.1391909790039065
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,16,1,power_law_1.01,4.044149169921875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,16,1,power_law_1.01,3.4671585083007814
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,16,1,power_law_1.01,4.367869567871094
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,16,1,power_law_1.01,4.188200988769531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,16,1,power_law_1.01,3.9054931640625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,16,1,power_law_1.01,4.660406494140625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,16,1,power_law_1.01,4.792869262695312
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,16,1,power_law_1.01,4.615946350097657
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,16,1,power_law_1.01,4.453591003417968
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,16,1,power_law_1.01,6.161605224609375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,16,1,power_law_1.01,6.4499969482421875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,16,1,power_law_1.01,7.433215942382813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,16,1,power_law_1.01,8.736591186523437
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,16,1,power_law_1.01,10.343869628906251
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,16,1,power_law_1.01,11.88683349609375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,16,1,power_law_1.01,15.59110107421875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,16,1,power_law_1.01,20.927812499999998
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,16,1,power_law_1.01,26.988649902343752
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,16,1,power_law_1.01,33.2919775390625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,16,1,power_law_1.2,0.11666751861572267
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,16,1,power_law_1.2,0.18040704727172852
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,16,1,power_law_1.2,0.2700761604309082
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,16,1,power_law_1.2,0.38366783142089844
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,16,1,power_law_1.2,0.6397235107421875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,16,1,power_law_1.2,0.6961856079101563
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,16,1,power_law_1.2,0.7155680084228516
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,16,1,power_law_1.2,0.7572723388671875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,16,1,power_law_1.2,0.07013951778411866
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,16,1,power_law_1.2,0.7972032165527343
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,16,1,power_law_1.2,0.8593952178955078
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,16,1,power_law_1.2,0.9307424163818359
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,16,1,power_law_1.2,1.001002883911133
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,16,1,power_law_1.2,1.1389676666259765
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,16,1,power_law_1.2,1.2438803100585938
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,2688,22,512,16,1,power_law_1.01,40.5814794921875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,16,1,power_law_1.2,1.3697401428222657
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,16,1,power_law_1.2,1.3150277709960938
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,16,1,power_law_1.2,1.4622752380371093
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,16,1,power_law_1.2,1.6133657836914062
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,16,1,power_law_1.2,1.7810336303710936
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,16,1,power_law_1.2,2.0856851196289066
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,16,1,power_law_1.2,2.7146725463867187
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,16,1,power_law_1.2,3.3275021362304686
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,16,1,power_law_1.2,4.5995416259765625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,16,1,power_law_1.2,5.833775024414062
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,16,1,power_law_1.2,8.369603881835937
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,16,1,power_law_1.2,0.4911135864257813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,16,1,power_law_1.2,0.9310124969482422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,16,1,power_law_1.2,0.2630265617370605
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,16,1,power_law_1.2,1.9913888549804688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,16,1,power_law_1.2,3.879109191894531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,16,1,power_law_1.2,4.136867980957031
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,16,1,power_law_1.2,10.890611572265625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,16,1,power_law_1.2,4.3153515625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,16,1,power_law_1.2,1.2522656250000002
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,16,1,power_law_1.2,4.141452941894531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,16,1,power_law_1.2,4.589306945800781
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,16,1,power_law_1.2,4.898590087890625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,16,1,power_law_1.2,4.489256896972656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,16,1,power_law_1.2,4.8715930175781255
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,16,1,power_law_1.2,5.349535522460938
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,16,1,power_law_1.2,5.006024169921875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,16,1,power_law_1.2,5.538275146484375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,16,1,power_law_1.2,6.1724365234375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,16,1,power_law_1.2,6.4005078125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,16,1,power_law_1.2,7.13336669921875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,16,1,power_law_1.2,7.741970825195312
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,16,1,power_law_1.2,8.557150268554688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,2688,22,512,16,1,power_law_1.2,13.484895019531251
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,16,1,power_law_1.2,11.091550292968751
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,16,1,power_law_1.2,12.100836181640625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,16,1,power_law_1.2,15.611604003906251
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,16,1,power_law_1.2,19.590898437499998
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,16,1,power_law_1.2,29.6903076171875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,16,1,power_law_1.2,33.458232421874996
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,32,1,power_law_1.01,0.03294528007507324
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,32,1,power_law_1.01,0.04104896068572998
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,32,1,power_law_1.01,0.038483200073242185
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,32,1,power_law_1.01,0.0421343994140625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,32,1,power_law_1.01,0.06435904026031494
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,32,1,power_law_1.01,0.06486720085144043
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,32,1,power_law_1.01,0.06609471797943114
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,32,1,power_law_1.01,0.06598527908325195
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,32,1,power_law_1.01,0.07014400005340576
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,32,1,power_law_1.01,0.07837696075439453
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,32,1,power_law_1.01,0.08121472358703613
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,32,1,power_law_1.01,0.07109824180603028
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,32,1,power_law_1.01,0.08437888145446777
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,32,1,power_law_1.01,0.0886086368560791
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,32,1,power_law_1.01,0.09348735809326172
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,32,1,power_law_1.01,0.09642687797546387
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,32,1,power_law_1.01,0.09800383567810059
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,32,1,power_law_1.01,0.11021504402160645
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,32,1,power_law_1.01,0.12071359634399415
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,32,1,power_law_1.01,0.14670783996582032
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,32,1,power_law_1.01,0.17362623214721679
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,32,1,power_law_1.01,0.24354112625122068
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,32,1,power_law_1.01,0.29698495864868163
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,32,1,power_law_1.01,0.42342975616455075
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,32,1,power_law_1.01,0.538240623474121
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,32,1,power_law_1.01,0.777804183959961
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,2688,22,512,16,1,power_law_1.2,40.53327880859375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,32,1,power_law_1.01,0.03306688070297241
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,32,1,power_law_1.01,0.03505919933319092
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,32,1,power_law_1.01,0.036424961090087894
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,32,1,power_law_1.01,0.037128319740295415
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,32,1,power_law_1.01,0.04277440071105957
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,32,1,power_law_1.01,0.04325439929962158
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,32,1,power_law_1.01,0.0443884801864624
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,32,1,power_law_1.01,0.043598718643188476
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,32,1,power_law_1.01,0.04520063877105713
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,32,1,power_law_1.01,0.04562240123748779
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,32,1,power_law_1.01,0.04621823787689209
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,32,1,power_law_1.01,0.04757887840270996
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,32,1,power_law_1.01,0.048136320114135746
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,32,1,power_law_1.01,0.05387904167175293
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,32,1,power_law_1.01,0.05626815795898438
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,32,1,power_law_1.01,0.058734078407287595
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,32,1,power_law_1.01,0.0562937593460083
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,32,1,power_law_1.01,0.06113855838775635
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,32,1,power_law_1.01,0.07055103778839111
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,32,1,power_law_1.01,0.08762432098388671
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,32,1,power_law_1.01,0.10049152374267578
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,32,1,power_law_1.01,1.0089266967773436
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,32,1,power_law_1.01,0.14240703582763672
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,32,1,power_law_1.01,0.17444032669067383
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,32,1,power_law_1.01,0.24875904083251954
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,32,1,power_law_1.01,0.3245318222045898
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,32,1,power_law_1.01,1.2457810974121093
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,32,1,power_law_1.01,0.47322433471679687
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,32,1,power_law_1.01,0.01857471942901611
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,32,1,power_law_1.01,0.019773440361022947
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,32,1,power_law_1.01,0.020675840377807616
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,32,1,power_law_1.01,0.606355857849121
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,32,1,power_law_1.01,0.02056063890457153
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,32,1,power_law_1.01,0.021003520488739012
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,32,1,power_law_1.01,0.02064255952835083
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,32,1,power_law_1.01,0.021042559146881103
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,32,1,power_law_1.01,0.020634880065917967
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,32,1,power_law_1.01,0.020996479988098143
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,32,1,power_law_1.01,0.019402240514755247
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,32,1,power_law_1.01,0.7534239959716797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,32,1,power_law_1.01,0.021489279270172117
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,32,1,power_law_1.01,0.0206931209564209
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,32,1,power_law_1.01,0.02188352108001709
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,32,1,power_law_1.01,0.02315200090408325
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,32,1,power_law_1.01,0.02507136106491089
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,32,1,power_law_1.01,0.025053439140319822
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,32,1,power_law_1.01,0.02810879945755005
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,32,1,power_law_1.01,0.0328985595703125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,32,1,power_law_1.01,0.039413759708404536
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,32,1,power_law_1.01,0.05013184070587158
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,32,1,power_law_1.01,0.05998528003692627
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,32,1,power_law_1.01,0.0809171199798584
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,32,1,power_law_1.01,0.10116607666015624
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,32,1,power_law_1.01,0.1394559955596924
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,32,1,power_law_1.01,0.18256448745727538
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,32,1,power_law_1.2,0.03255872011184692
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,32,1,power_law_1.2,0.03911295890808105
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,32,1,power_law_1.2,0.04152959823608399
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,32,1,power_law_1.2,0.04188223838806152
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,32,1,power_law_1.2,0.0623686408996582
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,32,1,power_law_1.2,0.06410367965698242
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,32,1,power_law_1.2,0.06553664207458496
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,32,1,power_law_1.2,0.07147071838378907
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,32,1,power_law_1.2,0.07352255821228028
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,32,1,power_law_1.2,0.07684671878814697
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,32,1,power_law_1.2,0.08180928230285645
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,32,1,power_law_1.2,0.0817024040222168
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,32,1,power_law_1.2,0.08284031867980957
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,32,1,power_law_1.2,0.08979519844055175
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,32,1,power_law_1.2,0.09373184204101562
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,32,1,power_law_1.2,0.09737024307250977
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,32,1,power_law_1.2,0.09900544166564942
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,32,1,power_law_1.2,0.11088704109191894
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,32,1,power_law_1.2,0.12159168243408203
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,32,1,power_law_1.2,0.14621055603027344
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,32,1,power_law_1.2,0.17684223175048827
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,32,1,power_law_1.01,0.264771842956543
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,32,1,power_law_1.2,0.2432486343383789
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,32,1,power_law_1.01,0.3499571228027344
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,32,1,power_law_1.2,0.29821311950683593
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,32,1,power_law_1.2,0.42237823486328124
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,32,1,power_law_1.01,0.4347833633422852
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,32,1,power_law_1.2,0.542242546081543
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,32,1,power_law_1.2,0.032569599151611325
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,32,1,power_law_1.2,0.0338592004776001
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,32,1,power_law_1.2,0.03452415943145752
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,32,1,power_law_1.2,0.0361631989479065
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,32,1,power_law_1.2,0.04196735858917237
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,32,1,power_law_1.2,0.042756481170654295
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,32,1,power_law_1.2,0.04336063861846924
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,32,1,power_law_1.2,0.04433216094970703
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,32,1,power_law_1.2,0.044800000190734865
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,32,1,power_law_1.2,0.045540480613708495
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,32,1,power_law_1.2,0.04639999866485596
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,32,1,power_law_1.2,0.04725503921508789
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,32,1,power_law_1.2,0.048485760688781736
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,32,1,power_law_1.2,0.05295936107635498
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,32,1,power_law_1.2,0.05743872165679932
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,32,1,power_law_1.2,0.05917312145233154
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,32,1,power_law_1.2,0.05627711772918701
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,32,1,power_law_1.2,0.06077439785003662
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,32,1,power_law_1.2,0.07226816177368164
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,32,1,power_law_1.2,0.0857363224029541
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,32,1,power_law_1.2,0.10091135978698731
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,32,1,power_law_1.2,0.14280768394470217
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,32,1,power_law_1.2,0.7730886077880859
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,32,1,power_law_1.2,0.17378623962402345
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,32,1,power_law_1.2,0.24914688110351563
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,32,1,power_law_1.2,1.0092256164550781
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,32,1,power_law_1.2,0.3262598419189453
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,32,1,power_law_1.2,1.2386758422851563
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,32,1,power_law_1.2,0.01852671980857849
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,32,1,power_law_1.2,0.018910080194473267
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,32,1,power_law_1.2,0.0193612802028656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,32,1,power_law_1.2,0.020174078941345215
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,32,1,power_law_1.2,0.02058880090713501
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,32,1,power_law_1.2,0.020615038871765138
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,32,1,power_law_1.2,0.0208076810836792
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,32,1,power_law_1.2,0.020679039955139162
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,32,1,power_law_1.2,0.021006081104278564
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,32,1,power_law_1.2,0.020655360221862793
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,32,1,power_law_1.2,0.020967040061950683
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,32,1,power_law_1.2,0.4715564727783203
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,32,1,power_law_1.2,0.021811199188232423
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,32,1,power_law_1.2,0.023095040321350097
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,32,1,power_law_1.2,0.021416959762573243
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,32,1,power_law_1.2,0.025181438922882084
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,32,1,power_law_1.2,0.028040959835052493
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,32,1,power_law_1.2,0.023927679061889646
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,32,1,power_law_1.2,0.6082060623168946
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,32,1,power_law_1.2,0.7545145416259766
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,32,1,power_law_1.2,0.03381439924240112
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,32,1,power_law_1.2,0.03962111949920654
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,32,1,power_law_1.2,0.05048704147338867
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,32,1,power_law_1.2,0.061296000480651855
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,32,1,power_law_1.2,0.08325247764587403
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,32,1,power_law_1.2,0.1042854404449463
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,32,1,power_law_1.2,0.14614080429077148
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,32,1,power_law_1.2,0.1892300796508789
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,32,1,power_law_1.01,0.04193280220031738
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,32,1,power_law_1.01,0.061366400718688964
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,32,1,power_law_1.01,0.06271552085876465
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,32,1,power_law_1.01,0.06999872207641603
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,32,1,power_law_1.01,0.11338175773620604
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,32,1,power_law_1.01,0.11469696044921876
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,32,1,power_law_1.01,0.11559807777404785
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,32,1,power_law_1.01,0.11665472030639648
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,32,1,power_law_1.01,0.1177401638031006
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,32,1,power_law_1.01,0.11896639823913575
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,32,1,power_law_1.01,0.12249343872070313
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,32,1,power_law_1.01,0.12515711784362793
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,32,1,power_law_1.01,0.12797120094299316
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,32,1,power_law_1.01,0.13078975677490234
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,32,1,power_law_1.01,0.13455552101135254
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,32,1,power_law_1.01,0.13978752136230468
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,32,1,power_law_1.01,0.14272512435913084
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,32,1,power_law_1.01,0.1583174419403076
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,32,1,power_law_1.01,0.17389823913574218
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,32,1,power_law_1.01,0.2092255973815918
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,32,1,power_law_1.01,0.2746566390991211
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,32,1,power_law_1.01,0.373135986328125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,32,1,power_law_1.01,0.4703539276123047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,32,1,power_law_1.2,0.27532928466796874
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,32,1,power_law_1.01,0.658123550415039
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,32,1,power_law_1.01,0.8393465423583985
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,32,1,power_law_1.01,1.2168370819091796
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,32,1,power_law_1.01,1.5822431945800781
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,32,1,power_law_1.2,0.3634854507446289
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,32,1,power_law_1.01,0.04046080112457275
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,32,1,power_law_1.01,0.043587198257446294
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,32,1,power_law_1.01,0.04437888145446777
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,32,1,power_law_1.01,0.037025918960571294
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,32,1,power_law_1.01,0.053047041893005374
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,32,1,power_law_1.01,0.05341184139251709
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,32,1,power_law_1.01,0.05449535846710205
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,32,1,power_law_1.01,0.054611201286315914
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,32,1,power_law_1.01,0.05552192211151123
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,32,1,power_law_1.01,0.055840640068054205
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,32,1,power_law_1.01,0.05780608177185058
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,32,1,power_law_1.01,0.05845183849334716
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,32,1,power_law_1.01,0.06104512214660644
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,32,1,power_law_1.01,0.06322879791259765
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,32,1,power_law_1.01,0.06607359886169434
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,32,1,power_law_1.01,0.06968704223632813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,32,1,power_law_1.01,0.07163072109222413
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,32,1,power_law_1.01,0.08382975578308105
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,32,1,power_law_1.01,0.09599679946899414
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,32,1,power_law_1.01,0.11499839782714845
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,32,1,power_law_1.01,0.1333043193817139
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,32,1,power_law_1.01,0.19556287765502928
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,32,1,power_law_1.01,0.2394438362121582
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,32,1,power_law_1.01,0.3419756698608398
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,32,1,power_law_1.01,0.4454323196411133
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,32,1,power_law_1.2,0.4513267135620117
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,32,1,power_law_1.01,1.9576576232910157
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,32,1,power_law_1.01,0.6469273376464844
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,32,1,power_law_1.01,0.019423359632492067
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,32,1,power_law_1.01,0.0209881591796875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,32,1,power_law_1.01,0.022620799541473387
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,32,1,power_law_1.01,0.0227622389793396
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,32,1,power_law_1.01,0.023057279586791993
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,32,1,power_law_1.01,0.02321727991104126
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,32,1,power_law_1.01,0.023361918926239015
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,32,1,power_law_1.01,0.8309024047851562
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,32,1,power_law_1.01,0.02319808006286621
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,32,1,power_law_1.01,0.02392127990722656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,32,1,power_law_1.01,0.02058624029159546
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,32,1,power_law_1.01,1.0397567749023438
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,32,1,power_law_1.01,0.02434367895126343
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,32,1,power_law_1.01,0.02345088005065918
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,32,1,power_law_1.01,0.026330881118774414
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,32,1,power_law_1.01,0.02718143939971924
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,32,1,power_law_1.01,0.029787518978118897
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,32,1,power_law_1.01,0.03499648094177246
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,32,1,power_law_1.01,0.03113343954086304
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,32,1,power_law_1.01,0.04322624206542969
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,32,1,power_law_1.01,0.05462783813476563
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,32,1,power_law_1.01,0.0676262378692627
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,32,1,power_law_1.01,0.10797887802124025
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,32,1,power_law_1.01,0.07861824035644531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,32,1,power_law_1.01,0.1374067211151123
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,32,1,power_law_1.01,0.1958438491821289
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,32,1,power_law_1.01,0.25538623809814454
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,32,1,power_law_1.2,0.055968642234802246
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,32,1,power_law_1.2,0.0608403205871582
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,32,1,power_law_1.2,0.061413760185241695
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,32,1,power_law_1.2,0.07136896133422851
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,32,1,power_law_1.2,0.11345472335815429
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,32,1,power_law_1.2,0.1145907211303711
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,32,1,power_law_1.2,0.11540351867675783
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,32,1,power_law_1.2,0.11630975723266601
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,32,1,power_law_1.2,0.11768896102905273
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,32,1,power_law_1.2,0.11841728210449218
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,32,1,power_law_1.2,0.12213055610656738
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,32,1,power_law_1.2,0.12497471809387206
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,32,1,power_law_1.2,0.12801216125488282
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,32,1,power_law_1.2,0.13136256217956543
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,32,1,power_law_1.2,0.13462528228759765
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,32,1,power_law_1.2,0.14053888320922853
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,32,1,power_law_1.2,0.1422489643096924
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,32,1,power_law_1.2,0.1581100845336914
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,32,1,power_law_1.01,0.3735308837890625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,32,1,power_law_1.2,0.17699264526367187
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,32,1,power_law_1.2,0.2091571235656738
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,32,1,power_law_1.2,0.27711488723754885
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,32,1,power_law_1.2,0.3734259033203125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,32,1,power_law_1.01,0.49329792022705077
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,32,1,power_law_1.2,0.47216320037841797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,32,1,power_law_1.2,0.6546002960205077
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,32,1,power_law_1.01,0.6152959823608398
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,32,1,power_law_1.2,0.8461650848388672
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,32,1,power_law_1.2,0.036563839912414554
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,32,1,power_law_1.2,0.039827840328216554
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,32,1,power_law_1.2,0.04145664215087891
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,32,1,power_law_1.2,0.04364160060882568
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,32,1,power_law_1.2,0.05225791931152344
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,32,1,power_law_1.2,0.053002238273620605
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,32,1,power_law_1.2,0.05384255886077881
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,32,1,power_law_1.2,0.05457215785980225
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,32,1,power_law_1.2,0.05499392032623292
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,32,1,power_law_1.2,0.055395197868347165
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,32,1,power_law_1.2,0.05709695816040039
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,32,1,power_law_1.2,0.05790272235870362
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,32,1,power_law_1.2,0.05997504234313965
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,32,1,power_law_1.2,0.06268735885620116
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,32,1,power_law_1.2,0.06687744140625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,32,1,power_law_1.2,0.07011007785797119
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,32,1,power_law_1.2,0.07427840232849121
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,32,1,power_law_1.2,0.08364031791687013
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,32,1,power_law_1.2,0.09610752105712891
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,32,1,power_law_1.2,0.11480128288269043
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,32,1,power_law_1.2,1.217281951904297
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,32,1,power_law_1.2,0.13463104248046875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,32,1,power_law_1.2,0.19602048873901368
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,32,1,power_law_1.2,0.23882816314697264
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,32,1,power_law_1.2,0.3437126541137695
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,32,1,power_law_1.2,1.5824517822265625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,32,1,power_law_1.2,0.44706558227539067
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,32,1,power_law_1.2,1.9586585998535155
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,32,1,power_law_1.2,0.019378559589385985
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,32,1,power_law_1.2,0.020599040985107422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,32,1,power_law_1.2,0.020972158908843994
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,32,1,power_law_1.2,0.022613120079040528
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,32,1,power_law_1.2,0.022252800464630126
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,32,1,power_law_1.2,0.02302783966064453
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,32,1,power_law_1.2,0.022687358856201174
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,32,1,power_law_1.2,0.02303807973861694
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,32,1,power_law_1.2,0.0230732798576355
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,32,1,power_law_1.2,0.023510398864746093
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,32,1,power_law_1.2,0.02348992109298706
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,32,1,power_law_1.2,0.024995839595794676
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,32,1,power_law_1.2,0.026207358837127687
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,32,1,power_law_1.2,0.02758527994155884
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,32,1,power_law_1.2,0.649534683227539
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,32,1,power_law_1.2,0.030081279277801513
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,32,1,power_law_1.2,0.8315705871582031
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,32,1,power_law_1.2,0.035460479259490967
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,32,1,power_law_1.2,0.03231231927871704
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,32,1,power_law_1.2,0.05563007831573487
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,32,1,power_law_1.2,1.0382303619384765
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,32,1,power_law_1.2,0.04381504058837891
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,32,1,power_law_1.2,0.06876863956451415
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,32,1,power_law_1.2,0.08112511634826661
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,32,1,power_law_1.2,0.11209792137145995
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,32,1,power_law_1.2,0.14184576034545898
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,32,1,power_law_1.2,0.20627264022827146
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,32,1,power_law_1.2,0.2640224075317383
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,32,1,power_law_1.01,0.04398719787597656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,32,1,power_law_1.01,0.047662081718444826
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,32,1,power_law_1.01,0.07811647891998291
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,32,1,power_law_1.01,0.09659071922302245
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,32,1,power_law_1.01,0.11934528350830079
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,32,1,power_law_1.01,0.14773119926452635
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,32,1,power_law_1.01,0.1701510429382324
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,32,1,power_law_1.01,0.1772857666015625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,32,1,power_law_1.01,0.18185600280761718
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,32,1,power_law_1.01,0.18540224075317382
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,32,1,power_law_1.01,0.1925472068786621
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,32,1,power_law_1.01,0.2051468849182129
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,32,1,power_law_1.01,0.2132172775268555
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,32,1,power_law_1.01,0.2364735984802246
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,32,1,power_law_1.01,0.3313587188720703
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,32,1,power_law_1.01,0.34060672760009764
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,32,1,power_law_1.01,0.3565599822998047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,32,1,power_law_1.01,0.3863539123535156
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,32,1,power_law_1.01,0.41559806823730466
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,32,1,power_law_1.01,0.4767007827758789
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,32,1,power_law_1.01,0.5363801574707031
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,32,1,power_law_1.01,0.700228500366211
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,32,1,power_law_1.01,0.8683430480957032
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,32,1,power_law_1.2,0.3890700912475586
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,32,1,power_law_1.01,1.1462073516845703
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,32,1,power_law_1.01,1.4377471923828125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,32,1,power_law_1.01,2.0334002685546873
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,32,1,power_law_1.01,2.6098239135742185
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,32,1,power_law_1.01,3.2275167846679684
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,32,1,power_law_1.01,0.04678463935852051
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,32,1,power_law_1.01,0.06988800048828125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,32,1,power_law_1.01,0.09797056198120117
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,32,1,power_law_1.01,0.05131968021392822
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,32,1,power_law_1.01,0.13560959815979004
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,32,1,power_law_1.01,0.1887980842590332
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,32,1,power_law_1.01,0.22176448822021486
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,32,1,power_law_1.01,0.22893823623657225
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,32,1,power_law_1.01,0.23208959579467772
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,32,1,power_law_1.01,0.23638591766357422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,32,1,power_law_1.01,0.243874568939209
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,32,1,power_law_1.01,0.2522259140014648
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,32,1,power_law_1.01,0.25467071533203123
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,32,1,power_law_1.01,0.2634092712402344
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,32,1,power_law_1.01,0.3124959945678711
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,32,1,power_law_1.01,0.3200147247314453
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,32,1,power_law_1.01,0.33004032135009764
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,32,1,power_law_1.01,0.3551980972290039
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,32,1,power_law_1.01,0.38060161590576175
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,32,1,power_law_1.01,0.43306751251220704
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,32,1,power_law_1.01,0.49946815490722657
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,32,1,power_law_1.01,0.6118143844604492
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,32,1,power_law_1.01,0.7128844451904297
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,32,1,power_law_1.01,0.9444319915771484
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,32,1,power_law_1.01,1.1800889587402343
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,32,1,power_law_1.01,1.6488301086425783
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,32,1,power_law_1.01,2.0807653808593747
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,32,1,power_law_1.01,2.550390472412109
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,32,1,power_law_1.2,0.5141446304321289
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,32,1,power_law_1.01,0.07427072048187255
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,32,1,power_law_1.01,0.10060352325439452
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,32,1,power_law_1.01,0.07754367828369141
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,32,1,power_law_1.01,0.11607872009277345
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,32,1,power_law_1.01,0.13169983863830564
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,32,1,power_law_1.01,0.16167232513427734
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,32,1,power_law_1.01,0.16849407196044922
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,32,1,power_law_1.01,0.1708723258972168
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,32,1,power_law_1.01,0.17429311752319337
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,32,1,power_law_1.01,0.18169343948364258
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,32,1,power_law_1.01,0.1892416000366211
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,32,1,power_law_1.01,0.19351743698120116
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,32,1,power_law_1.01,0.2018764877319336
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,32,1,power_law_1.01,0.21156864166259765
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,32,1,power_law_1.01,0.21956928253173827
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,32,1,power_law_1.01,0.23848960876464842
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,32,1,power_law_1.2,0.6399622344970703
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,32,1,power_law_1.01,0.08767807960510254
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,32,1,power_law_1.01,0.2811820793151855
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,32,1,power_law_1.01,0.3234431838989258
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,32,1,power_law_1.01,0.41491710662841796
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,32,1,power_law_1.01,0.5038438415527343
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,32,1,power_law_1.01,0.6750822448730469
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,32,1,power_law_1.01,0.8758457946777345
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,32,1,power_law_1.01,1.2935891723632813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,32,1,power_law_1.01,1.735382995605469
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,32,1,power_law_1.2,0.044004478454589845
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,32,1,power_law_1.2,0.0457203197479248
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,32,1,power_law_1.2,0.0786790418624878
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,32,1,power_law_1.2,0.09337151527404786
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,32,1,power_law_1.2,0.11536319732666016
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,32,1,power_law_1.2,0.14460991859436034
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,32,1,power_law_1.01,2.614837646484375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,32,1,power_law_1.2,0.16770944595336915
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,32,1,power_law_1.2,0.1753683280944824
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,32,1,power_law_1.2,0.17847808837890625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,32,1,power_law_1.2,0.19435007095336915
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,32,1,power_law_1.2,0.18571584701538085
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,32,1,power_law_1.2,0.21603776931762697
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,32,1,power_law_1.2,0.20576576232910154
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,32,1,power_law_1.2,0.24298751831054685
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,32,1,power_law_1.2,0.3344255828857422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,32,1,power_law_1.2,0.34539905548095706
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,32,1,power_law_1.2,0.35781951904296877
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,32,1,power_law_1.2,0.3918982315063476
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,32,1,power_law_1.2,0.4196249771118164
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,32,1,power_law_1.01,3.4362030029296875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,32,1,power_law_1.2,0.48464126586914064
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,32,1,power_law_1.2,0.5414720153808593
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,32,1,power_law_1.2,0.7038944244384766
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,32,1,power_law_1.2,0.8784786987304688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,32,1,power_law_1.2,1.1552236938476563
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,32,1,power_law_1.2,1.4569798278808594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,32,1,power_law_1.2,2.0484095764160157
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,7168,2048,8,256,32,1,power_law_1.01,4.153285827636719
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,32,1,power_law_1.2,0.05176064014434815
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,32,1,power_law_1.2,0.04567359924316407
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,32,1,power_law_1.2,0.07087359905242921
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,32,1,power_law_1.2,0.09310976028442383
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,32,1,power_law_1.2,2.6387405395507812
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,32,1,power_law_1.2,0.12915200233459473
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,32,1,power_law_1.2,0.1843756866455078
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,32,1,power_law_1.2,0.21892927169799803
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,32,1,power_law_1.2,0.22608320236206056
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,32,1,power_law_1.2,0.23087615966796876
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,32,1,power_law_1.2,0.23674943923950192
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,32,1,power_law_1.2,0.2416985511779785
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,32,1,power_law_1.2,3.2374969482421876
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,32,1,power_law_1.2,0.2501919937133789
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,32,1,power_law_1.2,0.25676736831665037
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,32,1,power_law_1.2,0.26579200744628906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,32,1,power_law_1.2,0.31695039749145504
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,32,1,power_law_1.2,0.3259891128540039
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,32,1,power_law_1.2,0.3355456161499023
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,32,1,power_law_1.2,0.3687923049926758
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,32,1,power_law_1.2,0.3943328094482422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,32,1,power_law_1.2,0.45406208038330076
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,32,1,power_law_1.2,0.5180441665649413
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,32,1,power_law_1.2,0.6274828720092773
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,32,1,power_law_1.2,0.7549887847900391
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,32,1,power_law_1.2,0.9555763244628906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,32,1,power_law_1.2,1.1957421112060547
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,32,1,power_law_1.2,1.6616461181640623
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,32,1,power_law_1.2,0.07467135906219483
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,32,1,power_law_1.2,2.1266514587402345
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,32,1,power_law_1.2,0.0775820779800415
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,32,1,power_law_1.2,2.56453125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,32,1,power_law_1.2,0.08598527908325196
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,32,1,power_law_1.2,0.09854847908020019
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,32,1,power_law_1.2,0.11273856163024902
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,32,1,power_law_1.2,0.1599468803405762
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,32,1,power_law_1.2,0.1664678382873535
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,32,1,power_law_1.2,0.17165056228637696
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,32,1,power_law_1.2,0.18341760635375975
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,32,1,power_law_1.2,0.17604608535766603
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,32,1,power_law_1.2,0.19077823638916017
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,32,1,power_law_1.2,0.19796863555908203
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,32,1,power_law_1.2,0.20958784103393552
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,32,1,power_law_1.2,0.22313535690307615
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,32,1,power_law_1.2,0.23296960830688476
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,32,1,power_law_1.2,0.26612480163574215
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,32,1,power_law_1.2,0.12845376014709473
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,32,1,power_law_1.2,0.3912947082519531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,32,1,power_law_1.2,0.5397612762451172
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,32,1,power_law_1.2,0.32352256774902344
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,32,1,power_law_1.2,0.6647468566894531
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,32,1,power_law_1.2,1.3329286193847656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,32,1,power_law_1.2,0.9910399627685547
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,32,1,power_law_1.2,1.9790675354003906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,32,1,power_law_1.01,0.036275839805603026
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,32,1,power_law_1.01,0.0525708818435669
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,32,1,power_law_1.01,0.05895872116088867
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,32,1,power_law_1.01,0.06694719791412354
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,32,1,power_law_1.01,0.07956416130065917
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,32,1,power_law_1.01,0.0817030429840088
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,32,1,power_law_1.01,0.08376128196716308
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,32,1,power_law_1.01,0.08735872268676757
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,32,1,power_law_1.01,0.08847999572753906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,32,1,power_law_1.01,0.09369088172912597
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,32,1,power_law_1.01,0.13675264358520506
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,32,1,power_law_1.01,0.13991616249084474
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,32,1,power_law_1.01,0.14645119667053225
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,32,1,power_law_1.01,0.1494688034057617
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,32,1,power_law_1.01,0.15392512321472168
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,32,1,power_law_1.01,0.16284671783447266
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,32,1,power_law_1.2,2.7460037231445313
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,32,1,power_law_1.01,0.1841529655456543
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,32,1,power_law_1.01,0.2381907272338867
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,32,1,power_law_1.01,0.20052736282348632
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,32,1,power_law_1.01,0.27794496536254887
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,32,1,power_law_1.01,0.3629542541503906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,32,1,power_law_1.01,0.4523142242431641
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,32,1,power_law_1.01,0.6247724914550782
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,32,1,power_law_1.01,0.7989810943603516
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,32,1,power_law_1.01,1.1486579132080077
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,32,1,power_law_1.01,1.496170196533203
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,32,1,power_law_1.01,0.036280319690704346
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,32,1,power_law_1.01,0.04084735870361328
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,32,1,power_law_1.01,0.048533120155334473
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,32,1,power_law_1.01,1.8486534118652345
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,32,1,power_law_1.01,0.05749375820159912
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,32,1,power_law_1.01,0.07226240158081054
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,32,1,power_law_1.01,0.08475135803222657
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,32,1,power_law_1.01,0.08613951683044434
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,32,1,power_law_1.01,0.08859711647033691
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,32,1,power_law_1.01,0.09144512176513672
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,32,1,power_law_1.01,0.09100864410400392
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,32,1,power_law_1.01,0.09497471809387206
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,32,1,power_law_1.01,0.11012736320495606
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,32,1,power_law_1.01,0.11287424087524414
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,32,1,power_law_1.01,0.11862527847290039
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,32,1,power_law_1.01,0.12159232139587402
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,32,1,power_law_1.01,0.12524864196777344
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,32,1,power_law_1.01,0.14951487541198732
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,32,1,power_law_1.01,0.1357196807861328
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,32,1,power_law_1.01,0.17653247833251953
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,32,1,power_law_1.01,0.2110540771484375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,32,1,power_law_1.01,0.24744384765625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,32,1,power_law_1.01,0.3124044799804687
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,32,1,power_law_1.01,0.3894611358642578
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,32,1,power_law_1.01,0.5338105773925781
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,32,1,power_law_1.01,0.6650450897216797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,32,1,power_law_1.01,0.9551974487304686
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,32,1,power_law_1.01,1.245525131225586
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,32,1,power_law_1.01,1.523498840332031
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,32,1,power_law_1.2,0.05225344181060791
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,32,1,power_law_1.2,0.056887040138244624
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,32,1,power_law_1.2,0.06525440216064453
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,32,1,power_law_1.2,0.07957759857177735
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,32,1,power_law_1.2,0.08133631706237793
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,32,1,power_law_1.2,0.08480192184448242
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,32,1,power_law_1.2,0.08800064086914063
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,32,1,power_law_1.2,0.08949248313903808
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,32,1,power_law_1.2,0.03382400035858154
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,32,1,power_law_1.2,0.09405887603759766
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,32,1,power_law_1.2,0.13385343551635742
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,32,1,power_law_1.2,0.14037440299987794
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,32,1,power_law_1.2,0.14296256065368654
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,32,1,power_law_1.2,0.14974528312683105
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,32,1,power_law_1.2,0.15439104080200194
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,32,1,power_law_1.2,0.16349119186401367
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,32,1,power_law_1.2,0.1837068748474121
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,32,1,power_law_1.2,0.2010553550720215
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,32,1,power_law_1.2,0.23862592697143556
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,32,1,power_law_1.2,0.2787251281738281
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,32,1,power_law_1.2,0.03808511972427368
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,32,1,power_law_1.2,0.3616089630126953
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,32,1,power_law_1.2,0.4605254364013672
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,32,1,power_law_1.2,0.6256556701660156
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,32,1,power_law_1.2,0.7966105651855468
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,32,1,power_law_1.2,0.03618688106536865
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,32,1,power_law_1.2,0.040687360763549806
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,32,1,power_law_1.2,0.04843071937561035
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,32,1,power_law_1.2,0.05665664196014404
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,32,1,power_law_1.2,0.07265791893005372
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,32,1,power_law_1.2,0.09087615966796875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,32,1,power_law_1.2,0.08518400192260742
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,32,1,power_law_1.2,0.08691904067993164
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,32,1,power_law_1.2,0.08911999702453613
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,32,1,power_law_1.2,0.09032192230224609
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,32,1,power_law_1.2,0.0925004768371582
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,32,1,power_law_1.2,0.11077887535095214
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,32,1,power_law_1.2,0.1130502414703369
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,32,1,power_law_1.2,0.11938943862915039
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,32,1,power_law_1.2,0.12391936302185058
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,32,1,power_law_1.2,0.12706303596496582
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,32,1,power_law_1.2,0.13646335601806642
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,32,1,power_law_1.2,0.15319168090820312
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,32,1,power_law_1.2,0.1804256057739258
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,32,1,power_law_1.2,0.21599359512329103
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,32,1,power_law_1.2,0.24927808761596681
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,32,1,power_law_1.2,0.3126233673095703
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,32,1,power_law_1.2,0.39319232940673826
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,32,1,power_law_1.2,1.1499180603027344
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,32,1,power_law_1.2,0.5351327896118164
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,32,1,power_law_1.2,0.6651769256591796
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,32,1,power_law_1.2,1.5009036254882813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,32,1,power_law_1.2,0.9505273437499999
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,32,1,power_law_1.2,1.8581272888183595
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,32,1,power_law_1.2,1.2393920135498047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,32,1,power_law_1.2,1.52114501953125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,32,1,power_law_1.01,0.043160319328308105
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,32,1,power_law_1.01,0.05762688159942627
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,32,1,power_law_1.01,0.07675199985504151
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,32,1,power_law_1.01,0.09048704147338868
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,32,1,power_law_1.01,0.10917311668395997
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,32,1,power_law_1.01,0.13569984436035157
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,32,1,power_law_1.01,0.13973952293395997
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,32,1,power_law_1.01,0.14447999954223634
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,32,1,power_law_1.01,0.14809215545654297
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,32,1,power_law_1.01,0.15222847938537598
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,32,1,power_law_1.01,0.16420671463012695
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,32,1,power_law_1.01,0.17426048278808595
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,32,1,power_law_1.01,0.23060480117797852
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,32,1,power_law_1.01,0.23918975830078124
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,32,1,power_law_1.01,0.24442560195922852
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,32,1,power_law_1.01,0.2511724853515625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,32,1,power_law_1.01,0.26627264022827146
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,32,1,power_law_1.01,0.2956224060058594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,32,1,power_law_1.01,0.3251667022705078
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,32,1,power_law_1.01,0.38720638275146485
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,32,1,power_law_1.01,0.48133312225341796
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,32,1,power_law_1.01,0.6354079818725585
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,32,1,power_law_1.01,0.7666989135742187
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,32,1,power_law_1.01,1.0613177490234373
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,32,1,power_law_1.01,1.3531622314453124
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,32,1,power_law_1.01,1.9304454040527343
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,32,1,power_law_1.01,2.5126931762695315
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,32,1,power_law_1.01,0.04648320198059082
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,32,1,power_law_1.01,3.088131103515625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,32,1,power_law_1.01,0.051813120841979976
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,32,1,power_law_1.01,0.08676735877990722
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,32,1,power_law_1.01,0.11434432029724122
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,32,1,power_law_1.01,0.07283455848693847
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,32,1,power_law_1.01,0.15251008033752442
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,32,1,power_law_1.01,0.15470911979675292
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,32,1,power_law_1.01,0.15928832054138184
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,32,1,power_law_1.01,0.16298944473266602
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,32,1,power_law_1.01,0.16479488372802736
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,32,1,power_law_1.01,0.16881088256835936
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,32,1,power_law_1.01,0.19871488571166993
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,32,1,power_law_1.01,0.1807481575012207
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,32,1,power_law_1.01,0.21520063400268558
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,32,1,power_law_1.01,0.21424127578735352
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,32,1,power_law_1.01,0.23819904327392577
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,32,1,power_law_1.01,0.20564992904663085
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,32,1,power_law_1.01,0.2609369659423828
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,32,1,power_law_1.01,0.2829747200012207
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,32,1,power_law_1.01,0.3290572738647461
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,32,1,power_law_1.01,0.38663105010986326
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,32,1,power_law_1.01,0.49617343902587885
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,32,1,power_law_1.01,0.6007865524291992
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,32,1,power_law_1.01,0.8233452606201173
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,32,1,power_law_1.01,1.0387225341796875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,32,1,power_law_1.01,1.4514796447753906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,32,1,power_law_1.01,1.8938092041015626
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,32,1,power_law_1.01,2.3248704528808597
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,32,1,power_law_1.2,0.04595200061798096
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,32,1,power_law_1.2,0.06890880107879639
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,32,1,power_law_1.2,0.07603328227996826
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,32,1,power_law_1.2,0.08691840171813965
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,32,1,power_law_1.2,0.10579584121704103
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,32,1,power_law_1.2,0.13548864364624022
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,32,1,power_law_1.2,0.1393702411651611
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,32,1,power_law_1.2,0.14615232467651368
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,32,1,power_law_1.2,0.1504089641571045
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,32,1,power_law_1.2,0.1568051242828369
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,32,1,power_law_1.2,0.16678207397460937
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,32,1,power_law_1.2,0.17710016250610353
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,32,1,power_law_1.2,0.23317951202392578
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,32,1,power_law_1.2,0.2412428855895996
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,32,1,power_law_1.2,0.24592639923095705
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,32,1,power_law_1.2,0.2526016044616699
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,32,1,power_law_1.2,0.2973427200317383
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,32,1,power_law_1.2,0.2668934440612793
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,32,1,power_law_1.2,0.3260812759399414
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,32,1,power_law_1.2,0.4018463897705078
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,32,1,power_law_1.2,0.48248512268066407
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,32,1,power_law_1.2,0.6367692947387695
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,32,1,power_law_1.2,0.7743315124511719
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,32,1,power_law_1.2,1.0610124969482422
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,32,1,power_law_1.2,1.356615753173828
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,32,1,power_law_1.2,0.04641471862792969
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,32,1,power_law_1.2,0.05473983764648438
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,32,1,power_law_1.2,0.07022016048431397
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,32,1,power_law_1.2,0.08419391632080078
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,32,1,power_law_1.2,0.11100095748901366
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,32,1,power_law_1.2,0.15146368026733398
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,32,1,power_law_1.2,0.15481216430664063
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,32,1,power_law_1.2,0.16032320022583008
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,32,1,power_law_1.2,1.942938232421875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,32,1,power_law_1.2,0.16636991500854492
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,32,1,power_law_1.2,0.16761663436889648
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,32,1,power_law_1.2,0.17455999374389647
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,32,1,power_law_1.2,0.17977088928222656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,32,1,power_law_1.2,0.2162944030761719
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,32,1,power_law_1.2,0.2166752052307129
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,32,1,power_law_1.2,0.2275775909423828
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,32,1,power_law_1.2,0.23460927963256836
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,32,1,power_law_1.2,0.2468435287475586
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,32,1,power_law_1.2,0.2665164756774902
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,32,1,power_law_1.2,0.2893721580505371
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,32,1,power_law_1.2,0.33974590301513674
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,32,1,power_law_1.2,0.39074241638183593
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,32,1,power_law_1.2,0.5155648040771484
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,32,1,power_law_1.2,0.6065497589111328
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,32,1,power_law_1.2,0.8330150604248047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,32,1,power_law_1.2,1.0414873504638673
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,32,1,power_law_1.2,2.512750701904297
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,32,1,power_law_1.2,1.4641165161132812
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,32,1,power_law_1.2,3.1075540161132813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,32,1,power_law_1.2,1.8969190979003905
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,32,1,power_law_1.2,2.3263218688964846
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,32,1,power_law_1.01,0.046419200897216795
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,32,1,power_law_1.01,0.052172160148620604
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,32,1,power_law_1.01,0.07699007987976074
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,32,1,power_law_1.01,0.10241984367370605
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,32,1,power_law_1.01,0.13483519554138185
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,32,1,power_law_1.01,0.17869440078735352
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,32,1,power_law_1.01,0.19293632507324218
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,32,1,power_law_1.01,0.22183296203613284
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,32,1,power_law_1.01,0.23418624877929686
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,32,1,power_law_1.01,0.24090303421020506
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,32,1,power_law_1.01,0.2501612854003906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,32,1,power_law_1.01,0.2543212890625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,32,1,power_law_1.01,0.2722163200378418
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,32,1,power_law_1.01,0.2954636764526367
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,32,1,power_law_1.01,0.31769088745117186
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,32,1,power_law_1.01,0.33383167266845704
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,32,1,power_law_1.01,0.4491059112548828
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,32,1,power_law_1.01,0.48060031890869137
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,32,1,power_law_1.01,0.5129625701904297
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,32,1,power_law_1.01,0.5784339141845704
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,32,1,power_law_1.01,0.6550924682617187
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,32,1,power_law_1.01,0.8600134277343751
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,32,1,power_law_1.01,0.9840300750732421
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,32,1,power_law_1.01,1.2892562866210937
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,32,1,power_law_1.01,1.5932505798339842
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,32,1,power_law_1.01,2.2129254150390625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,32,1,power_law_1.01,0.04894207954406739
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,32,1,power_law_1.01,0.055036802291870114
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,32,1,power_law_1.01,0.06625472068786621
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,32,1,power_law_1.01,0.10045439720153808
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,32,1,power_law_1.01,2.8370623779296875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,32,1,power_law_1.01,0.15030271530151368
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,32,1,power_law_1.01,0.21540992736816406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,32,1,power_law_1.01,3.4367129516601564
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,32,1,power_law_1.01,0.2645484733581543
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,32,1,power_law_1.01,0.3011155128479004
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,32,1,power_law_1.01,0.30972095489501955
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,32,1,power_law_1.01,0.3188678359985352
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,32,1,power_law_1.01,0.3272147369384766
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,32,1,power_law_1.01,0.33354686737060546
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,32,1,power_law_1.01,0.3446227264404297
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,32,1,power_law_1.01,0.3625689697265625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,32,1,power_law_1.01,0.365486068725586
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,32,1,power_law_1.01,0.381130256652832
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,32,1,power_law_1.01,0.4531788635253906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,32,1,power_law_1.01,0.4749081420898437
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,32,1,power_law_1.01,0.49986625671386714
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,32,1,power_law_1.01,0.5582112121582031
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,32,1,power_law_1.01,0.6102732849121094
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,32,1,power_law_1.01,0.7328710174560547
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,32,1,power_law_1.01,0.8603884887695312
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,32,1,power_law_1.01,1.1055136108398438
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,32,1,power_law_1.01,1.3429452514648437
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,32,1,power_law_1.01,2.3077293395996095
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,32,1,power_law_1.01,2.7828640747070312
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,32,1,power_law_1.01,0.0939136028289795
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,32,1,power_law_1.01,0.09384256362915039
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,32,1,power_law_1.01,0.0960211181640625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,32,1,power_law_1.01,0.09479040145874024
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,32,1,power_law_1.01,0.09922368049621581
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,32,1,power_law_1.01,0.10610752105712891
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,32,1,power_law_1.01,0.11273664474487304
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,32,1,power_law_1.01,0.11976192474365235
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,32,1,power_law_1.01,0.12605631828308106
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,32,1,power_law_1.01,0.1329203224182129
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,32,1,power_law_1.01,0.14619968414306642
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,32,1,power_law_1.01,0.1696294403076172
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,32,1,power_law_1.01,0.19258304595947268
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,32,1,power_law_1.01,0.2306707191467285
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,32,1,power_law_1.01,0.29593023300170895
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,32,1,power_law_1.01,0.33785022735595704
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,32,1,power_law_1.01,0.4311231994628907
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,32,1,power_law_1.01,0.6004134368896484
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,32,1,power_law_1.01,0.7727442932128906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,32,1,power_law_1.01,1.1114943695068358
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,32,1,power_law_1.01,1.4521658325195312
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,32,1,power_law_1.01,2.1443379211425784
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,32,1,power_law_1.01,2.8363909912109375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,32,1,power_law_1.2,0.04764863967895508
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,32,1,power_law_1.2,0.04977344036102295
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,32,1,power_law_1.2,0.08033599853515624
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,32,1,power_law_1.2,0.09904064178466797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,32,1,power_law_1.2,0.12639231681823732
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,32,1,power_law_1.01,4.238698120117188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,32,1,power_law_1.2,0.15976192474365233
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,32,1,power_law_1.2,0.19251007080078125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,32,1,power_law_1.2,0.2214451217651367
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,32,1,power_law_1.2,0.22849536895751954
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,32,1,power_law_1.2,0.2372243118286133
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,32,1,power_law_1.2,0.24261247634887698
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,32,1,power_law_1.2,0.2629145622253418
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,32,1,power_law_1.2,0.2703462409973144
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,32,1,power_law_1.2,0.3009779167175293
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,32,1,power_law_1.2,0.3220345687866211
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,32,1,power_law_1.2,0.3378886413574219
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,32,1,power_law_1.2,0.4497248077392578
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,32,1,power_law_1.2,0.4847155380249023
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,32,1,power_law_1.2,0.5143699264526367
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,32,1,power_law_1.2,0.5897068786621094
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,32,1,power_law_1.2,0.6719468688964844
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,32,1,power_law_1.2,0.8643859100341796
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,32,1,power_law_1.2,0.9922656249999999
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,32,1,power_law_1.2,1.3017619323730467
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,32,1,power_law_1.2,1.59833984375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,32,1,power_law_1.2,2.2248915100097655
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,32,1,power_law_1.2,2.8343966674804686
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,32,1,power_law_1.2,0.0493446397781372
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,32,1,power_law_1.2,0.05537471771240234
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,32,1,power_law_1.2,0.07184063911437988
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,32,1,power_law_1.2,0.10334207534790038
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,32,1,power_law_1.2,0.14910976409912108
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,32,1,power_law_1.2,3.481155700683594
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,32,1,power_law_1.2,0.20776704788208006
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,32,1,power_law_1.2,0.2545523262023926
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,32,1,power_law_1.2,0.2966060829162598
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,32,1,power_law_1.2,0.3057088088989258
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,32,1,power_law_1.2,0.3120780754089355
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,32,1,power_law_1.2,0.3214944076538086
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,32,1,power_law_1.2,0.3309497451782227
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,32,1,power_law_1.2,0.34199871063232423
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,32,1,power_law_1.2,0.359156494140625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,32,1,power_law_1.2,0.36552574157714846
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,32,1,power_law_1.2,0.37876670837402343
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,32,1,power_law_1.2,0.45104896545410156
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,32,1,power_law_1.2,0.482611198425293
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,32,1,power_law_1.2,0.5110240173339844
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,32,1,power_law_1.2,0.5698015975952149
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,32,1,power_law_1.2,0.6323859024047851
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,32,1,power_law_1.2,0.769793930053711
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,32,1,power_law_1.2,0.8845638275146485
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,32,1,power_law_1.2,1.1271788787841797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,32,1,power_law_1.2,1.375810546875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,32,1,power_law_1.2,1.8488365173339845
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,32,1,power_law_1.2,2.333684539794922
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,32,1,power_law_1.2,0.09384896278381347
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,32,1,power_law_1.2,2.809570617675781
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,32,1,power_law_1.2,0.09393856048583984
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,32,1,power_law_1.2,0.09474431991577148
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,32,1,power_law_1.2,0.09591808319091796
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,32,1,power_law_1.2,0.09944512367248534
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,32,1,power_law_1.2,0.10610879898071288
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,32,1,power_law_1.2,0.11273088455200195
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,32,1,power_law_1.2,0.11937408447265625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,32,1,power_law_1.2,0.12585536003112793
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,32,1,power_law_1.2,0.13324864387512209
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,32,1,power_law_1.2,0.14618240356445314
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,32,1,power_law_1.2,0.17511039733886719
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,32,1,power_law_1.2,0.18666175842285154
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,32,1,power_law_1.2,0.22865983963012698
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,32,1,power_law_1.2,0.29743488311767574
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,32,1,power_law_1.2,0.33793792724609373
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,32,1,power_law_1.2,0.432184944152832
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,32,1,power_law_1.2,0.6020633697509765
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,32,1,power_law_1.2,0.7702617645263672
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,32,1,power_law_1.2,1.1098777770996093
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,32,1,power_law_1.2,1.4521087646484374
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,32,1,power_law_1.2,2.1468844604492188
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,32,1,power_law_1.2,2.829980163574219
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,32,1,power_law_1.01,0.027596800327301024
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,32,1,power_law_1.2,4.237530212402343
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,32,1,power_law_1.01,0.03257535934448243
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,32,1,power_law_1.01,0.05830016136169434
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,32,1,power_law_1.01,0.06983424186706542
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,32,1,power_law_1.01,0.042269439697265626
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,32,1,power_law_1.01,0.09113535881042481
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,32,1,power_law_1.01,0.09796735763549805
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,32,1,power_law_1.01,0.10195455551147461
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,32,1,power_law_1.01,0.1061894416809082
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,32,1,power_law_1.01,0.10900863647460937
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,32,1,power_law_1.01,0.11763968467712402
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,32,1,power_law_1.01,0.13359040260314942
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,32,1,power_law_1.01,0.13870464324951173
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,32,1,power_law_1.01,0.13966848373413085
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,32,1,power_law_1.01,0.14431872367858886
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,32,1,power_law_1.01,0.15233023643493654
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,32,1,power_law_1.01,0.17009727478027342
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,32,1,power_law_1.01,0.1876198387145996
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,32,1,power_law_1.01,0.1309331226348877
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,32,1,power_law_1.01,0.24037567138671875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,32,1,power_law_1.01,0.28388864517211915
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,32,1,power_law_1.01,0.34718463897705076
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,32,1,power_law_1.01,0.4376614379882812
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,32,1,power_law_1.01,0.5730912017822265
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,32,1,power_law_1.01,0.7335948944091797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,32,1,power_law_1.01,1.054300765991211
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,32,1,power_law_1.01,1.3733094787597655
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,2688,1856,6,128,32,1,power_law_1.01,1.6778854370117187
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,32,1,power_law_1.2,0.03578815937042236
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,32,1,power_law_1.2,0.04399680137634278
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,32,1,power_law_1.2,0.05544000148773194
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,32,1,power_law_1.2,0.06763135910034179
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,32,1,power_law_1.2,0.09167424201965332
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,32,1,power_law_1.2,0.09809663772583008
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,32,1,power_law_1.2,0.10345024108886718
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,32,1,power_law_1.2,0.10719807624816895
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,32,1,power_law_1.2,0.11205951690673828
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,32,1,power_law_1.2,0.11957247734069824
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,32,1,power_law_1.2,0.029679360389709475
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,32,1,power_law_1.2,0.13268544197082519
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,32,1,power_law_1.2,0.13620479583740236
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,32,1,power_law_1.2,0.14124416351318358
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,32,1,power_law_1.2,0.1431379222869873
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,32,1,power_law_1.2,0.1468876838684082
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,32,1,power_law_1.2,0.15496831893920898
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,32,1,power_law_1.2,0.17231807708740235
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,32,1,power_law_1.2,0.189366397857666
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,32,1,power_law_1.2,0.2502239990234375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,32,1,power_law_1.2,0.2844588851928711
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,32,1,power_law_1.2,0.3499103927612305
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,32,1,power_law_1.2,0.43432064056396485
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,32,1,power_law_1.2,0.5747238540649414
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,32,1,power_law_1.2,0.7334381103515625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,32,1,power_law_1.2,1.0572102355957032
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,32,1,power_law_1.2,1.3806796264648438
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,2688,1856,6,128,32,1,power_law_1.2,1.6862489318847658
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,32,1,power_law_1.01,0.06039103984832763
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,32,1,power_law_1.01,0.09641535758972168
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,32,1,power_law_1.01,0.1505990409851074
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,32,1,power_law_1.01,0.2281305694580078
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,32,1,power_law_1.01,0.5060851287841797
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,32,1,power_law_1.01,0.5663897705078125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,32,1,power_law_1.01,0.5900479888916015
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,32,1,power_law_1.01,0.5344198226928711
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,32,1,power_law_1.01,0.6074886322021484
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,32,1,power_law_1.01,0.6433081817626953
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,32,1,power_law_1.01,0.7038566589355468
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,32,1,power_law_1.01,0.7431807708740235
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,32,1,power_law_1.01,0.8651821136474609
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,32,1,power_law_1.01,0.9242189025878906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,32,1,power_law_1.01,1.0343775939941406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,32,1,power_law_1.01,1.2171193695068359
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,32,1,power_law_1.01,0.31881343841552734
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,32,1,power_law_1.01,0.9867385864257813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,32,1,power_law_1.01,1.076468505859375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,32,1,power_law_1.01,1.3282675170898437
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,32,1,power_law_1.01,2.1109368896484377
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,32,1,power_law_1.01,2.5800326538085936
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,32,1,power_law_1.01,1.5793171691894532
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,32,1,power_law_1.01,3.6138763427734375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,32,1,power_law_1.01,4.616569519042969
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,32,1,power_law_1.01,6.644660034179688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,32,1,power_law_1.01,8.699760131835937
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,2688,22,512,32,1,power_law_1.01,10.80223876953125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,32,1,power_law_1.01,0.40294719696044917
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,32,1,power_law_1.01,1.2077484893798829
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,32,1,power_law_1.01,1.5514341735839845
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,32,1,power_law_1.01,0.2489516830444336
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,32,1,power_law_1.01,2.8549383544921874
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,32,1,power_law_1.01,2.7767346191406252
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,32,1,power_law_1.01,3.2688775634765626
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,32,1,power_law_1.01,2.830513916015625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,32,1,power_law_1.01,2.9932415771484377
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,32,1,power_law_1.01,0.6293529510498047
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,32,1,power_law_1.01,3.387343444824219
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,32,1,power_law_1.01,3.1482226562499998
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,32,1,power_law_1.01,3.2831405639648437
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,32,1,power_law_1.01,3.530611877441406
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,32,1,power_law_1.01,3.533077697753906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,32,1,power_law_1.01,3.6211883544921877
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,32,1,power_law_1.01,5.523494262695313
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,32,1,power_law_1.01,4.839936523437499
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,32,1,power_law_1.01,6.718048706054688
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,32,1,power_law_1.01,3.0389337158203125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,32,1,power_law_1.01,7.6359130859375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,32,1,power_law_1.01,7.728136596679687
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,32,1,power_law_1.01,10.990943603515625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,32,1,power_law_1.01,18.57846923828125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,32,1,power_law_1.01,20.51652587890625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,32,1,power_law_1.01,32.558134765625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,32,1,power_law_1.01,11.81815185546875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,32,1,power_law_1.2,0.06116608142852783
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,32,1,power_law_1.2,0.10043071746826171
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,32,1,power_law_1.2,0.1500038433074951
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,32,1,power_law_1.2,0.21455360412597657
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,32,1,power_law_1.2,0.5441580963134766
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,32,1,power_law_1.2,0.5605184173583984
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,32,1,power_law_1.2,0.29264575958251954
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,32,1,power_law_1.2,0.5909094238281251
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,32,1,power_law_1.2,0.5020409774780273
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,32,1,power_law_1.2,0.6669356536865234
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,32,1,power_law_1.2,0.6169823837280274
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,32,1,power_law_1.2,0.7272268676757813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,32,1,power_law_1.2,0.7800096130371094
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,32,1,power_law_1.2,0.8837696075439453
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,32,1,power_law_1.2,0.966192626953125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,32,1,power_law_1.2,1.0630790710449218
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,32,1,power_law_1.2,1.2458534240722656
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,32,1,power_law_1.2,1.0075494384765624
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,32,1,power_law_1.2,1.1064249420166017
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,2688,22,512,32,1,power_law_1.01,41.0838623046875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,32,1,power_law_1.2,1.3698643493652345
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,32,1,power_law_1.2,1.6311859130859374
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,32,1,power_law_1.2,2.126367340087891
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,32,1,power_law_1.2,2.6088339233398434
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,32,1,power_law_1.2,3.6264993286132814
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,32,1,power_law_1.2,4.634810791015625
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,32,1,power_law_1.2,0.23753215789794924
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,32,1,power_law_1.2,0.36267967224121095
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,32,1,power_law_1.2,0.6543007659912109
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,32,1,power_law_1.2,6.723301391601562
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,32,1,power_law_1.2,1.6244557189941407
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,32,1,power_law_1.2,2.8270309448242186
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,32,1,power_law_1.2,2.9332135009765627
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,32,1,power_law_1.2,3.0392474365234374
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,32,1,power_law_1.2,0.9585958099365234
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,32,1,power_law_1.2,3.139075927734375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,32,1,power_law_1.2,2.9335647583007813
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,32,1,power_law_1.2,3.328663635253906
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,32,1,power_law_1.2,3.0963214111328123
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,32,1,power_law_1.2,3.4445874023437497
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,32,1,power_law_1.2,3.3400082397460933
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,32,1,power_law_1.2,8.785941162109376
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,32,1,power_law_1.2,3.7874270629882814
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,32,1,power_law_1.2,3.5639602661132814
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,32,1,power_law_1.2,4.577673034667969
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,32,1,power_law_1.2,5.019734497070313
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,32,1,power_law_1.2,5.847569580078125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,32,1,power_law_1.2,7.097124633789062
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,32,1,power_law_1.2,6.382416381835937
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,32,1,power_law_1.2,9.586696166992187
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,32,1,power_law_1.2,11.49183349609375
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,32,1,power_law_1.2,11.98477294921875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,float16,20480,4096,2688,22,512,32,1,power_law_1.2,10.829461669921875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,32,1,power_law_1.2,18.379696044921875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,32,1,power_law_1.2,26.38796142578125
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,32,1,power_law_1.2,25.7728466796875
SGLang,0.5.8.post1,NVIDIA GB200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,2688,22,512,32,1,power_law_1.2,31.76097412109375
